Active acoustic attenuation system with reduced convergence time

Claim 2

In an acoustic system having an input for receiving an input acoustic wave produced by an acoustic source, and an output for radiating an output acoustic wave, an active attenuation method for attenuating undesirable said output acoustic wave by introducing a cancelling acoustic wave from an output transducer, comprising: sensing said input acoustic wave with an input transducer and providing an input signal; sensing the combined said output acoustic wave and said cancelling acoustic wave from said output transducer with an error transducer and providing an error signal; modeling said acoustic system with an adaptive filter model having a model input from said input transducer and an error input from said error transducer and outputting a correction signal to said output transducer to introduce the cancelling acoustic wave such that said error signal approaches a given value as said adaptive filter model responds to said input signal by updating weights of a coefficient weight vector from an initial nonconverged value in a nonconverged condition of said filter model at turn-on thereof to a converged value in a converged condition of said filter model; reducing convergence time of said filter model at turn-on thereof by starting said weights at a starting value which is closer to said converged value than said initial nonconverged value is to said converged value, said filter model converging to said input signal in a shorter time as said weights change and are updated from said starting value to said converged value; selecting said starting value according to a characteristic of said acoustic source.

Claim 3

The invention according to claim 1 comprising modeling said acoustic system with an adaptive filter model having a transfer function with both poles and zeros.

Claim 4

The invention according to claim 2 comprising modeling said acoustic system with an adaptive recursive filter model.

Claim 5

The invention according to claim 3 comprising modeling said acoustic system with a recursive least mean square filter model.

Claim 6

The invention according to claim 4 comprising adaptively modeling said acoustic path on-line with said adaptive filter model without dedicated off-line pretraining, and also adaptively modeling the feedback path from said output transducer to said input transducer for both broadband and narrowband noise without dedicated off-line pretraining and without a separate model dedicated solely to said feedback path and pretrained thereto.

Claim 7

The invention according to claim 1 comprising providing an auxiliary noise source and introducing auxiliary noise from said auxiliary noise source into said adaptive filter model, such that said error transducer also senses the auxiliary noise from said auxiliary noise source, said auxiliary noise being random and uncorrelated with said input acoustic wave.

Claim 8

The invention according to claim 1 comprising predetermining said starting value of said weights according to one or more sensed parameters of said acoustic source.

Claim 9

The invention according to claim 7 comprising providing a plurality of sets of predetermined starting values of said weights and selecting one of said sets according to given ranges of sensed parameter values.

Claim 10

The invention according to claim 8 wherein one of said sets is from the converged value of said weights of the previous modeling of said acoustic system by said adaptive filter model.

Claim 11

In an acoustic system having an input for receiving an input acoustic wave produced by an acoustic source, and an output for radiating an output acoustic wave, an active attenuation method for attenuating undesirable said output acoustic wave by introducing a cancelling acoustic wave from an output transducer, comprising: sensing said input acoustic wave with an input transducer and providing an input signal; sensing the combined said output acoustic wave and said cancelling acoustic wave from said output transducer with an error transducer and providing an error signal; modeling said acoustic system with an adaptive filter model having a model input from said input transducer and an error input from said error transducer and outputting a correction signal to said output transducer to introduce the cancelling acoustic wave such that said error signal approaches a given value as said adaptive filter model responds to said input signal by updating weights of a coefficient weight vector from an initial nonconverged value in a nonconverged condition of said filter model at turn-on thereof to a converged value in a converged condition of said filter model, said filter model also responding to said input signal as the latter changes in response to changes of said input acoustic wave such that said weights are further updated during further convergence of said model to a new converged value; reducing convergence time of said filter model to a new converged condition in response to changes of said input acoustic wave, by providing new updated weight values according to a sensed parameter of said acoustic source without waiting for convergence of said filter model, such that said filter model converges in a shorter time.

Claim 12

In an acoustic system having an input for receiving an input acoustic wave produced by an acoustic source, and an output for radiating an output acoustic wave, an active attenuation method for attenuating undesirable said output acoustic wave by introducing a cancelling acoustic wave from an output transducer, comprising: sensing said input acoustic wave with an input transducer and providing an input signal; sensing the combined said output acoustic wave and said cancelling acoustic wave from said output transducer with an error transducer and providing an error signal; modeling said acoustic system with an adaptive filter model having a model input from said input transducer and an error input from said error transducer and outputting a correction signal to said output transducer to introduce the cancelling acoustic wave such that said error signal approaches a given value as said adaptive filter model responds to said input signal by updating weights of a coefficient weight vector from an initial nonconverged value in a nonconverged condition of said filter model at turn-on thereof to a converged value in a converged condition of said filter model, said filter model also responding to said input signal as the latter changes in response to changes of said input acoustic wave such that said weights are further updated during further convergence of said model to a new converged value; reducing convergence time of said filter model to a new converged condition in response to changes of said input acoustic wave, by providing new updated weight values according to a sensed parameter of said acoustic source without waiting for convergence of said filter model, such that said filter model converges in a shorter time.

Claim 13

The invention according to claim 10 comprising providing a plurality of sets of predetermined values of weights, and selecting one of said sets according to given ranges of sensed parameter values, without waiting for said filter model to otherwise converge in response to the newly changed input acoustic wave.

Claim 14

The invention according to claim 10 comprising also reducing convergence time of said filter model at turn-on thereof by starting said weights at a starting value which is closer to said converged value than said initial nonconverged value is to said converged value, said filter model converging in a shorter time at turn-on as said weights change and are updated from said starting value to said converged value.

Claim 15

The invention according to claim 10 comprising modeling said acoustic system with an adaptive filter model having a transfer function with both poles and zeros.

Claim 16

The invention according to claim 13 comprising modeling said acoustic system with an adaptive recursive filter model.

Claim 17

The invention according to claim 14 comprising modeling said acoustic system with a recursive least mean square filter model.

Claim 18

The invention according to claim 15 comprising adaptively modeling said acoustic path on-line with said adaptive filter model without dedicated off-line pretraining, and also adaptively modeling the feedback path from said output transducer to said input transducer for both broadband and narrowband noise without dedicated off-line pretraining and without a separate model dedicated solely to said feedback path and pretrained thereto.

Claim 19

The invention according to claim 10 comprising providing an auxiliary noise source and introducing auxiliary noise from said auxiliary noise source into said adaptive filter model, such that said error transducer also senses the auxiliary noise from said auxiliary noise source, said auxiliary noise being random and uncorrelated with said input acoustic wave.