Speech analysis for assessing physical competence under the concept of physical literacy (PhD Academy Award)

Abstract

The primary focus of my thesis was to investigate the relationship between speech features and physical activity, with the aim of developing a novel method for assessing physical competence and predicting exercise performance through speech analysis. I used a combination of machine learning and deep learning techniques to extract and model speech features that are indicative of physical exertion. This innovative approach allowed for the prediction of physical competence across various states of exercise, offering a non-invasive alternative to traditional fitness assessments. The methodology was designed to be applicable in real-world, dynamic environments, addressing the need for a more accessible and cost-effective solution that may offer a practical way to assess physical performance outside of laboratory settings. The motivation behind this research stemmed from the need to develop an approach that has the potential to be accessible, non-invasive and cost-effective method for assessing physical competence in diverse environments. Traditional methods, such as heart rate monitoring or physical performance tests, are often impractical in remote or resource-limited settings.1 Speech, being a naturally occurring and easily accessible output of human exertion, provides a unique opportunity to capture real-time physiological responses to physical activity.2 By analysing vocal changes induced by exercise, this study aimed to create a practical solution …