论文部分内容阅读
针对语音识别系统受噪声干扰识别率急剧下降的问题,通过分析传统的鲁棒语音特征提取方法在语音信号谱估计方面的不足,提出一种在不同信噪比下都具有较好鲁棒性和识别性能的语音特征提取算法.该算法结合多信号分类法(MUSIC)和最小模法(minimum-norm method,MNM)来进行谱估计.接着在移动机器人平台上进行验证实验,结果表明:该算法能有效的提高语音识别率,增强语音识别鲁棒性能.
Aiming at the sharp decline of recognition rate of speech recognition system caused by noise interference, by analyzing the shortcomings of traditional robust speech feature extraction methods in speech signal spectrum estimation, a new speech recognition algorithm with robustness under different SNR Which can be used to estimate the performance of the speech feature extraction algorithm.This algorithm combined with multi-signal classification (MUSIC) and minimum-norm method (MNM) to perform spectral estimation.Finally, the verification experiment on the mobile robot platform shows that the algorithm It can effectively improve the speech recognition rate and enhance the robustness of speech recognition.