相空间重构的情感语音特征提取及优化

发布时间：2018-04-22 00:40

  本文选题：相空间重构 + 非线性几何特征　； 参考：《西安电子科技大学学报》2017年06期

【摘要】：针对现有语音情感特征在表征情感信息上的不完整,将相空间重构理论引入到情感语音的特征提取中.通过分析不同语音情感状态下相空间重构的几何特性,提取了该重构相空间下基于轨迹的描述轮廓的5种非线性几何特征作为新的情感语音特征参数,并根据情感与特征映射的关系提出一种特征参数优化方法.首先,选用德语柏林语音库中的高兴、悲伤、中性和生气4种情感作为实验样本;其次,提取非线性几何特征和非线性属性特征(最小延迟时间、关联维数、Kolmogorov熵、最大Lyapunov指数和Hurst指数);最后,根据设计方案采用支持向量机进行情感语音识别.实验结果表明,该特征相较于非线性属性特征在情感语音识别上有较强的优势度,联合非线性属性特征后,通过特征参数优化的方法获得了最优的非线性特征集合,验证了该方法的实用性.
[Abstract]:Aiming at the incomplete representation of emotional information in existing speech emotional features, the theory of phase space reconstruction is introduced into the feature extraction of emotional speech. By analyzing the geometric characteristics of phase space reconstruction in different speech emotion states, five kinds of nonlinear geometric features based on trajectory describing contour in the reconstructed phase space are extracted as new emotional speech feature parameters. According to the relationship between emotion and feature mapping, a feature parameter optimization method is proposed. Firstly, the happy, sad, neutral and angry emotions in the German Berlin language corpus are selected as experimental samples; secondly, the nonlinear geometric features and nonlinear attribute features (minimum delay time, correlation dimension and Kolmogorov entropy) are extracted. The maximum Lyapunov exponent and Hurst exponent are obtained. Finally, support vector machine is used for emotional speech recognition according to the design scheme. The experimental results show that the feature has a strong superiority in emotional speech recognition compared with the nonlinear attribute feature. After combining the nonlinear attribute feature, the optimal nonlinear feature set is obtained by the method of feature parameter optimization. The practicability of the method is verified.
【作者单位】： 太原理工大学信息工程学院;
【基金】：国家自然科学基金资助项目(61371193)
【分类号】：TN912.3
，

本文编号：1784869