面向口语对话系统的用户情感识别研究

With the rapid development of the information age, the natural harmony of Human-machine Dialogue has become the main trend and the urgent need for Human-computer Interaction. Traditional spoken dialogue systems mainly focus on specific task-oriented dialogue, rarely involve users' emotional state analysis. Consequently, they are unable to catch frustrated or depressed feelings of users, and cannot response properly according to users' emotional states. In this proposal, we focus on the recognition of user emotional states in spoken dialogue systems, which is the fundermental step for intelligent emotion dialogue. This proposal specifically includes: User emotion analysis and representation in dialogue. By coding of multi-dimensional emotions, the links between the basic emotions are established; Dialogue emotion feature analysis. Building a dialogue emotion database which contains multi-source emotion features and users' real feelings; User emotion object feature confusion and recognition modeling, and feature selection and dimensionality reduction for multi-source features; Proposing modeling user emotion sequence in dialogue. A strategy of using different user emotion factors is proposed to achieve a personalized users emotion recognition, and to reflect users' emotional overlay and decay. Research on these issues will provide new ideas and solutions for user emotion recognition in spoken dialogue systems, and lay the theoretical and technical basis to promote Human-computer Interaction.

自然和谐的人机对话是高速信息时代人机交互的主要发展趋势和迫切需求。传统的口语对话系统主要侧重于面向特定任务的对话，很少涉及对用户的情感状态进行分析，这使得系统无法实时掌握用户遇到挫折或感到沮丧的状态，也无法根据用户的情感状态进行适当的应对。本课题致力于研究面向口语对话系统的用户情感识别问题，它是实现智能情感会话的基础。具体内容包括：会话用户情感的分析与表示，提出通过对用户的多维情感进行编码建立基本情感之间的联系；会话情感特征分析，构建具有多源特征，并包含用户真实情感的会话情感库；用户情感对象特征融合与识别建模，设计针对多源特征的特征选择和特征降维方法；提出基于多源时序特征的用户情感序列识别模型，采用不同的用户情感影响因子，以实现个性化的用户情感识别，并能反映会话过程中用户情感叠加与衰减。对这些问题的研究将为口语对话系统中用户情感识别提供新的思路和解决方法，为促进人机交互打下理论和技术基础。

自然和谐的人机对话是高速信息时代人机交互的主要发展趋势和迫切需求。传统的口语对话系统主要侧重于面向特定任务的对话，很少涉及对用户的情感状态进行分析，这使得系统无法实时掌握用户遇到挫折或感到沮丧的状态，也无法根据用户的情感状态进行适当的应对。本课题致力于研究面向口语对话系统的用户情感识别问题，它是实现智能情感会话的基础。具体内容包括：.1) 会话用户情感的分析与表示，会话情感特征分析，设计情感会话语料库构建框架;.2) 从大规模数据中有效地获取来自不同信息源的情感特征；.3) 多模态数据融合算法，针对语音特征和语义特征的特征选择和特征降维方法，以及对各分类器置信度的评价方法..4) 设计基于多源时序特征的用户情感序列识别模型。.对这些问题的研究为口语对话系统中的用户情感识别的研究提供新的思路和解决方法，为促进人机交互打下理论和技术基础。