跨文化语音情感识别综述-《计算机技术与发展》

文章信息/Info

Author(s):: LU Chen-ye; WANG Wei; School of Education Science,Nanjing Normal University,Nanjing 210097,China

Keywords:: emotion perception; emotion expression; automatic speech emotion recognition; cross culture; emotional characteristics

摘要:: 每种文化都有自己的特性,它会影响情绪的方方面面。跨文化语音情感识别是以跨文化为前提对语音所传递情绪的识别,已引起了心理科学与计算机研究等多学科广泛的关注,对今后多文化交流与人机交互具有极高的应用价值。文章回顾了跨文化语音情感识别的相关文献,从情绪感知的文化效应和情绪表达的文化效应等角度进行归纳总结,并分析了在人机交互过程中如何降低文化对情绪表达的影响,以此来提高跨文化自动语音情感系统的识别性能。结果表明,在情绪感知中,文化会造成群体内优势;在情绪表达中,文化会造成语音声学特征的系统性差异。因此,在自动语音情感识别中可以通过对特征进行选择、减小文化带来的特征分布差异以及分离特征中的文化因素来降低文化对情绪表达的影响。

Abstract:: Every culture has its own characteristics,which affect all aspects of emotion. Cross-cultural speech emotion recognition is therecognition of emotions conveyed by speech on the premise of cross-culturalism. It has attracted wide attention from multiple disciplinessuch as psychological science and computer research. It has extremely high application value for multicultural communication and human-computer interaction in the future. The article reviews the relevant literature of cross-cultural speech emotion recognition,summarizes thecultural effects of emotional perception and emotional expression, and analyzes how to reduce the influence of culture on emotionalexpression in the process of human-computer interaction,improving the recognition performance of the cross-cultural automatic speech emotion system. The results shows that in emotional perception, culture can cause intra - group advantages. In emotional expression,culture can cause systemic differences in voice acoustic characteristics. Therefore,in automatic speech emotion recognition,the influenceof culture on emotional expression can be reduced by combining features, eliminating the difference in feature distribution caused byculture,and separating cultural factors in features.