Design of speech noise reduction system for handheld mobile devices

摘要

There has been some adaptive algorithms used in the microphone array speech enhancement technology, but most of them had shortcomings like program back substitution, slow convergence and smaller voice. In our study, an adaptive noise reduction algorithm for microphone array is proposed based on an improved recursive least-square methodology. Using low complexity QR-RLS algorithm based on inverse Givens rotation, a handheld mobile device noise reduction system is designed, which can reduce the number of microphone array to two, eliminate large amount of root computation, and make the original sound without distortion and weakened. Moreover, when SNR is small, the amplitude of the main volume still unchanged, the noise signal frequency is close to zero. Experimental results show that our algorithm is effective.