首页 | 本学科首页   官方微博 | 高级检索  
     

语音情感特征提取和识别的研究与实现
引用本文:詹永照,曹鹏. 语音情感特征提取和识别的研究与实现[J]. 江苏大学学报(自然科学版), 2005, 26(1): 72-75
作者姓名:詹永照  曹鹏
作者单位:江苏大学计算机科学与通信工程学院,江苏,镇江,212013;江苏大学计算机科学与通信工程学院,江苏,镇江,212013
基金项目:国家自然科学基金资助项目(60273040)江苏省高校自然科学基金资助项目(02KJB520003)
摘    要:针对语音情感识别的实时性与可用性问题,提出了利用汉明窗提取语音信号中情感特征的方法,从说话者中采集带有快乐、愤怒、惊奇、悲伤情感的语句,并从语句中提取8个情感特征参数.同时提出采用贡献分析法确定情感特征参数的权值,利用加权欧氏距离模板匹配识别语音情感.实验表明该方法识别率有提高,更适合实时识别的应用。

关 键 词:语音信号  汉明窗  情感特征分析  加权欧氏距离  情感识别
文章编号:1671-7775(2005)01-0072-04
修稿时间:2004-09-14

Research and implementation of emotional feature extraction and recognition in speech signal
ZHAN Yong-zhao,CAO Peng. Research and implementation of emotional feature extraction and recognition in speech signal[J]. Journal of Jiangsu University:Natural Science Edition, 2005, 26(1): 72-75
Authors:ZHAN Yong-zhao  CAO Peng
Abstract:For solving the usability problem for real time emotion recognition in speech, an emotional feature extraction method is presented by using Hamming window. The eight emotional features are extracted from speech signal that contains emotion such as happiness, anger, surprise and sadness. A speech emotion recognition method is also presented based on weighted Euclidean distance template matching, in which the weights of features are calculated by the method of contribution analysis. The experiments show that the proposed methods can improve recognition rate and are more suitable for real time emotion recognition.
Keywords:speech signal  hamming window  emotional feature analysis  weighted Euclidean distance  emotion recognition
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号