The use of Convolutional Neural Network and StarRGB technique for gait movements recognition in remote physiotherapy

Computer vision is present in several areas of science, from the classification of people to the identification of cancer cells, which have become published in several publications. One of these applications is the identification of people's movements and gestures, including people with disabilities, for example. In this work, a technique called StarRGB is used in conjunction with a Convolutional Neural Network (CNN) for classifying the gait movements made by patients in a rehabilitation process. The StarRGB technique condenses temporal information from all images of a video into a single RGB image. Then, the obtained image is passed as input to a developed CNN so that it extracts the features of the image and classifies the movement performed in six classes. After the experiments, the results achieve a mean accuracy over the six classes of 99.64% and an AUC of 0.997, which shows the efficacy of this proposal to classify gaits using only ordinary RGB cameras.