A Study on Human Activity Recognition Using Accelerometer Data from Smartphones

The use of on-body wearable sensors is widespread in several academic and industrial domains. Of great interest are their applications in ambulatory monitoring and pervasive computing systems; here, some quantitative analysis of human motion and its automatic classification are the main computational tasks to be pursued. In this paper, we discuss how human physical activity can be classified using on-body accelerometers, with a major emphasis devoted to the computational algorithms employed for this purpose. In particular, we motivate our current interest for classifiers based on Hidden Markov Models (HMMs). An example is illustrated and discussed by analysing a dataset of accelerometer time series.

The objective of this feasibility study is to evaluate the use of the 'Physilog' device, an ambulatory physical-activity recorder based on acceleration measurement, for the monitoring of daily physical activities. Accelerations measured at the level of the chest and the thigh are recorded by Physilog over a period of 1 h in five normal subjects. A specially designed studio-like room allowing the performance of most usual activities of everyday life is used. A video film synchronised with the Physilog is obtained for each subject to check the accuracy of the data derived from Physilog. Based on the analysis on the average and the deviation of the acceleration signal, an algorithm is developed to classify the activities in four categories, i.e. lying, sitting, standing and locomotion. Compared with the video observations, the results from the algorithm show an overall misclassification of 10.7%, which is mainly due to confusion between dynamic activities and the standing posture. In contrast, the misclassification between postures is negligible. It is concluded that Physilog can be used in the clinical setting for the reliable measurement and long-term recording of most usual physical activities.