“…Sensor-based human activity recognition (HAR) is the process of identifying human activities such as walking and cooking from ambient and wearable sensors. HAR is the key enabling technology for applications in health monitoring [1,2], smart home [3], traffic monitoring [4], and personalised training [5]. In recent years, HAR has been extensively investigated [6,7], and numerous machine learning approaches, particularly deep learning [8,9], have been developed to extract features from sensor signals from multiple modalities and discover complicated correlations between features and behaviours [10][11][12].…”