计算机工程与应用
計算機工程與應用
계산궤공정여응용
COMPUTER ENGINEERING AND APPLICATIONS
2014年
15期
187-190,239
,共5页
语音情感%智能模型%神经网络%主成分分析
語音情感%智能模型%神經網絡%主成分分析
어음정감%지능모형%신경망락%주성분분석
speech emotion%intelligent model%neural network%Principal Component Analysis(PCA)
为克服由传统语音情感识别模型的缺陷导致的识别正确率不高的问题，将过程神经元网络引入到语音情感识别中来。通过提取基频、振幅、音质特征参数作为语音情感特征参数，利用小波分析去噪，主成分分析（PCA）消除冗余，用过程神经元网络对生气、高兴、悲伤和惊奇四种情感进行识别。实验结果表明，与传统的识别模型相比，使用过程神经元网络具有较好的识别效果。
為剋服由傳統語音情感識彆模型的缺陷導緻的識彆正確率不高的問題，將過程神經元網絡引入到語音情感識彆中來。通過提取基頻、振幅、音質特徵參數作為語音情感特徵參數，利用小波分析去譟，主成分分析（PCA）消除冗餘，用過程神經元網絡對生氣、高興、悲傷和驚奇四種情感進行識彆。實驗結果錶明，與傳統的識彆模型相比，使用過程神經元網絡具有較好的識彆效果。
위극복유전통어음정감식별모형적결함도치적식별정학솔불고적문제，장과정신경원망락인입도어음정감식별중래。통과제취기빈、진폭、음질특정삼수작위어음정감특정삼수，이용소파분석거조，주성분분석（PCA）소제용여，용과정신경원망락대생기、고흥、비상화량기사충정감진행식별。실험결과표명，여전통적식별모형상비，사용과정신경원망락구유교호적식별효과。
To improve the problem of the low recognition accuracy caused by the defect of the traditional speech emotion recognition model, this algorithm of process neural networks is introduced to the speech emotion recognition. This paper extracts the speech emotion features of fundamental frequency, amplitude, sound characteristic, and uses the method of wavelet analysis to reduce noise, the Principal Component Analysis(PCA)to reduce the redundancy, and carries on the experiment of classification and recognition of the four speech emotions of anger, happiness, sadness and surprise. The result proves that the method of process neural network has better recognition effect on the four speech emotions compared with the traditional recognition model.
