A Deep Learning Approach to Speech Recognition for Detection of Mental Disorders

Abstract

Mental disorders are conditions that affect a person’s cognitive functions, behavior or thinking, thereby impairing daily functions. The dearth of trained psychologists to the high number of patients living with a mental disorder pose significant challenges in the field of mental health. This study investigates the application of deep learning techniques to speech recognition for the purpose of detecting mental disorders. The main objective of this study is to effectively identify speech patterns associated with various mental disorders and thereafter develop a robust and accurate deep learning model system that can detect risk of a mental disorder in an individual based on their voice. The research methodology involved the collection of a dataset consisting of speech recordings from individuals diagnosed with depression and post-traumatic stress disorder (PTSD). The dataset acquired was carefully curated to include symptom severity levels, and linguistic variations. The results of this study demonstrate the effectiveness of deep learning approaches in speech recognition for mental disorder detection. The trained models achieved 95% and 94% accuracy rates in identifying and differentiating speech patterns associated with depression and PTSD respectively. The findings of this study have significant implications for the field of mental health. The developed deep learning system offers a promising avenue for the early detection and monitoring of mental disorders. Further research is warranted to validate and refine the developed models using larger and more diverse dataset. Additionally, the integration of multimodal data, such as combining speech analysis with psychological or text-based data, could enhance the diagnostic accuracy and reliability of the system. Keywords: Deep Learning, Mental Disorder, Post-traumatic Stress Disorder, Speech Recognition, Depression.