Adaptive microphone array-based filter in the speech enhancement

摘要

This study applied microphone array technology to develop an adaptive hybrid filter for enhancing speech signals. In the past decade, numerous studies have confirmed the excellent performance of microphone array technology in speech source localization and noise reduction. The proposed hybrid filter applies adaptive wavelet filter and spectral subtraction method for signal processing before filtering the signals by microphone array. Both the spectral subtraction method and adaptive wavelet thresholding method are highly effective for signal denoising. However, spectral subtraction method is ineffective for low-SNR signals whereas adaptive wavelet thresholding method is ineffective for high-frequency signals. The filtering behavior between adaptive wavelet filter and spectral subtraction method is controlled by a feedforward fuzzy neural network. Experimental results reveal that the proposed filter outperforms other methods of speech signal enhancement.