Brief Analysis on Human Activity Recognition

Abstract

Human activity recognition is regarded as one of the most prominent fields of exploration in computer science research. The process to elucidate human body movements to determine human gestures has been widely applied in surveillance, health assistance and interaction of human with computers. A variety of methodologies have been adopted by different researchers in this domain like wearable devices, device-free tools and object trackers to successfully recognize human gestures. This paper gives a brief analysis on processing of sensors data in HAR by using two deep neural network (DNN) models: Convolutional and recurrent neural network and summarizing their respective accuracies. The data being efficiently used to create the appropriate model is provided by the Wireless Sensor Data Mining (WSDM) lab. The data collection was done from 30 people using a wearable sensor and performing six different activities: (1) Sitting (2) Walking (3) Downstairs (4) Upstairs (5) Standing (6) Jogging and performing them with n number of repetitions. In the end, we discuss the approaches ‘Convolutional Neural Network’ and ‘Recurrent Neural Network’ and on the basis of their precision of how they recognize the activities and movements.