基于BOOSTING框架的视觉语音多模态情感识别检测方法

发布时间：2018-04-27 12:42

本文选题：情感识别 + 表情识别　；参考：《现代电子技术》2017年23期

【摘要】：情感识别技术是智能人机交互的重要基础,它涉及计算机科学、语言学、心理学等多个研究领域,是模式识别和图像处理领域的研究热点。鉴于此,基于Boosting框架提出两种有效的视觉语音多模态融合情感识别方法:第一种方法将耦合HMM(coupled HMM)作为音频流和视频流的模型层融合技术,使用改进的期望最大化算法对其进行训练,着重学习难于识别的(即含有更多信息的)样本,并将Ada Boost框架应用于耦合HMM的训练过程,从而得到Ada Boost-CHMM总体分类器;第二种方法构建了多层Boosted HMM(MBHMM)分类器,将脸部表情、肩部运动和语音三种模态的数据流分别应用于分类器的某一层,当前层的总体分类器在训练时会聚焦于前一层总体分类器难于识别的样本,充分利用各模态特征数据间的互补特性。实验结果验证了两种方法的有效性。
[Abstract]:Emotion recognition is an important foundation of intelligent human-computer interaction. It involves many research fields such as computer science, linguistics, psychology and so on. It is a hot topic in the field of pattern recognition and image processing. In view of this, two effective emotion recognition methods for visual speech multimodal fusion based on Boosting framework are proposed. The first method uses coupled HMM(coupled HMMas as model layer fusion technology for audio and video streams. The improved expectation maximization algorithm is used to train it, which is difficult to identify (that is, with more information). The Ada Boost framework is applied to the training process of coupled HMM, and the Ada Boost-CHMM classifier is obtained. The second method is to construct a multilayer Boosted hmm classifier. The data streams of facial expression, shoulder motion and speech are applied to one layer of the classifier, respectively. The current overall classifier will focus on the samples which are difficult to identify in the previous layer, and make full use of the complementary characteristics among the modal feature data. The experimental results show the effectiveness of the two methods.
【作者单位】：成都理工大学信息科学与技术学院;
【基金】：四川省软件工程专业卓越工程师质量工程项目支持(11100-14Z00327)
【分类号】：TN912.34

【相似文献】