摘要
维度语音情感识别(Dim-SER)是情感计算领域的一个新兴分支,它从多维、连续的角度看待情感,将SER问题建模为连续值的预测回归任务。当前的Dim-SER系统在进行情感预测时缺少对语料间情感程度相对顺序的考虑,严重影响了人机交互系统对说话人情感变化趋势的把握。从该需求出发,本文以人类情感认知特性为参照,构建了一个对情感程度相对顺序敏感的Dim-SER系统,并引入Gamma统计对SER系统性能评价标准加以完善。系统构建过程中,本文构造了Top-rank概率分布对语料间的情感顺序进行描述,并使用Kullback-Leibler距离对预测造成的顺序一致性损失进行度量,最后提出顺序敏感的神经网络算法实现系统预测损失的最小化。情感预测实验结果表明,同常用的k近邻算法和支持向量回归算法相比,该系统有效地提高了语料间情感程度相对顺序的正确性。
Dimensional speech emotion recognition（Dim-SER） is a rising branch of emotion computing field.It views emotion from dimensional and continuous perspective,and formalizesthe SER problem as a regression task.Current Dim-SER researches never consider the relative order of emotional degree between utterances,which would makethe human-machine interface get wrong information about speaker＇ s emotion variation trend.Starting from this demand,this paper constructs an order sensitive Dim-SER system with the human emotion cognitive characteristics as reference,and employsGamma statisticto evaluate emotion recognition performance.Specifically, the Top-rank probability distribution is developed to describethe emotional ordering of utterances,and the Kullback-Leibler divergence is usedto measure the loss of order consistency caused by emotion recognition.Finally,the Order-Senstive Network（OSNet） algorithm is proposed to minimized prediction loss.Experimental results show that,compared with the commonly usedA-Nearest Neighbor （k-NN） and Support Vector Regression（SVR） approaches,the proposed system effectively improve thecorrectness of emotional relative order between utterances.
出处
《信号处理》
CSCD
北大核心
2011年第11期1658-1663,共6页
Journal of Signal Processing
基金
自然科学基金(60772076)
语言语音教育部微软重点实验室开放基金资助项目(HIT.KLOF.2009015)
高等学校博士学科点专项科研基金(No.20050213032)
