Research Of Microphone Array Based Speech Denoising Method

Electronic speech noise cancellation systems are widely used in people's dailylife. Microphone-array based noise cancellation systems have become significantparts of them. They are successfully integrated into car mobile phone set,voice-control system and hearing aid system. Not only expected useful signals, butbackground noise, room reverberant and far-end echoes are included in acquiredsignals. Such effects greatly degrade the intelligibility of speech and recognition rateof Automated Speech Recognition system (ASR). A General Singular ValueDecomposition (GSVD) based Multi-microphone speech denoising algorithm isgiven in this paper. And we also give key implementations on Field ProgrammableGate Array (FPGA). Some in-depth works have been done in several aspects.1. Microphone array based speech noise cancellation systems have beenanalyzed and compared. A GSVD based optimal denoising algorithmhas been proposed and applied in our system.2. We discussed time-delay estimation between microphone pairs.Several algorithms has been compared in this paper. We proposedNLMS algorithm based microphone pairs time-delay estimationadaptive method.3. FPGA based system has been implemented by using Verilog HDL, IPreuse, paralleled computing method and CORDIC algorithm whichimprove efficiency of implementations.