Audio Signal Based Stress Recognition System using AI and Machine Learning

Abstract

The detection of stress from speech signals has recently received a lot of interest. Individual speech is a verbal means for people to communicate with one another. The human speech reflects the speaker’s mental state. To ensure that the person is in a healthy state of mind, proper classification of these speech signals into stress categories is essential. Speech is frequently used to detect whether a person is in a stressful or routine scenario. These can lead to the reliable classification of speech signals into separate stress types, showing that the individual is in a healthy state of mind. In this paper, stress identification and classification algorithms are built using machine learning (ML), artificial intelligence (AI), and Mel frequency cepstral coefficient (MFCC) feature extraction methodologies. Because most current stress indicators are intrusive, requiring samples from patients’ bodies, this study was done to identify methods to detect stress without introducing instruments into the body. This study illustrates how stress can be detected by using speech signal analysis approaches.