Movement analysis in learning by repetitive recall. An approach for automatic assistance in physiotherapy

This paper describes a method based on computer vision to analyze and compare patterns of people movements. An stochastic analysis of invariant algebraic moments from binarized image sequences of persons is proposed along with an approach for extracting frequency features from the time series obtained, in order to classify and categorize patients movements using a neural network. Based on studies of repetitive recall, an experiment with occupational therapy exercises was realized to classify the movement patterns performed by an hemiplegic patient to evaluate performance of the committed body part.