基于MDSF的精神分裂症自动识别算法  

Detection algorithm for schizophrenia based on multi-dimensional spatial characteristics of flatten emotion

作　　者：李智倩 郭源蕊 贺子亮 邓丽华[1] 何凌[1] 李元媛 LI Zhi-qian;GUO Yuan-rui;HE Zi-liang;DENG Li-hua;HE Ling;LI Yuan-yuan(College of Electrical Engineering,Sichuan University,Chengdu 610065,China;Huaxi Mental Health Center,Sichuan University,Chengdu 610041,China)

机构地区：[1]四川大学电气工程学院,四川成都610065 [2]四川大学华西心理卫生中心,四川成都610041

出　　处：《计算机工程与设计》2021年第7期1882-1889,共8页Computer Engineering and Design

基　　金：成都市科技惠民技术研发基金项目(2015-HM01-00430-SF);国家自然科学基金青年基金项目(61503264);四川大学创新火花库基金项目(2082604401189);四川省科技厅基金项目(2019YFS0236)。

摘　　要：为解决精神分裂症检测上临床医师短缺、效率低等问题,提出多维度情感扁平化空间域特征(multi-dimensional spatial characteristics of flatten emotion,MDSF)算法,提取精神分裂症和正常人在情绪表达强度的差异信息。实验提取28个患者及28个对照组语音样本的MDSF,结合SVM实现精神分裂症的自动检测,分析MDSF特征在不同维度下的性能。实验结果表明,应用MDSF三维特征,其精神分裂症自动识别正确率为85.1%-89.1%,优于低维度MDSF的识别性能,且MDSF的识别正确率高于国内外现有技术的正确率。提出的MDSF能为临床医师提供客观有效的辅助诊断。At present,there are problems such as shortage of clinicians,low efficiency in the detection of schizophrenia.To solve these problems,a multi-dimensional spatial characteristics of flat emotion(MDSF)algorithm was proposed.Different information of emotional expression intensity between schizophrenia and normal people was extracted.The MDSF of 28 schizophrenics and 28 normal controls were extracted,and the schizophrenics were automatically detected through SVM.The performance of MDSF features in different dimensions was analyzed.Experimental results show that the recognition accuracies of three-dimensional MDSF range from 85.1%to 89.1%,which are better than that of low-dimensional MDSF.And the recognition accuracies of MDSF are higher than that of existing technologies at home and abroad.The MDSF algorithm can provide objective and effective auxiliary diagnosis for clinicians.

关 键 词：精神分裂症 阴性症状 情感淡漠 多维度情感扁平化空间域特征 支持向量机 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]