“…The problem of tracking, identifying and classifying human actions has received increasing interest over the years, as it plays a key role in many applied contexts, such as human-computer interface (Popoola and Wang, 2012 ; Sarig Bahat et al, 2015 ; Quitadamo et al, 2017 ; Bachmann et al, 2018 ), daily-life activity monitoring (Mannini and Sabatini, 2010 ; Cheng et al, 2015 ; Chetty and White, 2016 ), clinical assessment (Rawashdeh et al, 2016 ; Arifoglu and Bouchachia, 2017 ; Howell et al, 2017 ) and sports performance (Attal et al, 2015 ; Ghazali et al, 2018 ; Hsu et al, 2018 ). The development of unobtrusive technologies for motion capture (e.g., wearable inertial measurement units—IMUs), their widespread integration in relatively cheap, commercially available devices (e.g., smartphones, watches, activity trackers, heart rate monitors, sensorized insoles), and the push toward healthier, more active life styles, have generated a multitude of existing and potential applications where automatic movement classification and assessment is fundamental (Attal et al, 2015 ; Cheng et al, 2015 ; Cust et al, 2019 ).…”