关键词: 声学, 自适应子带分解, 自适应抵消, 功率谱密度

Abstract: Extraction of the signals, which can represent sound source information and meanwhile include noise signals, from acoustic sensors is the key technique for sound emission monitoring. To eliminate the noise efficiently and raise the signaltonoise ratio, this paper brings up a dualadaptive noisecancellation algorithm, which includes adaptive subband decomposition algorithm and adaptive noise cancellation algorithm. First of all, to reduce the complexity in computation and realize parallel algorithm, the method of the subband decomposition, which matches the noise powerspectrum density, is adopted for nonuniform signal subband decomposition. Then, to save the computer time, the effective filtering is performed according to the distribution of noise in the subband. The subbands with lownoise or essentially without noise do not need filtering, while the adaptive filtering algorithm is used for the other subbands. The simulation shows that this method can greatly save computer time in comparison with the traditional adaptive noise cancellation method with uniform subbands, and the effect of filtering has been improved.

Key words: acoustics, adaptive subband decomposition, adaptive noise cancellation, powerspectrum density