基于多模态信息融合的语音意图理解方法
首发时间:2011-08-02
摘要:为从语音中获取包括字面含义和说话人情绪状态在内的全面意图信息,提出了一种基于多模态信息融合的语音意图理解方法,并对其中的关键词抽取、命令解析、基于文本/韵律特征的情绪状态检测以及多模态信息融合等关键算法进行了设计。该方法从识别文本和语音信号中抽取不同模态的信息并进行融合,能够有效地从语音中获取丰富的意图信息,有助于建立自然的人机交互环境。
For information in English, please click here
A Speech Intention Understanding Method Based on Multimodal Information Integration
Abstract:In order to obtain comprehensive speech intention information containing both the literal meaning and speaker’s affective state, a speech understanding method based on multimodal information integration is proposed. Key algorithms including keywords extraction, command analyzing, text/prosody-based affective state determination and multimodal information integration are designed. The method is able to effectively obtain rich intention information by extracting information of different modality from recognition text and speech signal and merging them together, which is helpful to establish a natural human-computer interaction environment.
Keywords: speech intention understanding multimodal information extraction multimodal information integration
论文图表:
引用
No.****
同行评议
勘误表
基于多模态信息融合的语音意图理解方法
评论
全部评论0/1000