摘要
语音情感识别是人工智能的重要研究领域之一,特征参数提取的准确性直接影响识别的效果。分析了发音持续时间、平均振幅、基音频率,第一共振峰和Mel频率倒谱参数,并基于模糊熵理论提取了各参数的权重。再利用模糊熵进行有效的度量融合,最后通过改进后综合判决对情感语句做出识别判定。研究发现融合后的参数增强了情感识别的效果。
Emotion recognition of speech is signification for artificial intelligence research; the feature parameters distillate accuracy influences recognition-rate directly. The improved method based on fuzzy entropy, which analyzed and distilled the weight of speech duration, average magnitudes, pitch frequency, formants and Mel-frequency cesptral coefficients. It＇s effective to fuse the feature parameter by the use of fuzzy entropy. At last, integrated with the advanced fuzzy decision, the proposed method shows better performance than conventional methods.
出处
《微电子学与计算机》
CSCD
北大核心
2006年第12期168-171,174,共5页
Microelectronics & Computer
关键词
语音情感识别
模糊熵
Mel频率倒谱
参数融合
Emotion recognition of speech, Fuzzy entropy, Mel-frequency cesptral coefficients, Parameter fusion
