Adaptive noise cancellation
First Claim
1. A method for removing impulse noise from a circuit electrical signal s(t) that includes a desired signal s0 (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 s0 (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 ω
1, to produce a first noise signal n1 (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 ω
2, that is spaced apart from and does not overlap any part of the first frequency band, to produce a second noise signal n2 (t);
passing the bandpassed signal through a third selected signal processing channel that has a third frequency band, having a central frequency ω
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)+n3 (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 ω
1, ω
2 and ω
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 roll-off frequencies, respectively, to form first, second and third channel output signals, respectively;
forming a combined signal that is a symmetric, homogeneous function HS{n1 (t), n2 (t);
θ
} of degree one from the first and second channel output signals,where the first and second channel output signals have m adjustable parameters θ
=(θ
1, . . . ,θ
m;
m=1 or
2);
forming a linear combination signal LC{HS{n1 (t), n2 (t);
θ
};
s(t)+n3 (t)} of the combined signal HS{n1 (t), n2 (t);
θ
} and the third channel output signal s(t)+n3 (t);
adjusting the parameters θ
whereby the linear combination signal LC{HS{n1 (t), n2 (t);
θ
};
s(t)+n3 (t)} provides a best possible estimate of the desired signal s(t), according to a selected criterion, for an optimal choice θ
=θ
0 of these parameters; and
issuing the linear combination signal LC{HS{n1 (t), n2 (t);
θ
=θ
0 };
s(t)+n3 (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 noise-free signal is assumed to have a representative frequency ω3, but may have a range of frequencies adjacent to this frequency, and is assumed to have substantially zero amplitude for all frequencies ω<ω1 and/or for all frequencies ω>ω2, where ω1 <ω3 <ω2 or ω1 <ω2. An input (noisy) signal is filtered and analyzed in a narrow frequency region surrounding ω=ω1 and/or a narrow frequency region surrounding ω=ω2 to obtain one or two output signal components n1 (t) and/or n2 (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 ω=ω3 to obtain an output signal s(t)+n3 (t) component including the desired signal s(t). A linear combination of signals, such as S(t)=s(t)+n3 (t)±[n1 (t)n2 (1)] 1/2, or S(t)=s(t)+n3 (t)+αexp(-jφ) [n1 (t)n2 (1)]1/2,or S(t)=s(t)+n3 (t)±nk (t) (k=1,2), or S(t)=s(t)+n3 (t)+exp(-jφ)nk (t), or S(t)=s(t)+n3 (t)+[exp(-jψ1)n1 (t)+exp(-jψ2) n2 (t)]/2, or S(t)=s(t)+n3 (t)+α[exp(-jψ) n1 (t)+exp(-jψ) n2 (t)]/2, is formed as a circuit output signal, and the ± sign and/or the multiplier α and/or the phase angles φ, ψ1, ψ2 and ψ 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 n1 (t) and n2 (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 φ , ψ1, ψ2 and ψ are also disclosed.
136 Citations
26 Claims
-
1. A method for removing impulse noise from a circuit electrical signal s(t) that includes a desired signal s0 (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 s0 (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 ω
1, to produce a first noise signal n1 (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 ω
2, that is spaced apart from and does not overlap any part of the first frequency band, to produce a second noise signal n2 (t);passing the bandpassed signal through a third selected signal processing channel that has a third frequency band, having a central frequency ω
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)+n3 (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 ω
1, ω
2 and ω
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 roll-off frequencies, respectively, to form first, second and third channel output signals, respectively; forming a combined signal that is a symmetric, homogeneous function HS{n1 (t), n2 (t);
θ
} of degree one from the first and second channel output signals,where the first and second channel output signals have m adjustable parameters θ
=(θ
1, . . . ,θ
m;
m=1 or
2);forming a linear combination signal LC{HS{n1 (t), n2 (t);
θ
};
s(t)+n3 (t)} of the combined signal HS{n1 (t), n2 (t);
θ
} and the third channel output signal s(t)+n3 (t);adjusting the parameters θ
whereby the linear combination signal LC{HS{n1 (t), n2 (t);
θ
};
s(t)+n3 (t)} provides a best possible estimate of the desired signal s(t), according to a selected criterion, for an optimal choice θ
=θ
0 of these parameters; andissuing the linear combination signal LC{HS{n1 (t), n2 (t);
θ
=θ
0 };
s(t)+n3 (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)
-
-
17. A method for removing impulse noise from a circuit electrical signal that includes a desired signal s0 (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 s0 (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 ω
1, to produce a first noise signal n1 (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 ω
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)+n2 (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 roll-off frequencies, respectively, to form first and second channel output signals, respectively; forming a linear combination signal LC{n1 (t), s(t)+n2 (t);
θ
} of the first channel output signal n1 (t) and the second channel output signal s(t)+n2 (t), where this linear combination has m adjustable parameters θ
=(θ
1, . . . , θ
m;
m=1 or
2);adjusting the parameters θ
whereby the linear combination signal LC{n1 (t), s(t)+n2 (t);
θ
} provides a best possible estimate of the desired signal s(t), according to a selected criterion, for optimal choices θ
=θ
0 of these parameters;and issuing the linear combination signal LC{n1 (t), s(t)+n2 (t);
θ
=θ
0 } as an estimate of the desired signal s(t). - View Dependent Claims (18, 19, 20, 21, 22, 23, 24, 25, 26)
-
Specification