Gaussian核SVM在抗噪语音识别中的应用

Application of support vector machine with Gaussian kernel in noise-robust speech recognition

为提高机器学习的推广能力,解决语音识别系统在噪声环境中识别率变差等问题,采用改进的MFCC语音特征参数,用Gaussian核支持向量机(SVM)作为语音识别网络,对SVM多类分类问题采用"一对一"分类算法,实现了一个汉语孤立词非特定人中等词汇量的抗噪语音识别系统。通过实验,分析了Gaussian核参数和误差惩罚参数C对SVM推广能力的影响。实验结果表明,当工作在不同信噪比情况下,使用最优参数的Gaussion核SVM的识别率比使用RBF神经网络有较大的提高,训练时间能大为缩减,鲁棒性也较好。

To improve the generalization ability of the machine leaming and solve the problem that recognition rates of the speech recognition system become worse in the noisy environment, the improved MFCC speech characters and the support vector machine is used as the recognition network for speech recognition system, a one-against-one method for multi-class support vector machine is utilized and a noise-robust speech recognition system is realized for Chinese isolated words, non-specific person and middle glossary quantity. By experiments, the influences of the kernel parameter γ and the error penalty parameter C on support vector machine＇ s generalization ability are analyzed. Experiments showed that the Gaussian kernel SVM speech recognition system with the best parameters has higher correct recognition rates than ones of using RBF network in different SNRs, and is of shorter training time and much better robustness.