Active noise control system incorporating psychoacoustic and spectrum-tuning features.

摘要

Acoustic noise problem is gaining more and more attention in modern society. Traditionally, passive noise control devices are used to block the undesired sound. However, they are inconvenient and costly in some situations. Instead, active noise control (ANC) technique can attenuate the noise in a more flexible and more effective way.;ANC technique works on the principal of acoustic superposition with electrically controlled loudspeaker(s) sending out anti-noise signal to cancel out the undesired noise in a target zone. The core component of ANC system is the adaptive filter, which updates the filter coefficients to control the anti-noise sent out by loudspeaker(s).;It should be noted that the ultimate goal of ANC is to minimize the annoyance brought by environmental noise to human being. Therefore human hearing characteristics are important factors to improve ANC performance in term of human perception. Psychoacoustics focuses on the study of human perception of sound by objective models. In this dissertation, psychoacoustic considerations are incorporated in ANC systems in two ways. Noise weightings are included in ANC system considering the non-uniform sensitivity of human hearing system. A new ANC architecture is proposed to give listeners flexibility to adjust the spectrum of residual noise considering individual discrepant preferences. In the first scheme, two typical noise weightings, A-weighting and ITU-R 468 noise weighting, are incorporated in the ANC system based on filtered-error least mean square (FELMS) structure. Instead of sound pressure level (SPL), psychoacoustic metrics are utilized to evaluate the noise attenuation performance. In the second approach, we propose a spectrum-tuning active noise control (STANC) structure which could tune the noise spectrum with a tuning filter. In the mean time, the change of tuning filter has no influence on system adaptation, which enssures the system stability and makes online tuning possible.;Conventional ANC system always works on linear assumption. However, nonlinearity in realistic system is another problem which may degrade ANC performance. A new adaptive algorithm based on kernel trick is utilized here for nonlinear ANC system. Simulations show that this system has similar performance as conventional ANC in linear condition but outperforms conventional one in nonlinear situation.