Analysis of Mental Stress with Machine Learning Methods

This paper uses speech analysis and machine learning to detect a student’s stress level through speech signals. Stress can occur for various reasons. The speech or utterance of any individual carries a significant amount of information regarding that individual’s mental state. The speech signal helps to detect if a person is stressed or not. The level of stress can be determined by the person’s style or tone of voice when answering questions or speaking. The acoustic properties of speech could be considered as an appropriate means for the detection of stress levels of that individual. This paper focuses on identifying an individual’s stress level by considering speech acoustics through machine learning. This model can determine a person’s stress level according to the speech features, leveraging and machine learning. For this purpose, a system is constructed separately using the linear support vector machine (SVM) and K-nearest neighbour (KNN), considering the MFCC features. Among the machine learning classifiers, the highest accuracy that had been recorded was by K-nearest neighbour with an accuracy of 93.5%.