Predicting Human Activity from Mobile Sensor Data Using CNN Architecture

Abstract

Having a model for predicting motion related activities of humans have tremendous applications. Quite often a simple smartphone is enough for monitoring the liveliness of a person. This can be achieved by using Activity Recognition (AR). Smartphones are employed in a wider manner and it becomes one amongst the ways to spot the human’s environmental changes by using the sensors in smart mobile. A variety of sensors are embedded in a smartphone - for instance gyroscope, accelerometer. The contraption is demonstrated to look at the state of a person. Here, a framework Human Activity Recognition (HAR) collects the raw data from sensors and movements of humans observed using a deep learning approach. Deep learning models are proposed to spot motions of humans with plausible high accuracy by using sensed data. The performance of a framework is analyzed using Convolutional Neural Network. The act of the model is analyzed in terms of exactness and efficiency. The designed activity recognition model is manipulated to detect the activities of elderly humans at home and it can detect the activities of persons in a crowded area when the area is authorized.