Adaptive noise cancellation

• US 5,465,413 A
• Filed: 03/05/1993
• Issued: 11/07/1995
• Est. Priority Date: 03/05/1993
• Status: Expired due to Fees

• Current Assignee: Trimble Navigation Limited (Trimble Inc.)
• Original Assignee: Trimble Navigation Limited (Trimble Inc.)
• Inventors: Enge, Per, Westfall, Brian, Farmer, Dominic
• Primary Examiner(s): Urban, Edward F.
• Assistant Examiner(s): WISLER, MARK
• Application Number:
  US08/027,004
• Time in Patent Office: 977 Days
• Field of Search: 375/99, 375/101-103, 455/296, 455/303-307
• US Class Current: 455/307
• CPC Class Codes: H04B 1/123 using adaptive balancing or compensation means adaptive filter circuits and algorithms H03H

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)±n_k (t) (k=1,2), or S(t)=s(t)+n₃ (t)+exp(-jφ)n_k (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.

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; 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).

2. The method of claim 1, further comprising the step of choosing as said selected criterion minimization of the difference between the present value

LC{HS{n₁ (t), n₂ (t);

θ

};

s(t)+n₃ (t)}

and the median value

Med{s(t)+n₃ (t);

t₁,t₂ }

with respect to said parameters θ

over a selected time interval t₁ ≦

t≦

t₂, where Med denotes the median value of a set of sampled values of said third output channel signal s(t)+n₃ (t) over the selected time interval.

3. The method of claim 1, further comprising the step of choosing as said selected criterion minimization of the value

LMS{LC{HS{n₁ (t), n₂ (t);

θ

};

s(t)+n₃ (t)};

t₁, t₂ }

with respect to said parameters θ

over a selected time interval t₁ ≦

t≦

t₂, where LMS denotes the least mean squares variation of the function r(t)=LC{HS{n₁ (t), n₂ (t);

θ

};

s(t)+n₃ (t)} over the selected time interval.

4. The method of claim 3, wherein minimization of said value LMS{LC{HS{n₁ (t), n₂ (t);

θ

};

s(t)+n₃ (t)};

t₁, t₂ } is implemented by minimization of the value of the integral ##EQU4## with respect to said parameters θ

, where r₀ is a selected constant.

5. The method of claim 4, further comprising the step of selecting said constant r₀ =0.

6. The method of claim 4, further comprising the step of selecting said constant r₀ to be the value ##EQU5## where p is a selected positive constant.

7. The method of claim 1, further comprising the step of choosing as said selected criterion minimization of the value

Var{LC{HS{n₁ (t), n₂ (t);

θ

};

s(t)+n₃ (t)};

t₁, t₂ }

with respect to said parameters θ

over a selected time interval t₁ ≦

t≦

t₂, where Var denotes the cumulative variation of the function r(t)=LC{HS{n₁ (t), n₂ (t);

θ

};

s(t)+n₃ (t)} over the selected time interval.

8. The method of claim 1, further comprising the steps of choosing as said linear combination signal LC and said homogeneous symmetric function HS the combination

LC{HS{n₁ (t), n₂ (t);

θ

};

s(t)+n₃ (t)}=s(t)+n₃ (t)+θ

{n₁ (t) n₂ (t)}^1/2,

where m=1 and said parameter θ

has three possible values, θ

=±

1 and θ

=0.

9. The method of claim 1, further comprising the steps of choosing as said linear combination signal LC and said homogeneous symmetric function HS the combination

LC{HS{n₁ (t), n₂ (t);

θ

};

s(t)+n₃ (t)}=s(t)+n₃ (t)+θ

{n₁ (t) n₂ (t)}^1/2,

where m=1 and said parameter θ

=exp (-jφ

) and φ

is a real number (0≦

φ

<

2π

).

10. The method of claim 9, wherein said phase angle φ

is determined by the steps of;

determining a phase Φ

1 associated with said signal n₁ (t);

determining a phase Φ

2 associated with said signal n₂ (t); and

determining said phase angle φ

according to the relations

φ

=b{ω

₃ -(ω

₁ +ω

₂)/2},

b=(Φ

2-Φ

1)/ (ω

₂ -ω

₁).

11. The method of claim 1, further comprising the step of choosing as said linear combination signal LC and said homogeneous symmetric function HS the combination

LC{HS{n₁ (t), n₂ (t);

θ

};

s(t)+n₃ (t)}=s(t)+n₃ (t)+θ

{n₁ (t)+n₂ (t)}/2,

where m=1 and said parameter θ

has two possible values, θ

=1 and θ

=-1.

12. The method of claim 1, further comprising the step of choosing as said linear combination signal LC and said homogeneous symmetric function HS the combination

LC{HS{n₁ (t),n₂ (t);

θ

};

s(t)+n₃ (t)}=s(t)+n₃ (t)+θ

{n₁ (t)+n₂ (t)}/2,

where said parameters θ

=α

exp (-jψ

) and α and

ψ

are real numbers with 0≦

ψ

<

2π

.

13. The method of claim 1, wherein m=2, further comprising the step of choosing as said linear combination signal LC and said homogeneous symmetric function HS the combination

LC{HS{n₁ (t),n₂ (t);

θ

};

s(t)+n₃ (t)}=s(t)+n₃ (t)-{exp(-jψ

1) n₁ (t)+exp(-jψ

2) n₂ (t)}/2,

where said two parameters θ

1=exp (-jψ

1) and θ

2=exp(-jψ

2), and ψ

1 and ψ

2 are real numbers lying in the interval 0≦

ψ

i<

2π

(i=1,2).

14. The method of claim 13, wherein said phase angles ψ

1 and ψ

2 are determined by the steps of;

determining a phase Φ

1 associated with said signal n₁ (t);

determining a phase Φ

2 associated with said signal n₂ (t); and

determining said phase angles ψ

1 and ψ

2 according to the relations

ψ

1=b(ω

₃ -ω

₁),

ψ

2=b(ω

₃ -ω

₂),

b=(Φ

2-Φ

1)/(ω

₂ -ω

₁).

15. The method of claim 1, further comprising the steps of:

choosing as said homogeneous symmetric function HS the function HS(n₁,n₂ ;

θ

1,θ

2)={A (θ

1 n₁)^h +B (θ

2 n₂)^h }^1/h exp{C(θ

2 n₂ /θ

1 n₁)^k },

where h and k are selected real, non-zero numbers and A, B, and C are selected numbers; and

choosing as said linear combination signal LC the signal LC{HS{n₁ (t),n₂ (t);

θ

};

s(t)+n₃ (t)}=s(t)+n₃ (t)++θ

HS{n₁ (t), n₂ (t)},

where said parameters θ

=α

exp (-jφ

) and α and

φ

are real numbers (0≦

φ

<

2π

).

16. The method of claim 1, further comprising the steps of:

passing said circuit input signal through a second bandpass filter with a second selected frequency passband; and

shifting the output signal of the second bandpass filter by a selected fourth frequency shift to produce a frequency shifted bandpassed signal that is passed to said first bandpass filter.

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).

18. The method of claim 17, further comprising the step of choosing as said selected criterion minimization of the difference between the present value

LC{n₁ (t), s(t)+n₂ (t);

θ

}

and the median value

Med{ s(t)+n₂ (t);

t₁, t₂ }

with respect to said parameters θ

over a selected time interval t₁ ≦

t≦

t₂, where Med denotes the median value of the set of sampled values of the function r(t)=s(t)+n₂ (t) over the selected time interval.

19. The method of claim 17, further comprising the step of choosing as said selected criterion minimization of the value

LMS{ LC{n₁ (t), s(t)+n₂ (t);

θ

};

t₁, t₂ }

with respect to said parameters θ

over a selected time interval t₁ ≦

t≦

t₂, where LMS denotes the least mean squares variation of the function r(t)=LC{n₁ (t), s(t)+n₂ (t);

θ

} over the selected time interval.

20. The method of claim 19, wherein minimization of said value LMS}LC}n₁ (t),s(t)+n₂ (t);

θ

} is implemented by minimization of the value of the integral ##EQU6## with respect to said parameters θ

, where r₀ is a selected constant.

21. The method of claim 20, further comprising the step of selecting said constant r₀ =0.

22. The method of claim 20, further comprising the step of selecting said constant r₀ to be the value ##EQU7## where p is a selected positive constant.

23. The method of claim 17, further comprising the step of choosing as said selected criterion minimization of the value Var{LC{n₁ (t), s(t)+n₂ (t);

θ

};

t₁, t₂ }

with respect to said parameters θ

over a selected time interval t₁ ≦

t≦

t₂, where Var denotes the cumulative variation of the function r(t)=LC{n₁ (t), s(t)+n₂ (t);

θ

} over the selected time interval.

24. The method of claim 17, further comprising the step of:

choosing as said linear combination signal LC the signal LC{n₁ (t), s(t)+n₂ (t);

θ

}=s(t)+n₂ (t)+θ

n₁ (t),

where m=1, said parameter θ

1=exp (-jφ

) and φ

is a real number (0≦

φ

≦

2π

).

25. The method of claim 17, further comprising the steps of:

passing said bandpassed signal through a third selected filter, where the third filter has frequency support contained in a third frequency band, having a central frequency ω

₃, that is spaced apart from and does not overlap any part of, said first frequency band or said second frequency band, to produce a third noise signal n₃ (t);

determining a phase Φ

1 associated with said signal n₁ (t);

determining a phase Φ

3 associated with said signal n₃ (t); and

determining said phase angle φ

according to the relations

φ

=b(ω

₂ -ω

₁),

b=(Φ

3 -Φ

1)/(ω

₃ -ω

₁).

26. The method of claim 17, further comprising the steps of:

passing said circuit input signal through a second bandpass filter with a second selected frequency passband; and

shifting the output signal of the second bandpass filter by a selected fourth frequency shift to produce a frequency shifted bandpassed signal that is passed to said first bandpass filter.