Adaptive noise cancellation

0Associated
Cases 
0Associated
Defendants 
0Accused
Products 
118Forward
Citations 
0
Petitions 
3
Assignments
First Claim
1. A method for removing impulse noise from a circuit electrical signal s(t) that includes a desired signal s.sub.0 (t) plus a noise signal n(t), where each component may vary with time t, the method comprising the steps of:
 receiving a circuit input signal that is a sum of a desired signal s.sub.0 (t) plus a noise signal n(t) that includes impulse noise from a circuit electrical signal, where the noise signal n(t) is to be determined;
passing the circuit input signal through a bandpass filter, with a selected frequency passband PB, to form a bandpassed signal;
passing the bandpassed signal through a first selected signal processing channel that has frequency support contained in a first frequency band having a central frequency .omega..sub.1, to produce a first noise signal n.sub.1 (t);
passing the bandpassed signal through a second selected signal processing channel that has frequency support contained in a second frequency band, having a central frequency .omega..sub.2, that is spaced apart from and does not overlap any part of the first frequency band, to produce a second noise signal n.sub.2 (t);
passing the bandpassed signal through a third selected signal processing channel that has a third frequency band, having a central frequency .omega..sub.3, that is spaced apart from and does not overlap any part of the first frequency band and the second frequency band, to produce a third channel output signal s(t)+n.sub.3 (t), where the third channel has frequency support contained in the third frequency band that includes substantially all frequencies that contribute to the desired signal s(t), and where the frequencies .omega..sub.1, .omega..sub.2 and .omega..sub.3 lie in the selected frequency pass band PB;
where the first, second and third selected channels include frequency shift by selected first, second and third shift frequencies, respectively, and include passage through a low pass filter with selected first, second and third rolloff frequencies, respectively, to form first, second and third channel output signals, respectively;
forming a combined signal that is a symmetric, homogeneous function HS{n.sub.1 (t), n.sub.2 (t);
.theta.} of degree one from the first and second channel output signals, where the first and second channel output signals have m adjustable parameters .theta.=(.theta.1, . . . ,.theta.m;
m=1 or
2);
forming a linear combination signal LC{HS{n.sub.1 (t), n.sub.2 (t);
.theta.};
s(t)+n.sub.3 (t)} of the combined signal HS{n.sub.1 (t), n.sub.2 (t);
.theta.} and the third channel output signal s(t)+n.sub.3 (t);
adjusting the parameters .theta. whereby the linear combination signal LC{HS{n.sub.1 (t), n.sub.2 (t);
.theta.};
s(t)+n.sub.3 (t)} provides a best possible estimate of the desired signal s(t), according to a selected criterion, for an optimal choice .theta.=.theta..sub.0 of these parameters; and
issuing the linear combination signal LC{HS{n.sub.1 (t), n.sub.2 (t);
.theta.=.theta..sub.0 };
s(t)+n.sub.3 (t)} as an estimate of the desired signal s(t).
3 Assignments
0 Petitions
Accused Products
Abstract
Method and apparatus for reducing or cancelling impulse noise from a signal containing noise. The desired noisefree signal is assumed to have a representative frequency .omega..sub.3, but may have a range of frequencies adjacent to this frequency, and is assumed to have substantially zero amplitude for all frequencies .omega..omega..sub.2, where .omega..sub.1 <.omega..sub.3 <.omega..sub.2 or .omega..sub.1 <.omega..sub.2. An input (noisy) signal is filtered and analyzed in a narrow frequency region surrounding .omega.=.omega..sub.1 and/or a narrow frequency region surrounding .omega.=.omega..sub.2 to obtain one or two output signal components n.sub.1 (t) and/or n.sub.2 (t), respectively, that, ideally, contain no contribution from the desired signal. The input signal is also filtered and analyzed in a narrow frequency region surrounding .omega.=.omega..sub.3 to obtain an output signal s(t)+n.sub.3 (t) component including the desired signal s(t). A linear combination of signals, such as S(t)=s(t)+n.sub.3 (t).+.[n.sub.1 (t)n.sub.2 (1)] .sup.1/2, or S(t)=s(t)+n.sub.3 (t)+.alpha.exp(j.phi.) [n.sub.1 (t)n.sub.2 (1)].sup.1/2,or S(t)=s(t)+n.sub.3 (t).+.n.sub.k (t) (k=1,2), or S(t)=s(t)+n.sub.3 (t)+exp(j.phi.)n.sub.k (t), or S(t)=s(t)+n.sub.3 (t)+[exp(j.psi.1)n.sub.1 (t)+exp(j.psi.2) n.sub.2 (t)]/2, or S(t)=s(t)+n.sub.3 (t)+.alpha.[exp(j.psi.) n.sub.1 (t)+exp(j.psi.) n.sub.2 (t)]/2, is formed as a circuit output signal, and the .+. sign and/or the multiplier .alpha. and/or the phase angles .phi., .psi.1, .psi.2 and .psi. are chosen to minimize the contribution of noise to S(t), according to a selected quantitative error measure. Three such error measures are the displacement of a signal from the median of a reference signal,the cumulative variation of a signal, and the least mean square value of a signal. Other combinations of the signals n.sub.1 (t) and n.sub.2 (t), replacing the (complex) arithmetic mean and geometric mean, may be used for the linear combination signal S(t). Methods for computation of the phase angles .phi. , .psi.1, .psi.2 and .psi. are also disclosed.
134 Citations
REJECTION OF A CLOSEINFREQUENCY INTERFERER EMPLOYING A LOG DETECTOR AND CLASSICAL DOWN CONVERTER  
Patent #
US 20110053514A1
Filed 01/09/2009

Current Assignee
Clemson University

Sponsoring Entity
Clemson University

Pseudorange calculation method, position calculation method, computerreadable recording medium, and position calculation device  
Patent #
US 7,940,209 B2
Filed 04/02/2009

Current Assignee
Seiko Epson Corporation

Sponsoring Entity
Seiko Epson Corporation

RESOLUTION ENHANCEMENT SYSTEM (RES) FOR NETWORKED RADARS  
Patent #
US 20110102249A1
Filed 10/20/2010

Current Assignee
Colorado State University Research Foundation

Sponsoring Entity
Colorado State University Research Foundation

SENSITIVITY ENHANCEMENT SYSTEM  
Patent #
US 20110102250A1
Filed 10/20/2010

Current Assignee
Colorado State University Research Foundation

Sponsoring Entity
Colorado State University Research Foundation

CANCELLATION OF PILOT AND TRAFFIC SIGNALS  
Patent #
US 20100260238A1
Filed 06/25/2010

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Impulse detection and reduction in a frequency modulation radio receiver  
Patent #
US 7,835,468 B2
Filed 03/13/2006

Current Assignee
Silicon Laboratories Incorporated

Sponsoring Entity
Silicon Laboratories Incorporated

Cancellation of pilot and traffic signals  
Patent #
US 7,751,465 B2
Filed 09/23/2005

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Antenna selection system and method  
Patent #
US 7,751,785 B2
Filed 04/18/2007

Current Assignee
GE Medical Systems Information Technologies Incorporated

Sponsoring Entity
GE Medical Systems Information Technologies Incorporated

Adaptive Noise Cancellation  
Patent #
US 20090012786A1
Filed 07/02/2008

Current Assignee
Texas Instruments Inc.

Sponsoring Entity
Texas Instruments Inc.

PSEUDORANGE CALCULATION METHOD, POSITION CALCULATION METHOD, COMPUTERREADABLE RECORDING MEDIUM, AND POSITION CALCULATION DEVICE  
Patent #
US 20090262019A1
Filed 04/02/2009

Current Assignee
Seiko Epson Corporation

Sponsoring Entity
Seiko Epson Corporation

Antenna selection system and method  
Patent #
US 7,248,843 B2
Filed 11/07/2003

Current Assignee
GE Medical Systems Information Technologies Incorporated

Sponsoring Entity
GE Medical Systems Information Technologies Incorporated

ANTENNA SELECTION SYSTEM AND METHOD  
Patent #
US 20070184802A1
Filed 04/18/2007

Current Assignee
GE Medical Systems Information Technologies Incorporated

Sponsoring Entity
GE Medical Systems Information Technologies Incorporated

Impulse detection and reduction in a frequency modulation radio receiver  
Patent #
US 20070211830A1
Filed 03/13/2006

Current Assignee
Silicon Laboratories Incorporated

Sponsoring Entity
Silicon Laboratories Incorporated

Triple multiplexing spread spectrum receiver  
Patent #
US 7,295,633 B2
Filed 07/30/2004

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

Multipath processing for GPS receivers  
Patent #
US 7,301,992 B2
Filed 03/30/2004

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

Strong signal cancellation to enhance processing of weak spread spectrum signal  
Patent #
US 7,116,704 B2
Filed 11/12/2003

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

GPS receiver with crosstrack hold  
Patent #
US 6,236,937 B1
Filed 07/31/2000

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

Method and apparatus for cancelling echo originating from a mobile terminal  
Patent #
US 6,256,384 B1
Filed 12/02/1997

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Strong signal cancellation to enhance processing of weak spread spectrum signal  
Patent #
US 6,282,231 B1
Filed 12/14/1999

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

GPS receiver with crosstrack hold  
Patent #
US 6,421,609 B2
Filed 07/12/2001

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

GPS receiver with crosstrack hold  
Patent #
US 6,125,325 A
Filed 07/25/1997

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

GPS receiver with crosstrack hold  
Patent #
US 6,292,749 B2
Filed 12/08/2000

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

GPS receiver with crosstrack hold  
Patent #
US 6,574,558 B2
Filed 05/24/2002

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

Apparatus and method for transmission of high speed data over communication channels  
Patent #
US 6,047,022 A
Filed 02/28/1997

Current Assignee
STMicroelectronics NV

Sponsoring Entity
Orckit Communications Ltd.

Cancellation of pilot and traffic signals  
Patent #
US 20060034218A1
Filed 09/23/2005

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Methods and apparatus for mitigating the effects of microphone overload in echo cancelation systems  
Patent #
US 6,850,783 B1
Filed 08/07/1998

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Antenna selection system and method  
Patent #
US 20050101252A1
Filed 11/07/2003

Current Assignee
GE Medical Systems Information Technologies Incorporated

Sponsoring Entity
GE Medical Systems Information Technologies Incorporated

Spread spectrum receiver with multipath correction  
Patent #
US 6,917,644 B2
Filed 12/11/2000

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

Multipath processing for GPS receivers  
Patent #
US 6,760,364 B2
Filed 09/18/2002

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

Triple multiplexing spread spectrum receiver  
Patent #
US 6,788,735 B2
Filed 12/16/2002

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

Triple multiplexing spread spectrum receiver  
Patent #
US 6,522,682 B1
Filed 03/02/1999

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

GPS system for navigating a vehicle  
Patent #
US 6,633,814 B2
Filed 12/22/2000

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

Spread spectrum receiver with multibit correlator  
Patent #
US 6,393,046 B1
Filed 04/25/1997

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

Pseudonoise correlator for a GPS spread spectrum receiver  
Patent #
US 6,400,753 B1
Filed 09/05/2000

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

Detection circuit of tone signal  
Patent #
US 6,415,139 B1
Filed 11/24/1998

Current Assignee
Canon Ayutthaya Limited

Sponsoring Entity
OKI Electric Industry Company Limited

Multipath processing for GPS receivers  
Patent #
US 6,466,612 B2
Filed 02/06/2001

Current Assignee
Samsung Electronics Co. Ltd.

Sponsoring Entity
SiRF Technology

Spread spectrum receiver with multipath correction  
Patent #
US 6,198,765 B1
Filed 09/12/1997

Current Assignee
Samsung Electronics Co. Ltd.

Sponsoring Entity
SiRF Technology

Spread spectrum receiver with multipath correction  
Patent #
US 20010002203A1
Filed 12/11/2000

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
CSR Technology Incorporated

Multipath processing for GPS receivers  
Patent #
US 6,249,542 B1
Filed 03/27/1998

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

Adaptive noise canceller  
Patent #
US 6,285,718 B1
Filed 02/09/2000

Current Assignee
STMicroelectronics NV

Sponsoring Entity
Tioga Technologies Incorporated

GPS receiver  
Patent #
US 6,018,704 A
Filed 07/25/1997

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

Adaptive equalization of multipath signals  
Patent #
US 6,031,882 A
Filed 05/06/1997

Current Assignee
Trimble Navigation Limited

Sponsoring Entity
Trimble Navigation Limited

GPS car navigation system  
Patent #
US 6,041,280 A
Filed 04/25/1996

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

Spread spectrum receiver with multipath cancellation  
Patent #
US 6,047,017 A
Filed 07/25/1997

Current Assignee
CSR Technology Incorporated

Sponsoring Entity
SiRF Technology

Spread spectrum receiver with fast signal reacquisition  
Patent #
US 5,897,605 A
Filed 04/25/1996

Current Assignee
CSR plc

Sponsoring Entity


Triple multiplexing spread spectrum receiver  
Patent #
US 5,901,171 A
Filed 04/25/1996

Current Assignee
CSR Technology Incorporated

Sponsoring Entity


GMSK signal processors for improved communications capacity and quality  
Patent #
US 5,848,105 A
Filed 10/10/1996

Current Assignee
Apple Inc.

Sponsoring Entity
GARDNER WILLIAM A.

Generalized noise cancellation in a communication channel  
Patent #
US 5,621,768 A
Filed 11/29/1994

Current Assignee
TTI Inventions A LLC

Sponsoring Entity
Bell Communications Research Inc.

Adaptive multipath equalization  
Patent #
US 5,630,208 A
Filed 07/19/1994

Current Assignee
Trimble Navigation Limited

Sponsoring Entity
Trimble Navigation Limited

Sensitivity enhancement system  
Patent #
US 8,274,423 B2
Filed 10/20/2010

Current Assignee
Colorado State University Research Foundation

Sponsoring Entity
Colorado State University Research Foundation

Alarm system controller and a method for controlling an alarm system  
Patent #
US 8,369,967 B2
Filed 03/07/2011

Current Assignee
HOFFBERG FAMILY TRUST 1

Sponsoring Entity
STEVEN M. HOFFBERG 20041 GRAT

Cancellation of pilot and traffic signals  
Patent #
US 8,369,385 B2
Filed 06/25/2010

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Resolution enhancement system (RES) for networked radars  
Patent #
US 8,462,040 B2
Filed 10/20/2010

Current Assignee
Colorado State University Research Foundation

Sponsoring Entity
Colorado State University Research Foundation

Cancellation of pilot and traffic signals  
Patent #
US 8,594,157 B2
Filed 01/22/2013

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Adaptive pattern recognition based controller apparatus and method and humaninterface therefore  
Patent #
US 8,892,495 B2
Filed 01/08/2013

Current Assignee
HOFFBERG FAMILY TRUST 1

Sponsoring Entity
HOFFBERG FAMILY TRUST 1

MIC COVERING DETECTION IN PERSONAL AUDIO DEVICES  
Patent #
US 20150104032A1
Filed 12/22/2014

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Noise burst adaptation of secondary path adaptive response in noisecanceling personal audio devices  
Patent #
US 9,082,387 B2
Filed 12/20/2012

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Close talk detector for noise cancellation  
Patent #
US 9,094,744 B1
Filed 12/21/2012

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Ambient noise root mean square (RMS) detector  
Patent #
US 9,107,010 B2
Filed 02/08/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Sequenced adaptation of antinoise generator response and secondary path response in an adaptive noise canceling system  
Patent #
US 9,123,321 B2
Filed 12/27/2012

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Oversight control of an adaptive noise canceler in a personal audio device  
Patent #
US 9,142,207 B2
Filed 12/01/2011

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Leakagemodeling adaptive noise canceling for earspeakers  
Patent #
US 9,142,205 B2
Filed 12/03/2012

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Mobile communication device for delivering targeted advertisements  
Patent #
US 9,151,633 B2
Filed 03/24/2014

Current Assignee
Steven M Hoffberg

Sponsoring Entity
Steven M Hoffberg

Ambient noisebased adaptation of secondary path adaptive response in noisecanceling personal audio devices  
Patent #
US 9,208,771 B2
Filed 10/25/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Continuous adaptation of secondary path adaptive response in noisecanceling personal audio devices  
Patent #
US 9,214,150 B2
Filed 04/27/2012

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Reducing an acoustic intensity vector with adaptive noise cancellation with two error microphones  
Patent #
US 9,215,749 B2
Filed 03/14/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Coordinated gain control in adaptive noise cancellation (ANC) for earspeakers  
Patent #
US 9,226,068 B2
Filed 03/12/2015

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Power management of adaptive noise cancellation (ANC) in a personal audio device  
Patent #
US 9,230,532 B1
Filed 03/12/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Systems and methods for detection and cancellation of narrowband noise  
Patent #
US 9,264,808 B2
Filed 06/14/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Systems and methods for adaptive noise cancellation including secondary path estimate monitoring  
Patent #
US 9,294,836 B2
Filed 07/26/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Frequencyshaped noisebased adaptation of secondary path adaptive response in noisecanceling personal audio devices  
Patent #
US 9,319,784 B2
Filed 04/14/2014

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Adaptive noise canceling architecture for a personal audio device  
Patent #
US 9,318,094 B2
Filed 03/07/2012

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Frequency and directiondependent ambient sound handling in personal audio devices having adaptive noise cancellation (ANC)  
Patent #
US 9,319,781 B2
Filed 03/04/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Downlink tone detection and adaptation of a secondary path response model in an adaptive noise canceling system  
Patent #
US 9,318,090 B2
Filed 12/28/2012

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Sidetone management in an adaptive noise canceling (ANC) system including secondary path modeling  
Patent #
US 9,325,821 B1
Filed 11/27/2012

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Robust adaptive noise canceling (ANC) in a personal audio device  
Patent #
US 9,324,311 B1
Filed 03/14/2014

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Bandlimiting antinoise in personal audio devices having adaptive noise cancellation (ANC)  
Patent #
US 9,368,099 B2
Filed 03/28/2014

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Frequencydependent sidetone calibration  
Patent #
US 9,369,557 B2
Filed 03/05/2014

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Internal dynamic range control in an adaptive noise cancellation (ANC) system  
Patent #
US 9,369,798 B1
Filed 03/12/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Virtual microphone for adaptive noise cancellation in personal audio devices  
Patent #
US 9,392,364 B1
Filed 08/15/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Lowlatency multidriver adaptive noise canceling (ANC) system for a personal audio device  
Patent #
US 9,414,150 B2
Filed 08/15/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Systems and methods for hybrid adaptive noise cancellation  
Patent #
US 9,462,376 B2
Filed 07/23/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Systems and methods for adaptive noise cancellation by biasing antinoise level  
Patent #
US 9,460,701 B2
Filed 07/16/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Monitoring of speaker impedance to detect pressure applied between mobile device and ear  
Patent #
US 9,467,776 B2
Filed 03/15/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Systems and methods for enhancing performance of audio transducer based on detection of transducer status  
Patent #
US 9,479,860 B2
Filed 03/07/2014

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Systems and methods for use of adaptive secondary path estimate to control equalization in an audio device  
Patent #
US 9,478,212 B1
Filed 09/03/2014

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Systems and methods for hybrid adaptive noise cancellation  
Patent #
US 9,478,210 B2
Filed 06/24/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Robust adaptive noise canceling (ANC) in a personal audio device  
Patent #
US 9,502,020 B1
Filed 03/14/2014

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Dualmicrophone frequency amplitude response selfcalibration  
Patent #
US 9,532,139 B1
Filed 12/20/2012

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Internet appliance system and method  
Patent #
US 9,535,563 B2
Filed 11/12/2013

Current Assignee
HOFFBERG FAMILY TRUST 1

Sponsoring Entity
HOFFBERG FAMILY TRUST 1

Mobile communication device  
Patent #
US 9,551,582 B2
Filed 10/05/2015

Current Assignee
Blanding Hovenweep LLC

Sponsoring Entity
Blanding Hovenweep LLC

Systems and methods for performance and stability control for feedback adaptive noise cancellation  
Patent #
US 9,552,805 B2
Filed 12/19/2014

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Hybrid adaptive noise cancellation system with filtered error microphone signal  
Patent #
US 9,578,415 B1
Filed 08/21/2015

Current Assignee
Cirrus Logic International Semiconductor Ltd.

Sponsoring Entity
Cirrus Logic International Semiconductor Ltd.

Metric and tool to evaluate secondary path design in adaptive noise cancellation systems  
Patent #
US 9,578,432 B1
Filed 04/23/2014

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Headphone responsive to optical signaling  
Patent #
US 9,609,416 B2
Filed 06/09/2014

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Systems and methods for maintaining playback fidelity in an audio system with adaptive noise cancellation  
Patent #
US 9,620,101 B1
Filed 10/08/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Oversight control of an adaptive noise canceler in a personal audio device  
Patent #
US 9,633,646 B2
Filed 08/31/2015

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Speaker impedance monitoring  
Patent #
US 9,635,480 B2
Filed 06/28/2016

Current Assignee
Cirrus Logic International Semiconductor Ltd.

Sponsoring Entity
Cirrus Logic International Semiconductor Ltd.

Earcoupling detection and adjustment of adaptive response in noisecanceling in personal audio devices  
Patent #
US 9,646,595 B2
Filed 12/09/2014

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Control of audio output of headphone earbuds based on the environment around the headphone earbuds  
Patent #
US 9,648,410 B1
Filed 03/12/2014

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

System and method for dynamic range compensation of distortion  
Patent #
US 9,654,866 B2
Filed 01/28/2013

Current Assignee
Synaptics Incorporated

Sponsoring Entity
Synaptics Incorporated

Systems and methods for adaptive noise cancellation by adaptively shaping internal white noise to train a secondary path  
Patent #
US 9,666,176 B2
Filed 09/13/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Systems and methods for sharing secondary path information between audio channels in an adaptive noise cancellation system  
Patent #
US 9,704,472 B2
Filed 12/10/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Adaptive noise canceling architecture for a personal audio device  
Patent #
US 9,711,130 B2
Filed 04/15/2016

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Downlink tone detection and adaptation of a secondary path response model in an adaptive noise canceling system  
Patent #
US 9,721,556 B2
Filed 03/15/2016

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Power management of adaptive noise cancellation (ANC) in a personal audio device  
Patent #
US 9,773,493 B1
Filed 11/23/2015

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Source audio acoustic leakage detection and management in an adaptive noise canceling system  
Patent #
US 9,773,490 B2
Filed 06/09/2015

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Bandlimiting antinoise in personal audio devices having adaptive noise cancellation (ANC)  
Patent #
US 9,824,677 B2
Filed 05/16/2012

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Lowlatency multidriver adaptive noise canceling (ANC) system for a personal audio device  
Patent #
US 9,955,250 B2
Filed 07/06/2016

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Systems and methods for adaptive active noise cancellation for multipledriver personal audio device  
Patent #
US 10,013,966 B2
Filed 03/15/2016

Current Assignee
Cirrus Logic International Semiconductor Ltd.

Sponsoring Entity
Cirrus Logic International Semiconductor Ltd.

Feedback adaptive noise cancellation (ANC) controller and method having a feedback response partially provided by a fixedresponse filter  
Patent #
US 10,026,388 B2
Filed 08/19/2016

Current Assignee
Cirrus Logic International Semiconductor Ltd.

Sponsoring Entity
Cirrus Logic International Semiconductor Ltd.

Detection and alert of automobile braking event  
Patent #
US 10,127,816 B2
Filed 01/24/2017

Current Assignee
Blanding Hovenweep LLC

Sponsoring Entity
Blanding Hovenweep LLC

Systems and methods for selectively enabling and disabling adaptation of an adaptive noise cancellation system  
Patent #
US 10,181,315 B2
Filed 06/13/2014

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Systems and methods for multimode adaptive noise cancellation for audio headsets  
Patent #
US 10,206,032 B2
Filed 08/08/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Systems and methods for bandlimiting antinoise in personal audio devices having adaptive noise cancellation  
Patent #
US 10,219,071 B2
Filed 12/10/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Bandlimiting antinoise in personal audio devices having adaptive noise cancellation (ANC)  
Patent #
US 10,249,284 B2
Filed 10/18/2017

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Adaptive pattern recognition based control system and method  
Patent #
US 10,361,802 B1
Filed 02/02/2000

Current Assignee
HOFFBERG FAMILY TRUST 1

Sponsoring Entity
HOFFBERG FAMILY TRUST 1

Systems and methods for providing adaptive playback equalization in an audio device  
Patent #
US 10,382,864 B2
Filed 12/10/2013

Current Assignee
Cirrus Logic Incorporated

Sponsoring Entity
Cirrus Logic Incorporated

Twoinput crosstalkresistant adaptive noise canceller  
Patent #
US 4,649,505 A
Filed 07/02/1984

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
General Electric Company

Adaptive transmit preemphasis for digital modem computed from noise spectrum  
Patent #
US 5,008,903 A
Filed 05/25/1989

Current Assignee
Rembrandt Technologies LP

Sponsoring Entity
AT T Paradyne

Adaptive locallyoptimum detection signal processor and processing methods  
Patent #
US 5,018,088 A
Filed 10/02/1989

Current Assignee
Johns Hopkins University

Sponsoring Entity
Johns Hopkins University

Automatic noise reduction for individual frequency components of a signal  
Patent #
US 4,901,150 A
Filed 11/22/1988

Current Assignee
Sony Corporation

Sponsoring Entity
Sony Corporation

Moving iron instrument  
Patent #
US 4,912,398 A
Filed 03/10/1989

Current Assignee
Yokogawa Electric Corporation

Sponsoring Entity
Yokogawa Electric Corporation

Phasecontrol system for telecommunications signals received by an adaptive antenna  
Patent #
US 4,847,860 A
Filed 04/01/1986

Current Assignee
Socit Anonyme de Tlcommunications

Sponsoring Entity
Socit Anonyme de Tlcommunications

Adaptive radar signal processor for the detection of useful echo and the cancellation of clutter  
Patent #
US 4,719,466 A
Filed 12/13/1984

Current Assignee
Alenia Aeritalia Selenia SpA

Sponsoring Entity
SELENIA  INDUSTRIE ELETTRONICHE ASSOCIATE S.P.A.

Adaptive jitter canceller having sinusoidal accentuator and jitter prediction filter  
Patent #
US 4,792,964 A
Filed 01/27/1988

Current Assignee
NEC Corporation

Sponsoring Entity
NEC Corporation

ECG enhancement by adaptive cancellation of electrosurgical interference  
Patent #
US 4,537,200 A
Filed 07/07/1983

Current Assignee
Board of Trustees of the Leland Stanford Junior University

Sponsoring Entity
Board of Trustees of the Leland Stanford Junior University

Signal presence detector and method  
Patent #
US 4,363,138 A
Filed 11/27/1978

Current Assignee
TRW Limited

Sponsoring Entity
TRW Limited

Circuit for the adaptive suppression of narrow band interference  
Patent #
US 4,287,475 A
Filed 10/05/1979

Current Assignee
United States Of America As Represented By The Secretary Of The Air Force

Sponsoring Entity
United States Of America As Represented By The Secretary Of The Air Force

Noise detector employing plural delay circuits  
Patent #
US 4,220,926 A
Filed 08/23/1978

Current Assignee
Plessey Handel und Investments AG.

Sponsoring Entity
Plessey Handel und Investments AG.

Adaptive noise cancelling receiver  
Patent #
US 4,177,430 A
Filed 03/06/1978

Current Assignee
Rockwell International Corporation

Sponsoring Entity
Rockwell International Corporation

Automatic thresholding and reference circuit  
Patent #
US 4,067,013 A
Filed 11/12/1976

Current Assignee
The United States of America as represented by the Navy

Sponsoring Entity
The United States of America as represented by the Navy

System for obtaining pulse compression in the frequency domain  
Patent #
US 4,092,603 A
Filed 09/16/1976

Current Assignee
Hughes Aircraft Company

Sponsoring Entity
Hughes Aircraft Company

FREQUENCY DOMAIN SIGNAL PROCESSOR HAVING ADAPTIVE CAPABILITY  
Patent #
US 3,541,458 A
Filed 10/30/1968

Current Assignee
Klund William E.

Sponsoring Entity
Klund William E.

26 Claims
 1. A method for removing impulse noise from a circuit electrical signal s(t) that includes a desired signal s.sub.0 (t) plus a noise signal n(t), where each component may vary with time t, the method comprising the steps of:
 receiving a circuit input signal that is a sum of a desired signal s.sub.0 (t) plus a noise signal n(t) that includes impulse noise from a circuit electrical signal, where the noise signal n(t) is to be determined;
passing the circuit input signal through a bandpass filter, with a selected frequency passband PB, to form a bandpassed signal;
passing the bandpassed signal through a first selected signal processing channel that has frequency support contained in a first frequency band having a central frequency .omega..sub.1, to produce a first noise signal n.sub.1 (t);
passing the bandpassed signal through a second selected signal processing channel that has frequency support contained in a second frequency band, having a central frequency .omega..sub.2, that is spaced apart from and does not overlap any part of the first frequency band, to produce a second noise signal n.sub.2 (t);
passing the bandpassed signal through a third selected signal processing channel that has a third frequency band, having a central frequency .omega..sub.3, that is spaced apart from and does not overlap any part of the first frequency band and the second frequency band, to produce a third channel output signal s(t)+n.sub.3 (t), where the third channel has frequency support contained in the third frequency band that includes substantially all frequencies that contribute to the desired signal s(t), and where the frequencies .omega..sub.1, .omega..sub.2 and .omega..sub.3 lie in the selected frequency pass band PB;
where the first, second and third selected channels include frequency shift by selected first, second and third shift frequencies, respectively, and include passage through a low pass filter with selected first, second and third rolloff frequencies, respectively, to form first, second and third channel output signals, respectively;
forming a combined signal that is a symmetric, homogeneous function HS{n.sub.1 (t), n.sub.2 (t);
.theta.} of degree one from the first and second channel output signals, where the first and second channel output signals have m adjustable parameters .theta.=(.theta.1, . . . ,.theta.m;
m=1 or
2);
forming a linear combination signal LC{HS{n.sub.1 (t), n.sub.2 (t);
.theta.};
s(t)+n.sub.3 (t)} of the combined signal HS{n.sub.1 (t), n.sub.2 (t);
.theta.} and the third channel output signal s(t)+n.sub.3 (t);
adjusting the parameters .theta. whereby the linear combination signal LC{HS{n.sub.1 (t), n.sub.2 (t);
.theta.};
s(t)+n.sub.3 (t)} provides a best possible estimate of the desired signal s(t), according to a selected criterion, for an optimal choice .theta.=.theta..sub.0 of these parameters; and
issuing the linear combination signal LC{HS{n.sub.1 (t), n.sub.2 (t);
.theta.=.theta..sub.0 };
s(t)+n.sub.3 (t)} as an estimate of the desired signal s(t).  View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16)
 receiving a circuit input signal that is a sum of a desired signal s.sub.0 (t) plus a noise signal n(t) that includes impulse noise from a circuit electrical signal, where the noise signal n(t) is to be determined;
 17. A method for removing impulse noise from a circuit electrical signal that includes a desired signal s.sub.0 (t) plus a noise signal n(t), where each component may vary with time t, the method comprising the steps of:
 receiving a circuit input signal that is a sum of a desired signal s.sub.0 (t) plus a noise signal n(t) that includes impulse noise from a circuit electrical signal, where the noise signal n(t) is be determined;
passing the circuit input signal through a bandpass filter, with a selected frequency passband, to form a bandpassed output signal;
passing the bandpassed output signal through a first selected signal processing channel that has frequency support contained in a first frequency band having a central frequency .omega..sub.1, to produce a first noise signal n.sub.1 (t);
passing the bandpassed output signal through a second selected signal processing channel that has frequency support contained in a second frequency band, having a central frequency .omega..sub.2, that is spaced apart from and does not overlap any part of the first frequency band, to produce a second noise signal s(t)+n.sub.2 (t), where the second channel has frequency support contained in a second frequency band that includes substantially all frequencies that contribute to the desired signal s(t);
where the first and second selected channels include frequency shift by selected first and second shift frequencies, respectively, and include passage through a low pass filter with selected first and second rolloff frequencies, respectively, to form first and second channel output signals, respectively;
forming a linear combination signal LC{n.sub.1 (t), s(t)+n.sub.2 (t);
.theta.} of the first channel output signal n.sub.1 (t) and the second channel output signal s(t)+n.sub.2 (t), where this linear combination has m adjustable parameters .theta.=(.theta.1, . . . , .theta.m;
m=1 or
2);
adjusting the parameters .theta. whereby the linear combination signal LC{n.sub.1 (t), s(t)+n.sub.2 (t);
.theta.} provides a best possible estimate of the desired signal s(t), according to a selected criterion, for optimal choices .theta.=.theta..sub.0 of these parameters; and
issuing the linear combination signal LC{n.sub.1 (t), s(t)+n.sub.2 (t);
.theta.=.theta..sub.0 } as an estimate of the desired signal s(t).  View Dependent Claims (18, 19, 20, 21, 22, 23, 24, 25, 26)
 receiving a circuit input signal that is a sum of a desired signal s.sub.0 (t) plus a noise signal n(t) that includes impulse noise from a circuit electrical signal, where the noise signal n(t) is be determined;
1 Specification
This invention relates to adaptive processing of noisy electrical signals to reduce or cancel impulse noise.
BACKGROUND OF THE INVENTIONElectrical signals often include an undistorted signal component and a noise component, where the noise arises from several sources, including white or Gaussian noise, temperatureinduced Johnson noise, impulse noise, crosstalk noise, echo noise, intermodulation noise, amplitude noise, jitter noise, uncontrolled voltage variation, and several varieties of signal distortion. Gaussian noise has been studied widely, and many techniques exist to control or reduce the effects of Gaussian noise. Impulse noise has received less attention.
Klund, in U.S. Pat. No. 3,541,458, discloses use of an adaptive array of sensors to detect a weak (desired) signal in the presence of noise. The amplitude weights and phases assigned to the different sensors are varied to produce a signal that, ideally, is free of the spatially stationary component of the noise field. The invention uses a Fourier series representation for the waveform produced by each sensor and uses time averaging and other manipulations to identify the optimal amplitudes and phases assigned to each sensor.
An adaptive broadband noise cancellation signal receiver is disclosed in U.S. Pat. No. 4,177,430, issued to Paul. A desired radiofrequency passband is selected, and an input signal in this band is converted to an intermediate frequency broadband signal. The converted input signal is passed through two parallel signal processing channels having differing frequency characteristics. The first channel output represents desired signal plus noise, and the second channel output represents primarily noise, after filtering by an adaptive transversal filter. The two channel outputs are then subtracted from each other to produce a waveform that is primarily the desired signal. The adaptive filter shuts off if the first and second channel outputs have no audio component in common and emphasizes any broadband noise that is present.
Eaton et al disclose a circuit for suppression of narrowband interference noise (as opposed to white noise), in U.S. Pat. No. 4,287,475. The received signal, containing the desired signal plus narrowband noise plus wideband noise, is discrete Fourier transformed using a chirp algorithm illustrated in Fourier space in FIG. 3. Sidelobes are suppressed by about 40 dB relative to a center peak by additional filtering. The resulting signal is smoothed to produce a power spectral density function, apparently representing the narrowband noise, which is used to notch out this noise in the time domain.
U.S. Pat. No. 4,537,200, issued to Widrow, discloses a system for electrocardiogram signal enhancement by adaptive cancellation of noise created by operation of an electrosurgical instrument, such as a "Bovie", used for cauterizing tissue cut by an adjacent surgical knife. Gross interference noise is filtered out by use of passive lowpass and active lowpass filters. A first waveform, consisting of the desired signal plus a first noise signal, and a second waveform, consisting of a second noise signal only, are formed, with the two noise waveforms being correlated with each other but not with the desired signal. The second noise waveform only is passed through an adaptive filter, which consists of a tapped delay line, with the tapped off signals being weighted and summed to form a Least Mean Squares (LMS) estimate of the first noise waveform. The first and second waveforms are then subtracted from each other. The resulting difference signal, representing the desired signal, is fed back to the adaptive filter to determine, in a time varying manner, an LMS estimate of the first noise waveform.
Zinser et al, in U.S. Pat. No. 4,649,505, disclose a twoinput adaptive noise canceller that also begins with a first signal (desired signal plus first noise signal) and a second (noise only) signal, where the two noise signals are highly correlated. As best illustrated in FIG. 2, the second signal is passed through a first adaptive filter to produce a first "noise only" signal that is subtracted from the first signal (desired signal plus first noise signal). This first difference signal is then passed through a second adaptive filter, the output thereof is subtracted from the second (noise only) signal, and this second difference signal is used to adjust the second adaptive filter to improve the apparatus output signal, which is the first difference signal. The inventors suggest that this approach is needed to remove any vestiges of the desired signal that may be present in the second (noise only) signal.
An adaptive radar signal processor that cancels a clutter or noise signal that is present is disclosed in U.S. Pat. No. 4,719,466, issued to Farina et al. The apparatus is applied to distinguish a desired echo radar signal from a clutter signal and from thermal noise (both Gaussian) that are part of the return signal. A matrix D representing a steady state estimate is constructed by statistical manipulations, and an invertible covariance matrix M is constructed by the relation M.sup.1 =D.sup.T D*, where D.sup.T and D* are the transpose and complex conjugate of the matrix D. The initial signal uses time delay, multiplication and integration of various signal to determine a covariance matrix.
Yoshida discloses an adaptive jitter noise canceller in U.S. Pat. No. 4,792,964. An initial signal, containing the desired signal and superimposed jitter noise, is passed through a low pass filter, then sampled at a uniformly spaced sequence of times {t.sub.0 +n.DELTA.t.sub.0 } (n=0, 1, 2, . . . ; N a positive integer), then passed through a sinusoidal accentuator. The output signal is passed through an interpolator that produces quasisamples at a sequence of times {t.sub.1 +n.DELTA.t.sub.1 }, then passed through a second lowpass filter, then passed through a prediction filter to produce a filter output signal. This filter output signal is then subtracted from the initial signal to cancel the jitter noise.
U.S. Pat. No. 4,914,398, issued to Jove et al, disclose a system that suppresses additive signal disturbances in data channels that contain magnetoresistive transducers. The system passes an initial signal through a time delay and also passes the initial signal through positive and negative signal envelope detectors that are subtracted from each other to produce a second signal. This second signal is passed through a nonlinear adaptive filter to produce an estimate of the disturbance present. The adaptive filter output signal is subtracted from the time delayed initial signal to produce a difference signal that suppresses the additive disturbance.
An adaptive pretransmission filter for a modem, computed from the observed noise spectrum of the transmission channel, is disclosed by Betts et al in U.S. Pat. No. 5,008,903. A noise spectrum analyzer positioned at the receiver end calculates the difference between the transmitted signal and the received signal, and this difference is discretely Fourier transformed to produce an estimated noise spectrum. This noise spectrum is then transmitted back to the transmitter on a secondary channel, and pretransmit filter coefficients are determined and used to suppress the observed transmission channel noise.
In U.S. Pat. No. 5,018,088, Higbie discloses an adaptive, locallyoptimum signal detector and processor for spread spectrum communications. Here, the noise strength is much greater than the initial strength of the desired signal. The amplitude probability distribution function (APDF) of the noise is determined approximately, using the APDF of the desired signal plus noise as an estimate. This APDF estimate is then manipulated mathematically to produce one or more spectra that represent the noise alone, and a signal with enhanced desired signal component is produced. This approach changes adaptively with change in the present noise statistics.
Widrow et al, in "Adaptive Noise Cancelling: Principles and Applications", I.E.E.E. Proc., vol. 63 (1975) pp. 16921716, disclose use of a first or primary signal containing desired signal plus noise s(t)+n.sub.1 (t) and a second, noiseonly signal n.sub.2 (t), where the two noise signals have nonzero correlation with each other but have zero correlation with the desired signal. The second signal is passed through an adaptive filter whose output signal is subtracted from the first signal, and the output difference signal is used to adjust the adaptive filter to produce an output difference signal in which the first noise component n.sub.1 (t) is cancelled as completely as possible, using a Least Mean Squares algorithm to maximize noise cancellation. The basic problem can be generalized somewhat by adding third and fourth noise components n.sub.3 (t) and n.sub.4 (t) to the first and second signals, where these third and fourth noise components are uncorrelated with all other signals. This basic approach, where n.sub.3 (t)=n.sub.4 (t)=0, is used in many of the patents discussed above and in the invention. This article also discusses possible approaches when the second signal has nonzero correlation with the desired signal, a problem also discussed in the Zinsser et al patent (4,649,505) above.
In "Adaptive Noise Canceling Applied to Sinusoidal Interferences", I.E.E.E. Trans. Acoustics, Speech and Signal Processing, vol. 25 (1977) pp. 484491, Glover discloses a method for eliminating sinusoidal or other periodic interferences that distort a desired signal. The basic approach discussed by Widrow et al in the 1975 article, and the adaptive filter with feedback is approximated by a linear, timeinvariant filter. The noise or interference is a sinusoidal term or sum of such terms, whose presence is suppressed by appropriate choices of complex weighting coefficients in a related filter function.
Zeidler et al disclose a method for "Adaptive Enhancement of Multiple Sinusoids in Uncorrelated Noise", I.E.E.E. Trans. Acoustics, Speech and Signal Processing, vol;. 26 (1978) pp. 240254. Here, the noise is white noise and the desired signal is a finite sum of sinusoids with arbitrary frequencies. The method involves approximate inversion of certain scalar or matrix equations to determine certain weighting coefficients that define the desired signal.
Sambur, in "Adaptive Noise Canceling for Speech Signals", I.E.E.E. Trans. Acoustics, Speech and Signal Processing, vol. 26 (1978) pp. 419423, discloses an adaptive filtering technique that cancels out the desired signal, leaving a pure noise signal that can, presumably, be subtracted from the original combined signal to provide the desired signal. The method uses a variant on the usual Least Mean Squares (LMS) approach discussed in the Widrow et al 1975 article.
Ferrara and Widrow, in "The Time Sequenced Adaptive Filter", I.E.E.E. Trans. Acoustics, Speech and Signal Processing, vol. 29 (1981) pp. 679683, disclose an alternative to the LMS filter, using a timesequenced filter and associated algorithm. The algorithm provides a different weight matrix or vector for each of a recurring sequence of error surfaces defined by a least mean square error computed in the basic approach.
What is needed is an adaptive signal processing method that, ideally at least, will result in cancellation of most or all impulse noise and that does not require technically complex processing of the input signal. Preferably, the method should accept and process an input signal in real time, with little or no delay in issuance of an output signal with most or all of the noise removed.
SUMMARY OF THE INVENTIONThese needs are met by the invention, which provides several embodiments of method and apparatus for substantial reduction or complete removal of impulse noise. In one embodiment, the method invention uses three signal processing channels, one channel for the desired signal plus (impulsive) noise, s(t)+n.sub.3 (t), a first (pilot) channel for representative noise, n.sub.1 (t), and a second (pilot) channel for representative noise, n.sub.2 (t). The first, second and third channel signals are formed by passing a single input signal containing the desired signal plus impulse noise through parallel first, second and third channel filters, respectively, that have adjacent but spaced apart and nonoverlapping regions of notsubstantiallyzero amplitudes H.sub.1 (f), H.sub.2 (f) and H.sub.3 (f), respectively, in Fourier transform or frequency space with corresponding "frequency support regions" FS1, FS2 and FS3 for the filter amplitudes. The desired or noisefree signal s(t) has a Fourier transform S(f) that is substantially nonzero only within the frequency support region for the filter function H.sub.3 (f), and the frequency support regions for H.sub.1 (f), H.sub.2 (f) and H.sub.3 (f) are typically separated by 502,000 Hz. Passage of additive white Gaussian noise through these filters produce three output signals, y.sub.1G (t), y.sub.2G (t) and y.sub.3G (t), that are uncorrelated in time t: the expectation of the product, E{y.sub.iG (t) y.sub.kG (t)} with i.notident.k is zero.
However, passage of impulse noise, such as noise produced by lightning or another natural disturbance, through any two of the filters in parallel produces two output signals, h.sub.i (t) and h.sub.k (t), that are strongly correlated in time. The amplitudes of the two filter output signals h.sub.i (t) and h.sub.k (t) are proportional to the input signal amplitudes, and the phases of the two filter output signals h.sub.i (t) and h.sub.k (t) are proportional to the times of arrival of the impulse noise signal at the two filters.
In a generic embodiment, the input signal is received and processed by three parallel channels The third channel includes a first filter with frequency response H.sub.3 (f) whose frequency support region contains the frequency support region for the desired noisefree signal s(t). The first and second channels include first and second filters with respective frequency responses H.sub.1 (f) and H.sub.2 (f) whose frequency support regions are spaced apart from, do not overlap with, and flank (lie to the right and left, respectively, of) the frequency support region for the third filter. The centers of the first and second filter frequency support regions are spaced apart from the center of the third filter frequency support region by relatively small frequency differences .delta..omega..sub.1 and .delta..omega..sub.2, respectively, preferably of the order of 502,000 Hz.
More specifically, the input signal is passed through a bandpass filter and is split into three channels by three signal mixers operating at spaced apart frequencies .omega..sub.3 +.delta..omega..sub.1, .omega..sub.3 +.delta..omega..sub.2, and .omega..sub.3, where .omega..sub.3 is a central frequency for the desired or noisefree signal and .delta..omega..sub.i is a suitable frequency offset for channel i (i=1,2). Each processing channel includes multiplication of the input signal by a signal of the form exp{j(.omega..sup.3 +.delta..omega..sub.1)tj.theta..sub.LO)}, exp{j(.omega..sub.3 +.delta..omega..sub.2)tj.theta..sub.LO)} or exp{j(.omega..sub.3 t.theta..sub.LO)} (j.sup.2 =1), for image translation in the respective first, second and third channels, and passage of the resulting signals through an appropriate low pass filter, where .delta..omega..sub.i =(.omega..sub.i .omega..sub.3) (i=1,2) and .theta..sub.LO is the phase angle of the mixer or local oscillator.
In a general implementation of this embodiment, the first, second and third channel output signals are formed, and a combination signal N(t;.phi.)=exp(j.phi.){n.sub.1 (t) n.sub.2 (t)}.sup.1/2 is formed and added to the output signal s(t)+n.sub.3 (t) of the third channel. In this embodiment, the choice of a real phase angle .phi. (0.ltoreq..phi.<2.pi.) is made based upon an algorithm incorporated in the circuit. In one version, the choice of .phi. is restricted to .phi.=0 or .pi. is so that exp(j.phi.)=.+.1. In another version of this embodiment, .PHI. ranges over a discrete or continuous set of numbers drawn from the range 0.ltoreq..phi.<2.pi..
In a second general implementation, the output signal is s(t)+n.sub.3 (t)+M(t;.psi.1, .psi.2) is formed, with M(t;.psi.1, .psi.2)={n.sub.1 (t) exp(j.psi..sub.1)+n.sub.2 (t) exp(j.psi..sub.2)}/2, and choices for values of the real phase angles .psi.i(0.ltoreq..psi.i<2.pi.; i=1,2) are made, according to an algorithm, for a best possible estimation of the desired noisefree signal s(t). Other general combinations of the signals n.sub.1 (t) and n.sub.2 (t) can also be formed and used for the noise cancellation.
The phase angles .psi..sub.1, .psi..sub.2 and .phi. can be chosen using a general algorithm incorporated in the circuit, or these angles can be estimated from known behavior of the phase in the Fourier transform of a pure impulse signal. In the first situation, the phase angles are determined using feedback and an algorithm that provides a "best" estimation for the noise signal n.sub.3 (t) according to a quantitative measure of estimation error. In the second situation, the phase angles are estimated from observations of the signals n.sub.1 (t) and n.sub.2 (t), and the choice of passband central frequencies for the signal channels that form n.sub.1 (t) and n.sub.2 (t) is flexible. No feedback is required in this second situation.
BRIEF DESCRIPTION OF THE DRAWINGSFIG. 1 is a graph of the frequency support functions for three filters, H.sub.1 (.omega.), H.sub.2 (.omega.) and H.sub.3 (.omega.), and for the desired noisetree signal, S(f), used in several embodiments of the invention.
FIGS. 2A and 2B are graphs of filter output signals y.sub.1 (t), y.sub.2 (t) and y.sub.3 (t) for respective input signals representing a Gaussian noise signal and an impulse noise signal.
FIGS. 3, 5, 6, 7 and 9 are schematic views of several embodiments of the invention.
FIGS. 4 and 8 are graphs in the complex plane illustrating use of particular combinations of the channel signals that are used in the embodiments of FIGS. 3, 5, 6, 7 and 9.
FIGS. 10A, 10B, 10C, 10D, 10E and 10F are graphs of the magnitudes of the signals s(t)+n(t) (wideband, before passage through a low pass filter), s(t) (narrowband, after passage through a low pass filter), s(t)+n.sub.3 (t) (narrowband), s(t)+n.sub.3 (t)+N(t;.phi.=0), s(t)+n.sub.3 (t)N(t;.phi.=0) and s(t)+n.sub.3 (t).+.N(t;.phi.=0) (optimized), respectively, produced in the embodiment of FIG. 3 and used in the choice module in FIG. 3, where N(t;.phi.)=exp(j.phi.) {n.sub.1 (t) n.sub.2 (t)}.sup.1/2 is a particular combination of two noise channel signals discussed below, using displacement from the signal median as a quantitative measure of error.
FIGS. 11A, 11B, 11C, 11D, 11E and 11F are graphs of the phase angles for the signals s(t)+n(t) (wideband), s(t) (narrowband), s(t)+n.sub.3 (t) (narrowband), s(t)+n.sub.3 (t)+N(t;.phi.=0), s(t)+n.sub.3 (t)N(t;.phi.=0), and s(t).+.n.sub.3 (t).+.N(t;.phi.=0) (optimized), respectively, produced in the embodiment of FIG. 3 and used in the choice module in FIG. 3, using the error measure used in FIGS. 10A10F.
DESCRIPTION OF PREFERRED EMBODIMENTS OF THE INVENTIONFIG. 1 schematically illustrates the frequency support regions FS.sub.1, FS.sub.2 and FS.sub.3, centered at frequencies .omega..sub.1, .omega..sub.2, and .omega..sub.3, for three filters, H.sub.1 (.omega.), H.sub.2 (.omega.) and H.sub.3 (.omega.), respectively, and for the desired noisefree signal, S(.omega.), used for the invention. As used herein, the term "frequency support region" for a function F(.omega.) that is not identically zero, refers to the smallest connected (onepiece) region on the real line in the frequency domain (.omega.) for which F(.omega.) is substantially zero (or has very small magnitude) outside this region. The desired noisefree signal s(t) has a corresponding function S(.omega.) in the frequency domain, with frequency support region that is contained in the frequency support region for the third filter frequency response function H.sub.3 (.omega.). The frequency support regions for the frequency response functions H.sub.1 (.omega.) and H.sub.2 (.omega.) flank (lie on the left and right, respectively, of), do not overlap with, and are separated by distances of .delta..omega..sub.1 and .delta..omega..sub. 2 from, the center (.omega.=.omega..sub.3) of the frequency support region for H.sub.3 (.omega.), as shown. The separation distances may each typically be of the order of 502,000 Hz. These separation distances may be positive or negative. Optionally, the circuit may delete the first filter (H.sub.1 (.omega.)) or the second filter (H.sub.2 (.omega.) ), which contains noise only, in some embodiments of the invention.
Impulse noise is assumed to be represented by a timeshifted delta function A.delta.(t.tau.) in the input signal, or by a sum of such functions, where A and .tau. are parameters determined by the environment. It is further assumed that the output signal phase for any filter that receives the circuit input signal is linearly proportional to the frequency, or collection of frequencies, that make up the input signal; this assumption need only be satisfied in neighborhoods of the frequencies .omega..sub.1, .omega..sub.2 and .omega..sub.3.
FIG. 2A illustrates the effect of passing a Gaussian noise signal n.sub.G (t) through the filters of channels 1, 2 and 3. This produces output signals y.sub.1G (t), y.sub.2G (t), y.sub.3G (t) with substantially zero correlation in time: the expectation E{y.sub.iG (t) y.sub.kG (t)} is zero for i.notident.k.
FIG. 2B illustrates the effect of passing an impulse noise signal, represented by a timeshifted delta function with amplitude A, A.delta.(t.tau.), through the filters of channels 1, 2 and 3. This produces the respective impulse response functions Ah.sub.1 (t.tau.), Ah.sub.2 (t.tau.) and Ah.sub.3 (t.tau.), and any two of these three response functions are qualitatively similar and are highly correlated. Thus, knowledge of one or more of the functions Ah.sub.1 (t.tau.) and Ah.sub.2 (t.tau.) allows one to predict the shape and amplitude of the impulse function Ah.sub.3 (t.tau.) with reasonable accuracy and to remove or cancel the impulse noise contribution from the third channel output waveform s(t)+n.sub.3 (t)=s(t)+Ah.sub.3 (t.tau.).
FIG. 3 is a schematic view illustrating a first embodiment 11 of apparatus constructed according to the invention. The apparatus receives the sum of a desired noisefree signal s(t) plus an impulse noise signal n(t), where the impulse noise signal can be represented by a timeshifted delta function with amplitude A, A.delta.(t.tau.), namely where the amplitude A and time shift .tau. are characteristic of that impulse noise source. Alternatively, the impulse noise may be represented by a sum of such timeshifted delta functions, each delta function having its own amplitude and time shift parameters. The input signal s(t)+n(t) is received by a band pass filter (BPF) 13 that limits the contribution to a band of frequencies including the frequency support regions FS.sub.1, FS.sub.2 and FS.sub.3, illustrated in FIG. 1. The output waveform of the band pass filter 13 is received by a mixer 15 that multiplies this output signal by the function exp{j(.omega..sub.LO t+.theta..sub.LO)}, where .omega..sub.LO and .theta..sub.LO are the frequency and phase angle of the local oscillator, then by an intermediate frequency band pass filter 17 with selected frequency bandpass that includes the frequency support regions FS.sub.1, FS.sub.2 and FS.sub.3. The BPF 13 and/or the mixer 15 are characteristic of a double frequency conversion receiver, are optional, and may be deleted here.
The output signal s.sub.17 (t) from the BPF 17 is received by three signal processing channels 21, 31 and 41. In the first channel 21, the output signal s.sub.17 (t) is received by a local oscillator 23 that multiplies its input signal by exp{j(.DELTA..omega..sub.1 t+.theta..sub.LO)}, then passed through an LPF 25 with a rolloff frequency .omega..sub.u1. In the second channel 31, the output signal s.sub.17 (t) is received by a local oscillator 33 that multiplies its input signal by exp{j(.DELTA..omega..sub.2 t+.theta..sub.LO)}, then passed through an LPF 35 with a rolloff frequency .omega..sub.u2 (which is usually identical to .omega..sub.u1). In the third channel 41, the output signal s.sub.17 (t) is received by a local oscillator 43 that multiplies its input signal by exp{j(.DELTA..omega..sub.3 t+.theta..sub.LO)}, then passed through an LPF 45 with a rolloff frequency .omega..sub.u3. (which is usually identical to .omega..sub.u1) If a double frequency conversion receiver is used, the parameters .DELTA..omega..sub.1, .DELTA..omega..sub.2 and .DELTA..omega..sub.3 are defined by the relations
.DELTA..omega..sub.1 .apprxeq..omega..sub.1 .omega..sub.LO,(1)
.DELTA..omega..sub.2 .apprxeq..omega..sub.2 .omega..sub.LO,(2)
.DELTA..omega..sub.3 .apprxeq..omega..sub.3 .omega..sub.LO.(3)
If a single frequency conversion receiver is used, the frequency .omega..sub.LO is deleted in Eqs. (1), (2) and (3). The parameters .omega..sub.LO, .theta..sub.LO, and .omega..sub.u1 may be selected according to other criteria.
Output signals n.sub.1 (t) and n.sub.2 (t) from the two channels 21 and 31 are received by a square root/multiplier module 38 that issues the (complex) square root/product or geometric mean output function:
N.sub.1 (t)={n.sub.1 (t) n.sub.2 (t)}.sup.1/2. (4)
The output signal N.sub.1 (t) is received by one input terminal of a sum module 47 and by one input terminal of a difference module 49. A second input terminal of the sum module 47 and of the difference module 49 receives the output signal s(t)+n.sub.3 (t) from the third channel 41. The sum module 47 and the difference module 49 produce and issue the output signals s.sub.+ (t)=s(t)+n.sub.3 (t)+N.sub.1 (t), s(t)+n.sub.3 (t) and s.sub. (t)=s(t)+n.sub.3 (t)N.sub.1 (t), respectively. The three output signals s.sub.+ (t), s(t)+n.sub.3 (t) and s.sub. (t) are received by a choice module 51 that determines which of these three output signals provides a better representation of the desired noisefree signal s(t), using an algorithm that examines the output signals s.sub.+ (t), s(t)+n.sub.3 (t) and s.sub. (t) and makes this determination. The apparatus 11 then issues an output signal:
S.sub.1 (t)=s.sub.+ (t) or s(t)+n.sub.3 (t) or s.sub. (t),(5)
illustrating formation of the signal N.sub.1 (t) in the square root/multiplier module 39 and estimation of the noise component n.sub.3 (t).
In a second embodiment 61 of the invention, illustrated in FIG. 5, an input signal s(t)+n(t) again is received by the band pass filter 17, which issues an output signal s.sub.17 (t). Again, the BPF 13 and mixer 15 shown in FIG. 3 are optional here. The BPF output signal s.sub.17 (t) is received by the three signal processing channels 21, 31 and 41, as in FIG. 3, and the three channel output signals n.sub.1 (t), n.sub.2 (t) and s(t)+n.sub.3 (t) are produced as in FIG. 3. The channel output signals n.sub.1 (t) and n.sub.2 (t) are received at two input terminals of a complex square root/multiplier module 39, which produces and issues a complex square root/product output signal:
N(t;.phi.)=exp(j.phi.) N.sub.1 (t)=exp(j.phi.) {n.sub.1 (t)n.sub.2 (t)}.sup.1/2, (6)
where .phi. is a real number in the range 0.ltoreq..phi.<2.pi.. This output signal is multiplied by a real number .alpha. at a product module 63, then received at one input terminal of a sum module 65 that receives the third channel output signal s(t)+n.sub.3 (t) at its second input terminal. The sum module output signal:
S.sub.2 (t;.phi.,.alpha.)=s(t)+n.sub.3 (t)+.alpha.N(t;.phi.) =s(t)+n.sub.3 (t)+.alpha. exp(j.phi.) {n.sub.1 (t)n.sub.2 (t)}.sup.1/2 (7)
is then received by a choice module 67 that adjusts the real numbers .phi. and .alpha. by feedback to the second input terminal of the module 63, to obtain the best possible representation for the desired noisefree signal s(t) by the output signal S.sub.2 (t;.phi.,.alpha.). The choices (.alpha.=1, .phi.=0), (.alpha.=1, .phi.=.pi.) and (.alpha.=0) produce the candidate signals s(t)+n.sub.3 (t).+.{n.sub.1 (t)n.sub.2 (t)}.sup.1/2 and s(t)+n.sub.3 (t) used in FIG. 3. The phase angle .phi. can be drawn from a discrete, partly continuous or wholly continuous subset of the interval 0.ltoreq..phi.<2.pi..
In a third embodiment of the invention, illustrated in FIG. 6, one of the two channels 21 or 31 is deleted, product modules 29 and 63 form and issue an output signal .alpha. exp(j.phi.) n.sub.i (t), and a sum module 66 issues an output signal:
S.sub.3 (t;.phi.;.alpha.)=s(t)+n.sub.3 (t)+.alpha. exp(j.phi.) n.sub.i (t) (i=1 or 2) (8)
that is received by a choice module 69, which adjusts the phase .phi. and multiplier .alpha. to obtain the best possible representation for the desired noisefree signal s(t) by the output signal S.sub.3 (t;.phi.,.alpha.).
In a fourth embodiment 71 of the invention, illustrated in FIG. 7, an input signal s(t)+n(t) again is received by the BPF 17, which issues an output signal s.sub.17 (t)=s(t)+n.sub.3 (t). The BPF output signal s.sub.17 (t) is received and processed by the three signal processing channels 21, 31 and 41, as in FIG. 3, and the three channel output signals n.sub.1 (t), n.sub.2 (t) and s(t)+n.sub.3 (t) are produced as in FIG. 3. The channel output signals n.sub.1 (t) and n.sub.2 (t) are received at two input terminals of a complex arithmetic mean module 40, which produces and issues the (complex) arithmetic mean output function:
M(t;.psi.1,.psi.2)={n.sub.1 (t) exp(j.psi.1)+n.sub.2 (t) exp(j.psi.2)}/2,(9)
where .psi.1 and .psi.2 are real numbers in the range 0.ltoreq..psi.i<2.pi.(i=1,2).
FIG. 8 is a graph of the real and imaginary parts of the signals n.sub.1 (t), n.sub.2 (t) and n.sub.3 (t), illustrating formation of the signal M(t;.psi.1,.psi.2) in the complex arithmetic mean module 40.
The output function M(t;.psi.1,.psi.2) in FIG. 7 is received by one input terminal of a product module 73 that forms and issues an output signal .alpha.M(t;.psi.1,.psi.2), where the real numbers .alpha. and .psi.i (023 .psi.i<2.pi.; i=1 or 2) are determined by a choice module 77. A convenient choice here is .alpha.=1. The output signal of the product module 73 is received at one input terminal of a sum module 75 that receives the third channel output signal s(t)+n.sub.3 (t) at its second input terminal. The sum module output signal:
S.sub.4 (t;.psi.1,.psi.2,.alpha.)=s(t)+n.sub.3 (t)+.alpha.M(t;.psi.1,.psi.2)(10)
is then received by the choice module 77 that adjusts the real numbers .psi.1 and .psi.2 by feedback to the arithmetic mean module 40, to obtain the best possible representation for the desired noisefree signal s(t) by the output signal S.sub.4 (t;.psi.1,.psi.2,.alpha.).
A fifth embodiment is obtained in FIG. 7 by setting the phase angles .psi.1=.psi.2=.psi. and producing a sum module output signal:
S.sub.5 (t;.psi.,.alpha.)=s(t)+n.sub.3 (t)+.alpha.M(t;.psi.,.psi.),(11)
which is then optimized by appropriate choices of two parameters, .psi. and .alpha.. Normally, at most two of the three parameters .psi.1, .psi.2 and .alpha. are varied so that either .alpha.=1 and/or .psi.1=.psi.2=.psi. in Eq. (10).
The phase angles .psi.i (i=1,2) or .psi. used in the output signal M(t;.psi.1,.psi.2) are produced by the choice module 77 and introduced by feedback on an output line 78 to the product module 73. The phase angles .psi.i can be determined by a sequential search procedure that optimizes the apparatus output signal from the sum module 75 with respect to the choice of phase angles .psi.i. In the second, third, fourth and fifth embodiments, the multiplier .alpha. and/or one or more phase angles (.phi., .psi.1, .psi.2 or .psi. can be prescribed (for example, .alpha.=1 in Eqs. (7), (8) or (11)) rather than being determined by signal optimization.
Alternatively, the phase angles .psi.i and .phi. in FIGS. 5, 6 and 7 can be estimated by making use of a property of the phase .PSI. of the Fourier transform of an impulse function A.delta.(t.pi.). The phase .PSI. increases linearly with the frequency .omega. in the Fourier transform domain; that is,
.PSI.(.omega.)=a+b.omega., (12)
where a and b are constants. In this alternative, shown in FIG. 9, the output signals n.sub.1 (t) and n.sub.2 (t) of the respective first and second channels are carried by respective signal lines 27 and 37 to a phase estimation module 79 that produces phases .PSI.1 and .PSI.2 defined by the relations
.PSI.1=a+b.omega..sub.1, (13)
.PSI.2=a+b.omega..sub.2, (14)
where .PSI.1 and .PSI.2 are assumed to be measured by the phase estimation module 79. The coefficients a and b in Eqs. (13) and (14) are then determined by the relations
a={(.PSI..sub.1) (.omega..sub.2)(.PSI..sub.2) (.omega..sub.1)}/ ((.omega..sub.2 .omega..sub.1)) (15a)
b=(.PSI.2.PSI.1)/ (.omega..sub.2 .omega..sub.1), (15b)
and the phase associated with the noise portion of the signal s(t)+n.sub.3 (t) in the third channel 41 is estimated as
.PSI.3=a+b.omega..sub.3. (16)
If only the difference of two phase angles is sought, the coefficient a in Eq. (15a) is not used. A product module 74 receives the input signals n.sub.1 (t) and .PSI.1.PSI.3 and produces and issues an output signal exp[j(.PSI.1.PSI.3)]n.sub.1 (t) that is subtracted from the signal s(t)+n.sub.3 (t) by a difference module 76 to produce an estimate
S.sub.3 '(t;.PSI.1,.PSI.3,.alpha.=1)=s(t)+n.sub.3 (t)exp{j(.PSI.1.PSI.3)}n.sub.1 (t). (17)
This alternative approach for determination of the phase .PSI. can also be used to determine the phase angle .phi. in FIG. 5. The phase angle .phi. in Eq. (7) is chosen as
.phi.=.PSI.3(.PSI.1+.PSI.2)/2,. (18)
and this choice produces an estimate signal S.sub.2 '(t;.phi.,.alpha.). The phase angles .psi.1 and .psi.2 in Eq. (9) are chosen to be
.psi.i=.PSI.i.PSI.3 (i=1,2), (19)
and these choices produce an estimate signal S.sub.4 '(t;.psi.1,.psi.2,.alpha.).
The output signals n.sub.1 (t) and/or n.sub.2 (t) from the first and/or second channels in FIG. 5 can be replaced by output signals n.sub.4 (t) and/or n.sub.5 (t) from one or two new channels, constructed as the first and second channels are but with different selected central frequencies .omega..sub.4 and/or .omega..sub.5 and measured phases .PSI.4 and/or .PSI.5 replacing the frequency and phase angle parameters used in Eqs. (15a) and (15b). However, it is usually more convenient to use the channel signals n.sub.1 (t) and n.sub.2 (t) for this purpose.
In FIG. 1, the center frequencies .omega..sub.1 and .omega..sub.2 are shown flanking the center frequency .omega..sub.3 on the left and right. If the phase angle .phi. or .psi.1 or .psi.2 or .psi. is determined by the algorithm discussed in connection with Eqs. (15a) and (15b), the center frequencies .omega..sub.1, .omega..sub.2 and .omega..sub.3 may be chosen anywhere on the positive real axis, provided that .delta..omega..sub.1 and .delta..omega..sub.2 are not so large that phase linearity is lost. For example, one might choose .omega..sub.1 <.omega..sub.2 <.omega..sub.3 or .omega..sub.3 <.omega..sub.1 <.omega..sub.2 using this algorithm.
The functions N(t;.phi.) (complex geometric mean) and M(t;.psi.1,.psi.2) (complex arithmetic mean) set forth in Eqs. (6) and (9) are particular examples of a twoplace, symmetric, homogeneous function of the variables p.sub.1 and p.sub.2 of degree one with two associated adjustable parameters .theta.1 and .theta.2 (which may be identical), denoted HS(p.sub.1,p.sub.2 ;.theta.1,.theta.2), that satisfies the following conditions.
HS(p.sub.1,p.sub.2 ;.theta.1,.theta.2),=HS(p.sub.2,p.sub.1 ;.theta.2,.theta.1), (symmetry), (20)
HS(kp.sub.1,kp.sub.2 ;.theta.1,.theta.2),=k HS(p.sub.1,p.sub.2 ;.theta.1,.theta.2), (k.noteq.0) (homogeneity of degree one).(21)
Another example of such a function is the function
HS(p.sub.1,p.sub.2 ;.theta.1,.theta.2),={A(.theta.1p.sub.1).sup.h +B (.theta.2p.sub.2).sup.h }.sup.1/h exp{C (.theta.2p.sub.2 /.theta.1p.sub.1)}(h real and .noteq.0), (22)
where A, B and C are arbitrary constants. More generally, any twoplace symmetric, homogeneous function of degree one HS{n.sub.1 (t), n.sub.2 (t); .theta.1,.theta.2} with two adjustable parameters .theta.1 and .theta.2 (which may be identical) can be formed from the output noise signals from channels 21 and 31, and a general linear combination LC of the two signals s(t)+n.sub.3 (t) and HS{n.sub.1 (t), n.sub.2 (t); .theta.1,.theta.2} can be received at the input terminal of the choice or estimation module 51, 67, 77 or 79 in FIGS. 3, 5, 7 or 9, respectively.
One alternative for the choice algorithm incorporated in the choice or estimation module of FIGS. 3, 5, 6, 7 and 9 is illustrated by a wideband input test signal s(t)+n(t), whose magnitude is shown in FIG. 10A. The test signal consists of a constant function s(t)=constant, upon which are imposed six noise spikes, each of normalized height=1, as shown. The magnitudes of the five signals s(t) (narrowband, after passage through a lowpass filter), s(t)+n.sub.3 (t) (narrowband), s.sub.+ (t)=s(t)+n.sub.3 (t)+N.sub.1 (t), s.sub. (t)=s(t)+n.sub.3 (t)N.sub.1 (t) and the optimum choice S.sub.1 (t) are formed and shown graphically in FIGS. 10B, 10C, 10D, 10E and 10F, respectively. The phase of the wideband input test signal s(t)+n(t) is shown in FIG. 11A. The phases of the five signals s(t) (narrowband), s(t)+n.sub.3 (t) (narrowband), s.sub.+ (t)=s(t)+n.sub.3 (t)+N.sub.1 (t), s.sub. (t)=s(t)+n.sub.3 (t)N.sub.1 (t) and the optimum choice S.sub.1 (t) are shown graphically in FIGS. 11B, 11C, 11D, 11E and 11F, respectively.
With reference to FIGS. 10D and 10E, note that the magnitude .vertline.s.sub.+ (t).vertline.=.vertline.s(t)+n.sub.3 (t)+N.sub.1 (t).vertline. produces the mostnearlyconstant output signal magnitude for the approximate time ranges 600 .DELTA.t.ltoreq.t.ltoreq.800 .DELTA.t and 900 .DELTA.t.ltoreq.t.ltoreq.1250 .DELTA.t and 1600 .DELTA.t.ltoreq.t.ltoreq.2000 .DELTA.t; and the magnitude .vertline.s.sub. (t) .vertline.=.vertline.s(t)+n.sub.3 (t)N(t).vertline. produces the mostnearlyconstant output signal magnitude for the approximate time ranges 0<t.ltoreq.600 .DELTA.t and 800 .DELTA.t<t<900 .DELTA.t and 1250 .DELTA.t.ltoreq.t.ltoreq.1600 .DELTA.t. The phases of the signals s.sub.+ (t) and s.sub. (t) manifest similar behavior in FIGS. 11D and 11E. Because of this behavior, the optimum choice of output signal, s.sub.+ (t) or s.sub. (t) or s(t)+n.sub.3 (t), will change from time to time in the sampling interval. Often the signal whose magnitude (or phase) is closest to the signal median is determined to be the chosen signal.
This change with time of choice of the optimized output signal, s.sub.+ (t) or s.sub. (t) or s(t)+n.sub.3 (t), is believed to extend to a change with time of choice of the real numbers .alpha., .phi., .psi.1 .psi.2 and/or .psi. in the output function in Eqs. (6, (9))and (10), in order to produce an improved output function. The preceding approach can also be applied to the functions r(t)=S.sub.1 (t), S.sub.2 (t;.phi.,.alpha.), S.sub.3 (t;.phi.,.alpha.)), S.sub.4 (t;.psi.1,.psi.2,.alpha.), or S.sub.5 (t;.psi.,.alpha.) set forth in Eqs. (5), (7), (8), (10) or (11).
The quantitative error measure used to produce the optimized signal S.sub.2 (t), whose magnitude and phase are shown in FIGS. 10F and 11F, respectively, is the displacement from the signal median (magnitude or phase), denoted r.sub.med, of samples of the resulting signal. If {r(t.sub.i)}.sub.i denotes a set of P+1 sampled values of a signal r(t) at a set of consecutive sample times t.sub.0 <t.sub.1 < . . . <t.sub.P, the sampled values are rearranged, by sorting or another appropriate permutation operation .PHI., into a monotonic set {r(t.sub..PHI.i)}.sub.i for which r(t.sub..PHI.i).ltoreq.r(t.sub..PHI.k) whenever 1.ltoreq.i<k.ltoreq.P. The median value of the sampled signal r(t) is determined by the relations ##EQU1## With this choice of quantitative error measure, the choice of s.sub.+ (t) or s.sub. (t) is made for a particular time interval {t1, t2}, based upon which choice produces the instantaneous signal value (magnitude or phase) that is closest to the median value of the input signal s(t)+n.sub.3 (t) over the designated time interval. The median algorithm can be applied to the output functions r(t)=S.sub.1 (t), S.sub.2 (t;.phi.,.alpha.), S.sub.3 (t;.phi.,.alpha.), S.sub.4 (t;.psi.1,.psi.2,.alpha.) or S.sub.5 (t;.psi.,.alpha.) set forth in Eqs. (5), (7), (8), (10) or (11).
Another suitable error measure and associated algorithm, cumulative variation of the signal r(t), is determined by forming the sum ##EQU2## where r(t) represents the signal .vertline.S.sub.2 (t).vertline.=.vertline.s.sub.+ (t).vertline. or .vertline.s.sub. (t).vertline. or .vertline.s(t)+n.sub.3 (t).vertline. and P+1 is the number of distinct time points at which the signal r(t) is sampled in the interval {t.sub.0, t.sub.P }. The signal r(t) that produces the smallest value Var{r(t); t.sub.0,t.sub.P } is the mostnearlyconstant signal from among the candidate signals.
A third suitable error measure and associated algorithm that can be incorporated in the choice modules of FIGS. 3, 5, 6, 7 or 9 determines and compares a least mean square value over a selected time interval {t1, t2} ##EQU3## for the three signals r(t)=.vertline.s.sub.+ (t).vertline., r(t)=.vertline.s.sub. (t).vertline. and r(t)=.vertline.s(t)+n.sub.3 (t).vertline. to determine the appropriate choice of output signal for the apparatus. The value r0 can be set equal to zero, in which case the LMS algorithm produces a minimum power choice for the chosen signal. Alternatively, the value r0 can be chosen to be a pth power mean value of the input signal
Interesting choices here include p=1 and p=2. Other physically meaningful choices for the value r0 can also be made. The cumulative variation algorithm or the least mean squares algorithm discussed here can also be used with any of the embodiments shown in FIGS. 3, 5, 6, 7 and 9; that is, the parameters .alpha., .phi., .psi.1, .psi.2 and/or .psi. in the output functions r(t)=S.sub.1 (t), S.sub.2 (t;.phi.,.alpha.), S.sub.3 (t;.phi.), S.sub.4 (t;.psi.1,.psi.2,.alpha.) or S.sub.5 (t;.phi.,.alpha.) set forth in Eqs. (5), (7), (8),(10) or (11) can be determined by one of these algorithms.
Any of the three preceding error measures and associated algorithms can be applied to the output functions r(t)=S.sub.1 (t), S.sub.2 (t;.phi.,.alpha.), S.sub.3 (t;.phi.,.alpha.), S.sub.4 (t;.psi.1,.psi.2,.alpha.) or S.sub.5 (t;.psi.,.alpha.)) set forth in Eqs. (5), (7), (8), (10) or (11).
Optimization with respect to the magnitude and/or phase of a signal r(t) may be replaced by optimization with respect to the real part or the imaginary part of the signals, with the same quantitative measure of error being used. This approach compares the combination signals Re{s.sub.+ (t)}, Re{s.sub. (t)} and Re{s(t)+n.sub.3 (t)} or the combination signals Im{s.sub.+ (t)}, Im{s.sub. (t)} and Im{s(t)+n.sub.3 (t)} for a chosen quantitative measure of error, such as displacement from the median, least mean squares value, or minimum variation.
Another approach compares the combination signals w .vertline.s.sub.+ (t).vertline.+(1w) phase{s.sub.+ (t)}
and
w .vertline.s.sub. (t).vertline.+(1w) phase{s.sub. (t)}
and w.vertline.s(t)+n.sub.3 (t).vertline.+(1w) phase{s(t)+n.sub.3 (t)}, where w is a selected real or complex constant, and chooses as the optimized output signal the combination with the smallest quantitative error according to a selected error measure.