Stress Detection in Speech Signal Using Machine Learning and AI

Abstract

Individual person’s speech is verbal way to have conversation with others. Speech many time probably becomes to know that individual person is in stressful condition or normal. These can lead with appropriate assessment of the speech signals into different stress types to evoke that the individual person is in a fit state of mind. In this work, stress identification and classification algorithms are developed with the aid of machine learning (ML) and artificial intelligence (AI) together with MFCC feature extraction methods. The machine learning and AI-based approaches use an intelligent combination of feature selection and neural optimization algorithms to train and to improve the classification and identification accurateness of the system. Comparison is done with approach of classical neural networks and fuzzy inference classifiers. The proposed system is suitable for real-time speech and is language and word independent. The work is implemented using MATLAB 2014 version.