Study On Speech Enhancement With Microphone Arrays

Speech enhancement is a useful forward means to restrain the interference from the received speech signals which are inevitably interfered by noise. Speech enhancement using single microphone can't reduce noise in the environment with strong reverberation and nonstationary noise, so this paper focus on speech enhancement methods using microphone arrays because microphone array systems make full use of information in time\frequency domain and spatial domain.Time Delay Estimation (TDE) and Voice Activity Detection (VAD) are two key parts in most kinds of array speech enhancement methods and their performance directly affect the results of speech enhancement. Independent Component Analysis (ICA) can be used to separate the speech or noise without the knowledge of original signals or transmit functions. If ICA is combined with speech enhancement in array system, the performance must be improved. So the main works in this paper are as follows:(1) Crosspower Spectrum Phase (CSP) is a popular TDE technique and its performance degrades dramatically in the case of low Signal-to-Noise Ratio (SNR) and strong reverberation. So two improved algorithms are proposed in this paper. Simulation results show that the proposed methods outperform the original CSP method.(2) A VAD method for microphone array is proposed. More 'pure' voice can be obtained by using ICA. The VAD result of this 'pure' voice is used as reference for detecting those of all array signals. Simulation results illustrate the validity of the proposed method.(3) Generalized Sidelobe Canceller (GSC), as a basic adaptive beamforming algorithm in speech enhancement, only fits to cancel the coherent noise signals and the performance severely degrades in the real noise and reverberant environment. So a new GSC structure is proposed, which apply ICA with reference (rICA) in Robust Generalized Sidelobe Canceller (RGSC). Experiment results through real noisy signals prove that this new structure is robust for reverberation and any kind of noise, and it is applicable in real environment.