论文部分内容阅读
提出了利用偶数帧段输入隐马尔可夫模型（ＨＭＭ）提高在噪声环境下汉语连续语音识别系统鲁棒性的方法，并提出了对于传统谱相减降噪技术的修改方法。实验结果表明，本文的方法能有效地提高噪声背景下汉语连续语音识别系统的性能。
A method to improve the robustness of Chinese continuous speech recognition system under noisy environment by using HMM (Hidden Markov Model) is proposed. A modified method of traditional spectral subtraction and noise reduction is proposed. Experimental results show that the proposed method can effectively improve the performance of Chinese continuous speech recognition system under noisy background.