Statistical Analysis of Stress Prediction from Speech Signatures

Abstract

AbstractStress is a natural physical and psychological response to life’s events. It can be triggered because of being under lots of pressure, having responsibilities that you are finding overwhelming, not having much or any control over the outcome of a situation and times of uncertainty. The elevated levels of stress hormones and blood pressure, as well as the continuous and ongoing increase in heart rate, can have an impact on daily living and well-being. This type of long-term stress can put you at risk for high blood pressure, stroke, or heart attack. To avoid this, modern technology can be used to develop tools that can alert people to their rising stress levels, preventing long-term injury to the body. Stress has an effect on a number of physiological factors that can be examined in order to detect it effectively. The study of forecasting people’s mental states from their voice in stressful and non-stressful situations is known as voice analysis, as well as investigating how the voice changes in stressful situations in humans. The goal of this research is to use speech signals to predict stress. In MATLAB, several spectral properties from voice are retrieved. The gathered results are subjected to statistical analysis in order to arrive at a more accurate conclusion.KeywordsStressSpeechSpectral featuresVoice analysisStatistical analysis