Reducing feature space in human hand movement models
Abstract
Reducing feature space in human hand movement models
Incoming article date: 17.07.2021The article presents results of processing motion data obtained on a motion capture system using inertial microelectromechanical sensors (MEMS). Hardware and software complex compares reference movements with those that operator is performing. Reducing the feature space in motion model is an important task in the context of using many similar sensors that need to be processed on low-power devices. The main way to obtain the distance between two patterns is algorithm of dynamic time warping which has a computational complexity of O2, which means that it is expedient to select features. Statistics of distinction degree between different types of movements are provided to assess the quality of simplified models. Сomplexity of resulting models was demonstrate on Kohanovsky method.
Keywords: MEMS, motion capture, correlation analysis, wavelet transform, dynamic time-warping algorithm, accelerometer, gyroscope, inertial sensor, motion control, pattern matching