摘要
目的探索声学特征与青少年心境障碍抑郁发作患者自杀自伤行为的关系。方法招募193例青少年心境障碍抑郁发作患者,其中127例伴自杀自伤行为、66例不伴自杀自伤行为,采集患者朗读时的语音音频,提取包括基频包络、梅尔频谱、质谱心、频谱滚降系数及梅尔倒谱系数(Mel frequency cepstral coefficient,MFCC)在内的6552个声学特征,比较两组患者声学特征的差异,并对差异特征进行主成分分析,利用多因素logistic回归分析声学主成分与自杀自伤行为的关联。结果两组比较得到96个声学特征存在统计学差异(P<0.01),从中识别出12个声学主成分,logistic回归分析显示主成分1(OR=0.507,P=0.001)、主成分2(OR=0.535,P=0.008)、主成分5(OR=0.541,P=0.003)与自杀自伤行为负关联,这些主成分包括的特征为梅尔频谱、基频包络、高频谱滚降系数、MFCC8和MFCC1,主成分3(OR=1.505,P=0.042)、主成分12(OR=1.552,P=0.023)与自杀自伤行为正关联,这些主成分包括的特征为质谱心、低频谱滚降系数、MFCC0、梅尔频谱22、梅尔频谱24的最小值所在位置。结论青少年心境障碍抑郁发作患者自杀自伤行为与能量和频率等声学特征相关。
Objective To explore the relationship between acoustic features and suicidal and self-harming behaviors in adolescents with mood disorders(MDs)during depressive episode.Methods A total of 193 adolescent MDs patients with depressive episodes(127 cases with suicidal and self-harming behaviors and 66 cases without suicidal and self-harming behaviors)were recruited for this study.Acoustic data,including 6552 acoustic features such as fundamental frequency envelope,Mel spectrum,spectrum centroid,spectrum roll off coefficient,and Mel frequency cepstral coefficient(MFCC),were extracted from the patients'speech recordings.Differences in acoustic features between the two groups were compared,and principal component analysis was performed on the different features.Multifactor logistic regression was used to analyze the association between acoustic principal components and suicidal and self-harming behaviors.Results Ninety-six acoustic features were found to be significantly different between the two groups(P<0.01),from which 12 acoustic principal components were identified.Logistic regression analysis showed that principal component 1(PC1)(OR=0.507,P=0.001),PC2(OR=0.535,P=0.008),and PC5(OR=0.541,P=0.003)were negatively associated with suicidal and self-harming behaviors.These principal components included features such as fundamental frequency envelope,high frequency roll off coefficient,Mel spectrum,MFCC8,and MFCC1.On the other hand,PC3(OR=1.505,P=0.042)and PC12(OR=1.552,P=0.023)were positively associated with suicidal and self-harming behaviors,and included features such as spectrum centroid,low frequency roll off coefficient,the position of the minimum value of Mel spectrum 22 and 24,and MFCC0.Conclusion Suicidal and self-harming behaviors in adolescents with mood disorders and depressive episodes are associated with acoustic features such as energy and frequency.
作者
任永霞
方翰铮
张然
朱荣鑫
张锡哲
王菲
REN Yongxia;FANG Hanzheng;ZHANG Ran;ZHU Rongxin;ZHANG Xizhe;WANG Fei(Early Intervention Unit,Affiliated Nanjing Brain Hospital,Nanjing Medical University,Nanjing 210029,China)
出处
《中国神经精神疾病杂志》
CAS
CSCD
北大核心
2023年第3期141-145,共5页
Chinese Journal of Nervous and Mental Diseases
基金
江苏省科技计划项目(编号:BE2021617)
国家自然科学基金面上项目(编号:62176129)。
关键词
青少年
心境障碍
抑郁发作
自杀
自伤
声学特征
主成分分析
logistic回归分析
Adolescent
Mood disorders
Depressive episode
Suicide
Self-harming
Acoustic features
Principal component analysis
Logistic regression analysis
