维度语音情感识别研究综述

The Research Review of Dimensional Speech Emotion Recognition

维度语音情感识别是语音识别技术的重要研究方向,提取最能表达语音情感的特征码并构建具有模型泛化性和鲁棒性的声学模型是语音情感识别的重要研究内容。同时,其触及领域具备较强的多样性,心理学、模式识别以及认知科学等均属于其研究范围,而这些模块是其研究的重点,开展研究的目的主要是为了让机器具备人类情感,促使人机交互更加自然灵活。基于此,该文阐述了在情感心理学的研究基础上,分析情感语音数据库与数据标注,并对情感分类与回归加以探索,希望可以为维度语音情感识别提供新的思路。

Dimensional speech emotion recognition is an important research direction of speech recognition technology,and it is an important research content of speech emotion recognition to extract the feature code that can best express speech emotion and build an acoustic model with model generalization and robustness.At the same time,the fields it touches have a strong diversity,psychology,pattern recognition and cognitive science belong to its research scope,these modules are the focus of its research,and the main purpose of the research is mainly to make machines have human emotions and promote human-computer interaction to be more natural and flexible.Based on this,this paper expounds the analysis of the emotion speech database and data annotation on the basis of the research of emotion psychology,and explores the emotion classification and regression,hoping to provide new ideas for dimensional speech emotion recognition.