基于肢体动作序列三维纹理特征的情绪识别

Emotion recognition based on three-dimensional texture feature of body action sequence

目前基于人脸表情的情绪识别已经相对成熟,而根据人类肢体动作进行情绪识别的研究却不多。通过VLBP和LBP-TOP算子从三维空间中提取图像序列的肢体动作特征,分析愤怒、无聊、厌恶、恐惧、高兴、疑惑和悲伤七种自然情绪的特点,并用参数优化的支持向量机对情绪分类进行识别,识别率最高能够达到77. 0%。实验结果表明,VLBP和LBP-TOP算子具有较强的鲁棒性,能有效地从肢体动作中识别人的情绪。

At present,emotional recognition based on facial expression has been relatively mature,however,the research on human emotion recognition based on body movements is less.This paper utilized VLBP and LBP-TOP operator to extract the feature of the body movement from the three dimensional space.It analyzed the characteristics of 7 kinds of natural emotions,such as anger,boredom,disgust,fear,happiness,uncertainty and sadness,and used the SVM algorithm to classify emotion,the highest recognition rate could reach 77.0%.The experimental results show the strong robustness of VLBP and LBP-TOP operators and the algorithm can effectively identify human emotions from limb movements.