权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

Development of self-diagnosis system using audio, expression and action for depression prediction

开发利用音频、表情和动作进行抑郁症预测的自我诊断系统

基本信息

批准号：
20J13009
负责人：
LIU JIAQING
金额：
$ 0.58万
依托单位：
Ritsumeikan University
依托单位国家：
日本
项目类别：
Grant-in-Aid for JSPS Fellows
财政年份：
2020
资助国家：
日本
起止时间：
2020-04-24 至 2021-03-31
项目状态：
已结题

来源：
https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-20J13009/
关键词：
うつ病認識マルチモダリティAI

项目摘要

ストレスが多い現代社会では"うつ病”が大きな問題である．うつ病に対する早期発見と適切な治療により，大部分の改善が期待できることから，人工知能を用いた音声や表情などの特徴からうつ状態を検出する手法が期待されている. 本研究では，被験者の音声と表情動画像を用いたうつ状態検出法（Multi-modal Adaptive Fusion Transformer Network）を提案した．主な貢献点は, ①適応的マルチモダリティ特徴融合法を提案し，うつ状態検出に最も有効なモダリティ特徴を強調することによって高い検出精度の実現，②再帰型ニューラルネットワーク(RNN)の代わりにTransformerを用いたLong-range の時系列特徴を抽出による，検出精度の向上，③マルチタスク学習による検出精度の向上，である．提案法はstate-of-the-art精度を達成した．それらの成果は，国際学術誌IEEE Sensors Journal，IEEE Transactions on Instrumentation and Measurement，Sensors，および国際学会ACCVなどで発表した．

Modern society is full of problems. Early detection of the disease, appropriate treatment, most of the improvement is expected, artificial knowledge can be used, voice, expression, characteristics, state, detection methods are expected. In this study, we propose a Multi-modal Adaptive Fusion Transformer Network for the voice and expression animation of the victim. The main contribution points are as follows: (1) the proposed method of feature fusion for adaptive classification;(2) the implementation of high detection accuracy in the detection of the most effective feature;(3) the upward detection accuracy in the extraction of Long-range time series features for Transformer generation;(4) the upward detection accuracy in the detection of the most effective feature in the detection of state detection;(5) the upward detection accuracy in the detection of the most effective feature in the detection of state detection;(6) the upward detection accuracy in the detection of state detection;(7) the upward detection accuracy in the detection of state detection; and (8) the upward detection accuracy in the detection of state detection. Proposal law is state-of-the-art precision achieved. IEEE Sensors Journal, IEEE Transactions on Instrumentation and Measurement, Sensors, and ACCV.