A Human Activity Recognition Model Based on Wearable Sensor
Fan Zhou, Ruomei Wang, Hang Su, Shenyi Xu
Abstract
With the development of sensor technology, wearable sensors are gradually applied to human activity recognition due to its advantages of improved performance and portability. By identifying different human activity, it has a wide application prospect in medical treatment, abnormal activity monitoring, interactive games, intelligent home and other aspects. This paper designs a human motion activity recognition model based on wearable sensor. Firstly, by preprocessing the data collected by the sensor, a subdivided activity dataset composed of eight kinds of activity is established (Sit, stand, walk slowly, walk briskly, go upstairs, go downstairs, jog and run briskly). Second, a hybrid deep network model based on convolution neural network and recurrent neural network is proposed. The feature of convolutional neural network and recurrent neural network is used to classify the subdivision activity data set and realize the recognition of specific activity. The results of recognition can provide scientific statistical data for real-time activity monitoring, activity warning and reminder, health report, user health information management and other applications.