Feature Extraction and Prediction for Hand Hygiene Gestures with KNN Algorithm

This work focuses upon the analysis of hand gestures involved in the process of hand washing. There are six standard hand hygiene gestures for washing hands as provided by World Health Organisation hand hygiene guidelines. In this paper, hand features such as contours of hands, the centroid of the hands, and extreme hand points along the largest contour are extracted with the use of the computer vision library, OpenCV. These hand features are extracted for each data frame in a hand hygiene video. A robust hand hygiene dataset of video recordings was created in the project. A subset of this dataset is used in this work. Extracted hand features are further grouped into classes based on the KNN algorithm with a cross-fold validation technique for the classification and prediction of the unlabelled data. A mean accuracy score of >95% is achieved and proves that the KNN algorithm with an appropriate input value of K=5 is efficient for classification. A complete dataset with six distinct hand hygiene classes will be used with the KNN classifier for future work.