项目名称: 基于声学与韵律特征的社交人格感知
项目编号: No.61473041
项目类型: 面上项目
立项/批准年度: 2015
项目学科: 其他
项目作者: 谢湘
作者单位: 北京理工大学
项目金额: 80万元
中文摘要: 社交人格感知是目前人机交互领域中急需解决的问题,其应用富有实用价值。本课题提出适用于社交人格感知分析的数据获取方法,并构建基于中文电话录音的社交人格语音数据库,为深入研究和理解自动人格感知机理提供关键的实验数据;全面深入地研究声学特征与韵律特征的挑选与组合方法,比较不同特征在人格感知中所起作用的大小,并应用因子分析算法,寻求高效的声学与韵律特征获取方法;深入研究基于神经网络的社交人格预测新方法,使用机器学习算法构建基于语音声学和韵律特征的社交人格统计模型和分类器。最终建立一套能够快速准确地从电话录音中预测社交人格的原型系统。通过本项目的研究,一方面能够建立更有效的声学和韵律特征提取与挑选算法,得到更有效的人格预测模型,进一步提高人格感知的性能;另一方面基于声学与韵律特征的社交人格感知有相当广阔的应用前景。
中文关键词: 社交人格感知;语音分析;声学特征;韵律特征建模
英文摘要: Social personality attribution is an essential and practical issue in the field of human machine interaction. In this proposal, data and feature acquisition methods for social personality attribution are proposed, and are used for constructing a telephone recording conversational database of Mandarin. The database is used as a key experimental material in the research and understanding of automatic personality perception. Acoustic and prosodic features are selected and composed to be used in personality attribution. Factor analysis algorithms are applied and then machine learning algorithms are implemented for constructing statistical model of non-verbal acoustic and prosodic feature based social personality classifiers. A prototype system of instant social personality prediction from telephone conversations is constructed. By doing this research, more efficient acoustic and prosodic feature selection methods are proposed for personality prediction. On the other hand, the social personality perception has quite broad application prospects.
英文关键词: Social Personality Perception;Speech Analysis;Acoustic Feature;Prosodic Feature Modeling