基于随机森林和卷积特征学习的人机交互语音情感识别

针对情绪机器人的自动语音情感识别在不同类型人群之间的语音特征差异,提出了一种用于语音情感识别的随机森林,结合卷积特征学习对情绪化社交机器人系统进行了初步的仿真实验,结果表明情绪机器人能够实时跟踪兴奋、愤怒、哀伤、高兴、惊讶、恐惧、中性7种基本情绪。通过采用非个性化的语音情感特征,补充了原始的个性化语音情感特征,实现了对通用性情感和特殊性情感的提取,对于情感机器人来说,利用这些指标在模拟实验和应用实验中都具有一定的应用前景。

Human-computer Interaction Speech Emotion Recognition Based on Random Forest and Convolution Feature Learning

Focus on the different speech features of different types of people in the automatic speech emotion recognition of emotional robots,a random forest for speech emotion recognition is proposed,and a preliminary simulation experiment of emotional social robot system based on convolution feature learning is carried out.The results show that the emotional robot can track in real time,the seven basic emotions of excitement,anger,sadness,happiness,surprise,fear and neutrality.By using non personalized speech emotion features,the original personalized speech emotion features are supplemented,and the general emotion and special emotion are extracted.For emotional robot,using these indicators has a certain application prospect in the simulation experiment and application experiment.