该文实现了线性判别分析在汉语连续语音识别系统中的应用。通过将多帧原始特征联合后进行特征选择,特征之间的帧间相关性得到了有效的利用,从而提高了语音的识别率。实验结果表明,系统误识率下降16.90%。 This paper realizes the application of linear discriminant analysis in Chinese continuous speech recognition system. By combining the original features of multi-frame feature selection, the inter-frame correlation between features is effectively used, which improves the speech recognition rate. Experimental results show that the system error rate decreased by 16.90%.