Improved emotional speech recognition algorithms

Abstract

Depression and suicidality are the major public mental health concerns. Over 90% of people who die by suicide have depression or any other diagnosable mental disorder. However diagnosis of such conditions clinically is not always possible as they are unpredictable. Emotional speech recognition tools can be used efficiently to diagnose risks of depression and suicide. In this paper, a model of emotional speech recognition algorithm using Tamil language with the help of Linear Predictive Coding (LPC) and Parameters based method to diagnose the depression and suicide attempts is proposed. The system is trained using Artificial Neural Network (ANN) with multilayer perceptron back propagation method. The database of emotional speech is collected and the results are obtained for LPC based and parameters such as energy, average value based recognition system. LPC determination is easy and accurate and the best recognition rate achieved using LPC based recognition is 90%.