Acoustic noise suppression using regressive adaptive filtering

Experimental field tests dealing with the background acoustic noise in cars under various driving conditions are described. Analysis reveals that there is a high correlation between the acoustic noise in the area facing the driver´s seat and the noise in other locations in the car, which suggests the possibility of using the two-microphones noise cancellation approach. Results show that using a conventional finite-impulse response adaptive filter with the stochastic gradient adaptation algorithm leads to up to 12 dB of noise cancellation in the low end of the noise spectrum; some noise enhancement was noticed at the high end of the spectrum. This problem is discussed, along with a possible solution approach using proper filtering. A limitation of the two-microphones cancellation is that the optimal location of the secondary microphone varies, depending on the driving conditions. A multiple secondary microphones scheme is proposed as a solution. This scheme resulted in further reductions of the residual noise