摘要： 提出一种用于语音识别的鲁棒特征提取算法。该算法基于子带主频率信息，实现子带主频率信息与子带能量信息相结合，在特征参数中保留语谱中子带峰值位置信息。使用该算法设计抗噪孤立词语音识别系统，分别在白高斯噪声和背景语音噪声环境下，与传统特征算法做多种信噪比对比实验。试验结果表明该特征算法在2种噪声环境下的识别率有不同程度提高，具有良好的噪声鲁棒性。

关键词: 语音识别, 特征参数, 子带主频率, 噪声鲁棒性

Abstract: This paper proposes a robust feature extraction algorithm for speech recognition. This algorithm is based on the subband dominant frequency information. It combines the information of subband dominant frequency and the subband power as the new feature. So the information of the positions of subband spectrum peaks is saved. A robust isolated word speech recognition system based on this algorithm is designed. Experiments are conducted to compare the proposed algorithm with traditional feature extraction algorithms under different levels of babble noise and white gauss noise. The results indicate that the recognition accuracy of this system has been improved at some degrees under these noise conditions.

Key words: speech recognition, feature parameter, subband dominant frequency, noise robustness

中图分类号: 

• TN912.34