基於盲訊號分離語音增強技術之遠距離雜訊語音辨識

Speech Enhancement Technique Based on Blind Source Separation for Far-Field Noisy Speech Recognition

語音辨識在語音處理領域中，為其重要的一項研究領域項目之一，然而語音辨識中的辨識結果會隨著語音所在環境及語音距離而有所影響，在此本論文提出一套高適應性遠距離雜訊語音辨識系統，此系統結合獨立成分分析方法以及子空間語音增強方式，將雜訊語音進一步濾除噪聲並提升語音訊號強度以供辨識。實驗結果顯示，本論文所提出的語音辨識系統，可適用在各種噪聲環境當中，並有效的改善其辨識率，以及將原本信噪比0dB到10dB範圍中的帶噪語音，提升至20dB以上。

Speech recognition is one of the important parts of search field in speech processing. Nevertheless, the speech environment and speech distance will mainly affect the recognition result. In this paper, a high adaptation far-field noise speech recognition system is proposed. This system is combined with the methods of independent component analysis and subspace speech enhancement, and then further filtering the noise of speech to improve the speech quality for recognition. The experimental results show that the proposed system is suitable for several presented noisy environments, and it can effectively improve the recognition rate. For the SNR evaluation, this proposed system can make enhanced speech SNR with 20dB higher than original corrupted speech which ranges from 0dB to 10dB.