Smart Cup to Monitor Stroke Patients Activities During Everyday Life

Abstract

This paper presents a new platform to monitor stroke patients activities during their everyday life at home. This platform is intended to be a part of a smart objects ecosystem for home monitoring using common objects embedding sensors. The monitoring is performed with a self-contained smart cup that can be used to drink at different times of the day. The smart cup embeds various sensors in order to detect its movements and the liquid level. Activity analysis is performed on the collected data in order to provide information to the therapists on the patient's sedentariness and independence on the daily life tasks (sitting, walking, drinking and going up and down the stairs). This paper presents the design concept of the smart cup along with the implementation and mainly focuses on the activity analysis process. We used a linear classifier: the Support Vector Machine (SVM) classifier. Indeed, the classification of stroke patient's activities is a binary classification. Moreover, as we decided to use DCT features, SVM is the classifier that gives better classification performances. The results show a recognition precision above 92% on all activities with the smart cup. A comparative study has been carried out in order to assess the performances of the linear SVM classifier and a non-linear Multi-Layered Perceptron (MLP) classifier. The result of this study shows that the linear SVM classifier offers better performances on classifying everyday life activities with a smart cup.