Role of Spontaneous and Read Speech in Depression Detection for Adolescents
Abstract
Depression is more than just a low mood – it is a serious condition that negatively affects an individual’s physical and mental health. Depression is classified as a mood disorder which may be described as feelings of sadness, loss, or anger that interferes with a person’s day to day activities. With our work, we are trying to develop a system which will assist clinicians to detect depression without subjective assessment. In this paper, we hypothesize that a) speech features can be used as a biomarker to detect depression b) better results will be obtained by using spontaneous speech rather than read speech c) certain speech features will give better classification results and their performance is unaffected when used for individuals having different age group, sex and culture d) a small portion of speech will give the same classification result as using the whole part of the speech. By experimenting and evaluating classification results for the dataset of 54 depressed and 75 healthy individuals using different speech features, and using SVM as classification algorithm, we found that speech features can be used as a reliable biomarker for depression detection. Speech features like MFCC, pitch, jitter, shimmer and energy perform better in classifying an individual as a depressed or a controlled one. In depression detection, spontaneous speech performs remarkably better than read speech. We also found that in the classification, performance of a small portion of speech is similar to that of the complete portion of speech.