通信学报 ›› 2017, Vol. 38 ›› Issue (12): 109-120.doi: 10.11959/j.issn.1000-436x.2017294

• 学术论文 • 上一篇    下一篇

基于SAE和LSTM RNN的多模态生理信号融合和情感识别研究

李幼军1,2,3,黄佳进1,2,3,王海渊1,2,3,钟宁1,2,3,4   

  1. 1 北京工业大学国际WIC研究院,北京 100124
    2 磁共振成像脑信息学北京市重点实验室,北京 100124
    3 脑信息智慧服务北京市国际科技合作基地,北京 100124
    4 北京未来网络科技高精尖创新中心,北京 100124
  • 修回日期:2017-11-23 出版日期:2017-12-01 发布日期:2018-01-19
  • 作者简介:李幼军(1978-),男,河南栾川人,北京工业大学博士生,主要研究方向为生物信号分析、机器学习及情感计算等。|黄佳进(1977-),男,贵州遵义人,博士,北京工业大学助理研究员,主要研究方向为人工智能、推荐系统等。|王海渊(1981-),男,山西朔州人,博士,北京工业大学工程师,主要研究方向智能传感器、人工智能、智慧医疗系统的开发等。|钟宁(1956-),男,北京人,北京工业大学教授、博士生导师,主要研究方向为人工智能、Web智能、脑信息学、知识发现与数据挖掘、粒计算等。
  • 基金资助:
    国家自然科学基金资助项目(61420106005));国家重点基础研究发展计划基金资助项目(2014CB744600);国家国际科技合作专项基金资助项目(2013DFA32180)

Study of emotion recognition based on fusion multi-modal bio-signal with SAE and LSTM recurrent neural network

You-jun LI1,2,3,Jia-jin HUANG1,2,3,Hai-yuan WANG1,2,3,Ning ZHONG1,2,3,4   

  1. 1 Institute of International WIC,Beijing University of Technology,Beijing 100124,China
    2 Beijing Key Laboratory of Magnetic Resonance Imaging and Brain Informatics,Beijing 100124,China
    3 Beijing International Collaboration Base on Brain Informatics Wisdom and Services,Beijing 100124,China
    4 Beijing Advanced Innovation Center for Future Internet Technology,Beijing 100124,China
  • Revised:2017-11-23 Online:2017-12-01 Published:2018-01-19
  • Supported by:
    The National Natural Science Foundation of China(61420106005));The National Basic Research Program of China(2014CB744600);The International Science&Technology Cooperation Program of China(2013DFA32180)

摘要:

为了提高情感识别的分类准确率,提出一种将栈式自编码神经网络(SAE)和长短周期记忆单元循环神经网络(LSTM RNN)融合的多模态融合特征情感识别方法。该方法通过SAE对不同模态的生理特征进行信息融合和压缩,随后用LSTM RNN对长时间周期的融合进行情感分类识别。通过将该方法用到开源数据集中进行验证,得到情感分类准确率达到0.792 6。实验结果表明,SAE对多模态生理特征进行了有效融合,LSTM RNN能够有效地对长时间周期中的关键特征进行识别。

关键词: 多模态生理信号情感识别, 栈式自编码神经网络, 长短周期记忆循环神经网络, 多模态生理信号融合

Abstract:

In order to achieve more accurate emotion recognition accuracy from multi-modal bio-signal features,a novel method to extract and fuse the signal with the stacked auto-encoder and LSTM recurrent neural networks was proposed.The stacked auto-encoder neural network was used to compress and fuse the features.The deep LSTM recurrent neural network was employed to classify the emotion states.The results present that the fused multi-modal features provide more useful information than single-modal features.The deep LSTM recurrent neural network achieves more accurate emotion classification results than other method.The highest accuracy rate is 0.792 6

Key words: multi-modal bio-signal emotion recognition, stacked auto-encoder neural network, LSTM recurrent neural network, multi-modal bio-signals fusion

中图分类号: 

No Suggested Reading articles found!