摘要
随着科学技术的发展,越来越多的情感识别技术被应用于智能语音交互领域。然而,目前的情感识别技术存在着识别率不高、识别效果不稳定的现象,为了得到更好的识别效果,研究对语音信号的预处理方式进行探讨与优化。在此基础上,研究提出基于混合梅尔频率倒谱系数(MFCC-M)的语言情感识别算法,对不同频段下的语音情感识别算法进行优化,并使用Fisher比准则将优化算法进行结合。在算法比较实验中,所提算法平均识别率最高,为0.97,相较于其他3种识别算法,所提算法的平均识别率分别提高了7.8%、12.8%和16.9%。此外,比较不同算法的上下限差值,所提算法的差值最低,验证了该算法的稳定性。研究为智能语音交互提供了新的改进方向,并为语音情感分析基础的优化提供了数据支持。
With the development of science and technology,more and more emotion recognition technologies are applied in the field of intelligent speech interaction.However,the current emotion recognition technologies have the phenomena of low recog-nition rate and unstable recognition effect.In order to get better recognition effect,the preprocessing method of speech signal is discussed and optimized.On this basis,a language emotion recognition algorithm based on Mel-frequency cepstral coefficients-max(MFCC-M)is proposed to optimize speech emotion recognition algorithm in different frequency bands.The Fisher-ratio criterion is used to combine the optimization algorithm.In the algorithm comparison experiment,the proposed algorithm has the highest average recognition rate,it reaches O.97.Compared with the other three recognition algorithms,the average recog-nition rate of the proposed algorithm is increased by 7.8%,12.8%and 16.9%,respectively.In addition,by comparing the difference between upper and lower limits of different algorithms,the proposed algorithm has the lowest difference,which veri-fies the stability of the proposed algorithm,The research provides a new direction for the improvement of intelligent speech in-teraction and data support for the optimization of the basis of speech sentiment analysis.
作者
张贝贝
田甜
ZHANG Beibei;TIAN Tian(College of Information Technology,Shanghai Jian Qiao University,Shanghai 201306,China;College of Art Design,Shanghai Jian Qiao University,Shanghai 201306,China)
出处
《微型电脑应用》
2025年第5期300-304,共5页
Microcomputer Applications
关键词
人机交互
情感识别
语音信号
梅尔频率倒谱系数
Fisher比准则
human-computer interaction
emotion recognition
speech signal
Mel-frequency cepstral coefficient
Fisher-ratio criterion
