论文部分内容阅读
在研究传统语音特征参数线性预测倒谱系数(LPCC)和梅尔频率倒谱系数(MFCC)的基础上,加入基于人耳听觉特性的Bark子波滤波器组所提取的特征参数,来共同组成特征集。同时将基于遗传算法的相关性特征算法将特征集进行优化,分别采用贝叶斯和径向基神经网络算法进行语音识别。实验结果表明本方法与传统的LPCC和MFCC方法相比,平均识别率分别提高了4.66%和3.5%,最佳达到98.1%的识别率。
Based on the study of linear predictive cepstrum coefficients (LPCC) and Mel-Frequency Cepstral Coefficients (MFCC) of traditional speech feature parameters, the feature parameters extracted from the Bark wavelet filter bank based on human auditory characteristics are added Feature set. At the same time, the correlation feature algorithm based on genetic algorithm is used to optimize the feature set, and the speech recognition is carried out using Bayesian and RBF neural network respectively. Experimental results show that compared with the traditional LPCC and MFCC methods, the average recognition rate of this method is improved by 4.66% and 3.5%, respectively, and the best recognition rate is 98.1%.