Normalized frequency domain LMS adaptive filter

Claim 3

The method of claim 1 further comprising the step of iterating steps (ii) and (iii) for successive samples of the input signal taken over an observation time interval.

Claim 4

The method of claim 2 wherein said step (ii) of adaptively determining an estimate of the input spectral power comprises the step of recursively determining the estimate in dependence on the present magnitude of the input signal in the frequency bin and the spectral power estimate calculated for the prior iteration.

Claim 5

The method of claim 3 wherein said estimate for the (n+1) sample is determined in accordance with the recursive relationship σ.sup.2 (n+1)=(1-α)σ.sup.2 (n)+α|x(n)|.sup.2 where σ 2 (n+1) and σ 2 (n) represents the estimate of the input spectral power in the respective frequency bin for the nth and (nth+1) sample, α is a weighting constant determining the degree to which the estimate for the preceding sample affects the estimate for the current sample, and x(n) represents nth sample of the input data signal for the respective frequency bin.

Claim 6

The method of claim 2 wherein said step (ii) of adaptively determining an estimate of the input spectral power comprises the step of averaging the squares of the magnitudes of the preceding N samples of the input signal in the corresponding frequency bin.

Claim 7

The method of claim 2 wherein said step (ii) of adaptively determining an estimate of the input spectral power comprises the step of averaging the squares of the magnitudes of the current and preceding N-1 samples of the input signal in each said frequency bin.

Claim 8

A frequency domain Least Main Square adaptive filter, comprising: means for respectively transforming an input signal and a reference signal into the frequency domain; means for filtering the transformed input signal by an adaptive filter weight; means for comparing the filtered, transformed input signal and the transformed reference signal to provide an error signal; means for feeding said error signal back to said filtering means for adaptively adjusting said filter weight as a function of said error signal and a feedback coefficient; and means for determining said feedback coefficient, comprising: (i) means for fixing the value of a feedback constant; (ii) means for adaptively determining estimates of the spectral power of the input signal at each frequency bin of the transformed input signal; and (iii) means for normalizing said feedback constant by said power estimates within each frequency bin to provide individual feedback coefficients for each of the frequency bins.

Claim 9

A frequency domain Least Main Square adaptive filter, comprising: means for respectively transforming an input signal and a reference signal into the frequency domain; means for filtering the transformed input signal by an adaptive filter weight; means for comparing the filtered, transformed input signal and the transformed reference signal to provide an error signal; means for feeding said error signal back to said filtering means for adaptively adjusting said filter weight as a function of said error signal and a feedback coefficient; and means for determining said feedback coefficient, comprising: (i) means for fixing the value of a feedback constant; (ii) means for adaptively determining estimates of the spectral power of the input signal at each frequency bin of the transformed input signal; and (iii) means for normalizing said feedback constant by said power estimates within each frequency bin to provide individual feedback coefficients for each of the frequency bins.

Claim 10

The filter of claim 7 wherein said filter operates iteratively on successive samples of the input signal over an observation time interval.

Claim 11

The filter of claim 8 wherein said means for determining estimates of the input spectral power comprises means for recursively determining said estimates in dependence on the present magnitude of the input signal in the respective frequency bin and the spectral power estimate determined for the prior estimate.

Claim 12

The filter of claim 8 wherein said means for determining said estimates comprises means for averaging the squares of the magnitudes of the preceding N samples of the input signal in each said frequency bin to provide a mean average power estimate.

Claim 13

The filter of claim 8 wherein said means for determining said estimates comprises means for averaging the squares of the magnitudes of the current and preceding N-1 samples of the input signal in each said frequency bin to provide a mean average power estimate.