一种改进的特征提取方法在语音识别中的应用

针对梅尔频率倒谱系数(MFCC)参数在噪声环境中语音识别率下降的问题,提出了一种基于耳蜗倒谱系数(CFCC)的改进的特征参数提取方法.提取具有听觉特性的CFCC特征参数;运用改进的线性判别分析(LDA)算法对提取出的特征参数进行线性变换,得到更具有区分性的特征参数和满足隐马尔可夫模型(HMM)需要的对角化协方差矩阵;进行均值方差归一化,得到最终的特征参数.实验结果表明:提出的方法能有效地提高噪声环境中语音识别系统的识别率和鲁棒性.

Application of an improved feature extraction method in speech recognition

In order to solve the problem that the speech recognition rate in noise environment is used in Mel frequency cepstral coefficients(MFCC),an improved feature extraction method based on cochlear filter cepstral coefficient(CFCC)is proposed. First,extract the characteristic parameters of CFCC has the characteristics of hearing;then,using an improved linear discriminant analysis(LDA)linear transformation on the characteristic parameters extracted,get more distinguishing characteristic parameters of the diagonalization of the covariance matrix hidden Markov model(HMM)and meet the needs;finally,the mean variance normalization,get the characteristic parameters of the final. The experimental results show that the proposed method can effectively improve the recognition rate and robustness of speech recognition system in noisy environment.