噪声环境下的语音识别性能研究

在变强噪音的情况下，语音识别的正确率会迅速下降；当噪声较强并且强度不断发生变化的时候，端点检测是一个难题；提出了两种方法保证噪声较强而且强度不断发生改变情况下的语音识别率：基于LPC距离的端点检测算法和白适应的抗噪语音特征参数提取算法；经过实验证明，采用了这两种不同于传统算法的新方法后，变强噪音情况下的语音识别率得到了明显的改善。

Research of Speech Recognition under Noise

The correct rate of speech recognition will go down severely under changeable strong noise. Endpoint detection will become difficult when the noise is strong and changeable. Two methods to improve the recognition rate of the speech recognition under the strong and changeable noise is proposed: endpoint detection algorithm based on LPC distance and adaptive algorithm of anti-noise voice parameter distilling. It has been proved in the experiment that the recognition rate under strong and changeable was improved after the two method different from the traditional algorithm was adopted.