摘要
为探究基于梅尔频率倒谱系数(Mel-Frequency Cepstral Coefficients,MFCC)和隐马尔可夫模型(Hidden Markov Model,HMM)的语音识别优化方法,首先探讨语音识别系统的基本框架设计,其次分析MFCC特征提取方法,再次引入期望最大化(Expectation Maximization,EM)算法优化HMM参数,最后利用THCHS-30数据集进行实验验证。结果表明,引入EM算法优化HMM,可有效克服传统HMM在复杂语音环境下的识别困难问题,显著提升系统的识别精度和健壮性。
In order to explore the speech recognition optimization method based on Mel-Frequency Cepstral Coefficients(MFCC)and Hidden Markov Model(HMM),the basic framework design of the speech recognition system is first discussed.Secondly,the MFCC feature extraction method is analyzed,and the Expectation Maximization(EM)algorithm is introduced again to optimize HMM parameters.Finally,the THCHS-30 dataset is used for experimental verification.The results show that the introduction of EM algorithm to optimize HMM can effectively overcome the recognition difficulties of traditional HMM model in complex speech environment,and significantly improve the recognition accuracy and robustness of the system.
作者
郭佳淇
张继通
GUO Jiaqi;ZHANG Jitong(Zhengzhou University of Industrial Technology,Zhengzhou 451100,China)
出处
《电声技术》
2024年第10期83-85,共3页
Audio Engineering
