ANALYSIS OF NATURAL SPEECH FOR THE ASSESSMENT OF MOOD DISORDERS

Abstract

Psychiatric evaluation relies on subjective assessment. Biomarkers are objective indicators of illness that can aid in diagnostic classification and help guide treatment for individuals with psychiatric illnesses. Speech has been identified as an informative biomarker that is objective and easy to collect. Speech analysis has been shown to be effective in diagnostic classification, assessment of severity and prognosis, and early onset prediction of psychiatric illness. We aimed to synthesize results of published work and validate speech analysis methods for clinical application. We completed a systematic review to explore the state of the field and identify areas for further investigation. We present the collection and analysis methods for a speech study aimed at creating a corpus of high-quality speech data. We analyzed a preliminary sample using content speech features to differentiate bipolar from unipolar depression. Results from these preliminary analyses demonstrate the efficacy of our data collection procedure and the utility of content variables for tackling important classification problems in psychiatry.