基于模糊核聚类的多模式情感识别算法研究

Research on multimodal emotion recognition algorithm based on fuzzy kernel clustering

为了克服单模式情感识别存在的局限性,该文以语音信号和面部表情信号为研究对象,提出了一种新型的多模式情感识别算法,实现对喜悦、愤怒、惊奇和悲伤4种人类基本情感的识别。首先,将获取的信号进行预处理并提取情感特征参数,然后利用模糊核聚类算法对其进行聚类分析,即利用Mercer核,将原始空间通过非线性映射到高维特征空间,在高维特征空间中对多模式情感特征进行模糊核聚类分析。由于经过了核函数的映射,使原来没有显现的特征突现出来。实验结果验证了该方法的可行性和有效性。

In order to overcome the limitation of single mode emotion recognition. This paper described a novel multimodal emotion recognition algorithm, took speech signal and facial expression signal as the research subjects, and accomplished recognition for six kinds of human emotion （joy, anger, surprise, sadness）. First, made some pre-processing and extracted emotion feature for speech signal and facial expression signal. Second, used the fuzzy kernel clustering for clustering analysis. That is to say, by using Mercer kernel function, the data in original space were mapped to a high-dimensional eigen-space, and then used the fuzzy clustering for the speech features in the high-dimensional eigen-space. Because of the kernel mapping, the feature inherent in the emotion signals explores, which improves the discriminations of the different emotion category. Experimental results verify the feasibility and effectiveness of the proposed method.