基于语音音调域动态特征的精神分裂症自动检测算法  

AUTOMATIC DETECTION FOR SCHIZOPHRENIA BASED ON DYNAMIC CHARACTERISTICS OF VOICE TONE DOMAIN

作　　者：孟雨璇 袁尉 李雯 邓丽华[1] 何凌[1] 李元媛 Meng Yuxuan;Yuan Wei;Li Wen;Deng Lihua;He Ling;Li Yuanyuan(College of Electrical Engineering,Sichuan University,Chengdu 610065,Sichuan,China;Huaxi Mental Health Center,Sichuan University,Chengdu 610065,Sichuan,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610065

出　　处：《计算机应用与软件》2022年第8期171-180,225,共11页Computer Applications and Software

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

摘　　要：为了提供精神分裂症阴性症状临床诊断的客观生物学指标,实现精神分裂症自动检测,提出一种基于语音音调域动态特征的精神分裂症自动检测算法。基于频率覆盖度量和子带编码模型对语音信号进行基频提取,根据基频曲线及音调特征提出FDS-set(Fundamental frequency Digital Statistical feature Set)、EVR(Eight-tenths of the Vocal Range)、PVPD(Pitch Peak-Valley Profile Distance)、TFF(Tone Fluctuation Frequency)音调域声学动态特征参数,将其音调域动态声学特征参数集与SVM分类器结合,实现精神分裂症的自动检测。实验数据采集于28位精神分裂症患者及28位正常人,共448句语音样本。实验结果表明,提出的基于语音音调域动态特征的精神分裂症自动检测算法在精神分裂症自动检测的正确率达88.39%。相比国内外现状方法,该算法能够提取精神分裂症患者独特的语音音调特征,为临床医生提供客观的辅助诊断方法,具有重要的临床研究意义。In order to provide objective biological indicators for clinical diagnosis of negative symptoms of schizophrenia and realize automatic detection of schizophrenia,this paper proposes an automatic detection algorithm for schizophrenia based on dynamic characteristics of voice tone domain.We extracted the fundamental frequency of speech signals based on the frequency coverage metric and subband encoding model.According to the fundamental frequency curve and pitch feature,we proposed FDS-set,EVR,PVPD,TFF dynamic characteristic parameters in tone domain.The pitch domain dynamic acoustic feature parameter set was combined with SVM classifier to realize the automatic detection of schizophrenia.The experimental data were collected from 28 schizophrenic patients and 28 healthy people.A total of 448 segments of speech samples were obtained.The experimental results show that the proposed method achieves the accuracy of 88.39%for the automatic detection of schizophrenia.Compared with the current methods at home and abroad,the proposed method can extract the unique phonetic tone features of schizophrenia patients,provide objective aided diagnosis for clinicians,which has important clinical research significance.

关 键 词：精神分裂症 音调域动态特征 自动识别 基音频率 语音音调 

分 类 号：TN912.3[电子电信—通信与信息系统] R749.3[电子电信—信息与通信工程]