Study On Speech Signal Time Characteristics Of Depressed People

Depression is a mental disorder characterized by loss of interest and persistent feeling of sadness.Effective treatment is urgently needed for it high morbidity and mortality.However,scarce medical resources and social misconceptions about mental disorder lead to low treatment rates of depression,many patients with depression have not received timely professional guidance and treatment.At present,the main assessment methods of depression rely almost exclusively on various scales provided by patients and/or clinicians,this way need patient describe the illness accurately,has strong desire to recovery and doctors have rich clinical experience.As a result,this way has high misdiagnosis rate.To aid clinician master the patient’s condition more accurately,researchers have been looking for an objectively valid,non-invasive,and psychologically related measures for many years.Speech is an attractive candidate for its advantages: non-invasive,fast,convenient and practical(the ubiquity of mobile computing make it possible to monitor and collect patient’s data anytime,anywhere).Through clinical observation,clinician find that the language behavior of depression patients have so many characteristics,such as slow,pause.The slowing of speech speed can be described by the time characteristics of speech signals.This paper focused on time characteristics in speech signals,the main contributions are as follows:A data acquisition experiment with speaking style and emotional stimulation was designed,and 673 subjects data were collected over the past three years.It provides a big database for the analysis of speech signals of depressed patient.Nine time characteristics are constructed to measure the voice duration,pause duration and other information.In our paper,depressed patients has longer speech pause time than healthy controls,and the different is significant,medicine intervention can reduce speech pause time on depressed patients.So speech pause time is considered as a promising biomarker for depression detection.We built a model of depression recognition,and compared the time characteristics subset and the subset of audio part of the speech.The effectiveness of time characteristics in depression identification was verified.In the first and second phase male data,compared with phonetic features model,the accuracy of the time characteristics model increased by 6.1% and 9.2% respectively.In this paper,we construct time characteristics to describe the pause time,phonation time,and other information.the experimental results show that there is difference between depression group and healthy controls on the time characteristics of speech.In future research,to improve the recognition rate of patients with depression,we should consider making full use of the information carried by silent parts of speech.