Speech presence detector based on sparse time-random signal samples

• US 5,822,726 A
• Filed: 01/31/1995
• Issued: 10/13/1998
• Est. Priority Date: 01/31/1995
• Status: Expired due to Fees

• Current Assignee: Motorola, Inc. (Motorola Solutions, Inc.)
• Original Assignee: Motorola, Inc. (Motorola Solutions, Inc.)
• Inventors: Smallwood, Ralph Douglas, Taylor, Matthew Whiting
• Primary Examiner(s): Hudspeth, David R.
• Assistant Examiner(s): Smits, Talivaldis Ivars
• Application Number:
  US08/381,675
• Time in Patent Office: 1,351 Days
• Field of Search: 395/2.19, 395/2.23, 395/2.24, 395/2.37, 395/2.42, 381/46, 704/210, 704/214, 704/215, 704/228, 704/233
• US Class Current: 704/233
• CPC Class Codes: G10L 25/78 Detection of presence or absence of voice signals switching of direction of transmission by voice frequency in two-way loud-speaking telephone systems H04M9/10
• CPC Class Codes: H04B 1/46 by voice-frequency signals; by pilot signals

Abstract

A speech presence detector (101) includes a circuit input (105) to couple to a source of input signals. A control signal generator (113) generates control signals. A sampling circuit (112) samples the input signals as a function of the control signal. A processor (116, 616, or 916) is coupled to the sampling circuit to receive the signal samples. The processor is responsive to the sampled signals to identify a background noise threshold and detect the presence of speech. According to one embodiment, the method and apparatus employ a random control signal. According to another embodiment, the average frequency of the control signal is below the Nyquist rate. According to another embodiment, the processor is responsive to the received signal to count the number of samples within a predetermined amplitude range and identifies a background noise threshold from the number of samples which are in the predetermined range.

Claims

1. A speech presence detector, comprising:

a circuit input to couple to a source of input signals;

a signal generator to generate a random control signal at a control signal output;

a sampling circuit coupled to the circuit input to receive the input signals, the sampling circuit including a control input coupled to the control signal output to receive the random control signal and generating samples corresponding to the input signals as a function of the random control signal; and

a processor coupled to the sampling circuit to receive the samples, the processor responsive to the samples to identify a background noise threshold and the processor detecting presence of a speech signal in the input signals when the input signals exceed the background noise threshold.

2. The speech presence detector as defined in claim 1, wherein the processor compares instantaneous samples output by the sampling circuit to the background noise threshold to detect the presence of the speech signal.

3. The speech presence detector as defined in claim 1, further including an audio transducer, wherein the circuit input is coupled to the audio transducer and the random control signal has an average frequency that is within a frequency range of a human speech signal.

4. The speech presence detector as defined in claim 3, wherein the average frequency of the random control signal is less than 1000 Hz.

5. The speech presence detector as defined in claim 3, wherein the processor includes a bin associated with a predetermined amplitude range of input signals, the bin storing a value that is a count of samples having an amplitude in the predetermined range of the input signals.

6. The speech presence detector as defined in claim 5, further including a conditioning circuit generating digital representations corresponding to the input signals.

7. The speech presence detector as defined in claim 6, wherein a plurality of bins are used to generate a histogram, and wherein each of said bins is a count representing a number of samples in a respective predetermined range.

8. The speech presence detector as defined in claim 7, wherein the number of samples used to generate the background noise threshold is limited to a predetermined number.

9. The speech presence detector as defined in claim 8, wherein the conditioning circuit includes an A/D converter, and the sampling circuit is coupled to the A/D converter.

10. The speech presence detector as defined in claim 9, wherein the processor includes a bin associated with a subset of output levels of the A/D converter, and the background noise threshold is set by the count in the bin.

11. The speech presence detector as defined in claim 10, wherein the count of the bin associated with lowest amplitude signal samples is the only bin count used by the processor to determine a speech signal threshold.

12. The speech presence detector as defined in claim 10, wherein said bin count is the zeroth bin.

13. The speech presence detector as defined in claim 1,

wherein the control signal has a constant frequency and the signal generator is randomly enabled to generate the control signal and disabled to stop generation of the control signal.

14. An audio circuit comprising:

a microphone for converting audible signals to input signals;

a signal generator generating a control;

a sampling circuit coupled to the microphone and to the signal generator, the sampling circuit sampling the input signals as a function of the control signal and outputting a sample signal responsive thereto;

a processor coupled to the sampling circuit to receive the sample signal and responsive thereto to identify a background noise threshold, the processor identifying the presence of a speech signal in the input signals when the input signals exceed the background noise threshold; and

an A/D converter converting the input signals to digital representations, the A/D converter having output levels, and the sampling circuit being coupled to an output of the A/D;

wherein the processor includes a predetermined bin storing a value representing a count of the number of samples having an amplitude in a predetermined subset of the output levels of the A/D converter, and the bin count in the predetermined bin is the only bin count used by the processor to generate the background noise threshold.

15. The audio circuit as defined in claim 14, wherein the predetermined bin is the zeroth bin.

16. A telephone comprising:

a microphone for inputting audio signals;

a transmitter including an A/D converter coupled to the microphone, the A/D converter to convert the audio signals into digital representations;

an antenna coupled to an output of the transmitter

a controller generating a random control signal;

a sampling circuit coupled to the A/D converter to receive the digital representations and coupled to the controller to sample the digital representations according to the random control signal and generating signal samples; and

a processor coupled to the sampling circuit, the processor counting signal samples within a predetermined amplitude range and identifying a background noise threshold from a count of random samples which are in the predetermined amplitude range, and the processor detecting a speech signal when the random samples exceed the background noise threshold, and the processor coupled to the transmitter to enable the transmitter to output signals to the antenna when the speech signal is detected and to disable the transmitter from outputting signals to the antenna a predetermined time after the speech signal is no longer present in the digital representations.

17. A method of detecting the beginning of speech in input signals to an audio circuit, the method comprising the steps of:

randomly sampling the input signals to produce samples;

generating a background noise threshold from the samples;

comparing the input signals to the background noise threshold; and

detecting the beginning of speech when the input signals exceed the background noise threshold.

18. The method as defined in claim 17, further including:

counting a number of samples only in a predetermined amplitude range; and

identifying a current speech threshold from the number of samples in said predetermined amplitude range.

19. The method as defined in claim 18, further including:

generating the current speech threshold on a periodic basis;

holding the current speech threshold until a new speech threshold is generated; and

comparing the input signals with the current speech threshold until the new speech threshold is generated.