Microphone Array Speech Enhancement Based on Filter Bank Generalized Sidelobe Canceller

In order to accelerate the convergence of the microphone array speech enhancement system based on generalized sidelobe canceller (GSC), the input signals of the adaptive module are decomposed into their subbands by filter bank to perform the multichannel adaptive filter (MCAF) in each subband. The columns of the blocking matrix are derived from a prototype vector by cosine modulation, and these columns constitute a series of bandpass filters, which select signals with specific directions of arrival and frequencies. Further exploited by subband decomposition and suitably discarding the low-pass subbands prior to running independent unconstrained adaptive filters in each nonredundant subband, the computational complexity of a GSC implementation is greatly reduced compared to fully adaptive GSC schemes. Finally, based on the introduction of post-filtering technology, it eliminates the incoherent noise. Experiments indicate that the speech enhancement algorithm based on filter bank GSC have higher converging speed and better output signal to noise ratio (SNR) than the fullband GSC.