Hemiparesis is one of the most frequent poststroke conditions, which causes muscle weakness and/or inability to move one side of the body. Physical rehabilitation is the main treatment for hemiparesis recovery, and physiotherapists agree that using the impaired arm in the activities of daily living (ADLs)is crucial for a complete recovery. Currently, rehabilitation is assessed through diaries and self-questionnaires, which are subjective and do not tell the real condition of the patients throughout the day. Assistive devices can objectively evaluate the functional improvement of the impaired arm monitoring its activity. This work aimed to identify the purposeful arm movements during patient's ADLs. We consider arm's swing while walking as a non-purposeful movement. Firstly, the event-based approach was applied to separate movement and non-movement segments. Secondly, movement segments were used to detect the change-point (events)and their locations in time series signal. Two machine learning classifiers, Support Vector Machine (SVM)and Artificial Neural Network (ANN), were trained using 10-fold cross validation for the classification of purposeful and non-purposeful movements. Data from 10 healthy and 12 post-strokes volunteers from Institute Guttmann (Barcelona)were collected using the SensHand device. The volunteers, wearing one SensHand on each wrist, performed the following activities: resting, eating, pouring water, drinking, brushing, folding towel, grasp towel, grasp brush, grasp glass, continuous and walking. Developing a model based on the healthy subjects, the overall classification accuracy obtained from SVM classifier and ANN was 81.21 % and 97.06% respectively. Similarly, with poststroke subjects obtained accuracy with the SVM and ANN was 84.18% and 99.74% respectively. Considering the whole dataset, SVM and ANN obtained maximum accuracy equal to 86.21 % and 99.91 % respectively. In conclusion, our work showed promising results for the classification of purposeful and non-purposeful movements.

Daily activity recognition can help people to maintain a healthy lifestyle and robot to better interact with users. Robots could therefore use the information coming from the activities performed by users to give them some custom hints to improve lifestyle and daily routine. The pervasiveness of smart things together with advances in cloud robotics can help the robot to perceive and collect more information about the users and the environment. In particular thanks to the miniaturization and low cost of Inertial Measurement Units, in the last years, body-worn activity recognition has gained popularity. In this work, we investigated the performances with an unsupervised approach to recognize eight different gestures performed in daily living wearing a system composed of two inertial sensors placed on the hand and on the wrist. In this context our aim is to evaluate whether the system is able to recognize the gestures in more realistic applications, where is not possible to have a training set. The classification problem was analyzed using two unsupervised approaches (K-Mean and Gaussian Mixture Model), with an intra-subject and an inter-subject analysis, and two supervised approaches (Support Vector Machine and Random Forest), with a 10-fold cross validation analysis and with a Leave-One-Subject-Out analysis to compare the results. The outcomes show that even in an unsupervised context the system is able to recognize the gestures with an averaged accuracy of 0.917 in the K-Mean inter-subject approach and 0.796 in the Gaussian Mixture Model inter-subject one.