Physical inactivity is becoming a major public health concern and lead to a variety of chronic diseases. Since adequate moderate or vigorous activity can reduce the incidence of chronic diseases, noncommunicable disease and obesity. The evidence is supporting the importance of physical activity on health and well-being. However, many people nowadays live without adequate physical activity, and do not aware whether their daily activity is enough or not. The activity recognition and activity level can be used to survey the effectiveness and achievement of goals aimed at increasing physical activity. Physical activity monitoring has become a more proactive healthcare service that should build on the real-time reminding offered by healthcare solutions. Therefore, physical activity monitoring and activity level assessment are critical to maintain adequate physical activity and improve health. In this work, we present a motion patterns analysis for physical activity recognition and activity level assessment by using a wearable sensor. The proposed mechanism uses triaxial accelerometer as a sensing device. The sensor node is mounted in the right front waist, sensing and transmitting sensing data to server. The time series of raw data will be preprocessed through the aggregation technique of jumping window. The raw data will be divided into small segments and separated to gravity signal and body acceleration by filter. Through feature extraction and proposed classifier, motion pattern analysis is achieved. The classifier consists of activity recognition and activity level assessment algorithms. The results have demonstrated that the proposed methods can achieve 94.7%, 87.0% accuracy of activity recognition and activity level estimation respectively.
Adaptive human aware navigation based on motion pattern analysis
