IMU-based Human movement recognition using MINIROCKET and application in Dance performance
Movement recognition technology offers new possibilities for enhancing dance practice, education, and accessibility by enabling
real-time feedback, analysis, and documentation. Wearable Inertial
Measurement Unit (IMU) sensors provide an affordable, portable alternative to camera-based systems, capturing motion data
directly from the body. This multichannel time series data can be analyzed to recognize and differentiate movement. To classify these
signals efficiently, we apply Minimally Random Convolutional Kernel Transform (MINIROCKET)—a fast, lightweight, and accurate time-series classification method. Our goal is to develop a low-cost, real-time movement recognition tool that is adaptable to the demands of live performance and creative practice.