Structure selection algorithm for less musical-noise generation in integration systems of beamforming and spectral subtraction
Yu Takahashi, Yoshihisa Uemura, Hiroshi Saruwatari, Kiyohiro Shikano, Kazunobu Kondo · 2009 IEEE/SP 15th Workshop on Statistical Signal Processing · 2009
In this paper, we propose an appropriate structure selection algorithm for less musical-noise generation in integration methods of microphone array and spectral subtraction. In our previous work, we have analyzed musical-noise reduction structure in integration methods of microphone array and spectral subtraction based on higher order statistics. However, that analysis can be applied to only Gaussian and super-Gaussian noises. In this paper, we extend the analysis into sub-Gaussian noise case. As a result of the extended analysis, we find the fact that an appropriate structure is depending on the type of input noise. Based on this fact, we propose an appropriate structure selection algorithm in integration methods of microphone array and spectral subtraction for less musical-noise generation. The effectiveness of the proposed algorithm are shown via a computer simulation.