Human physical activity recognition algorithm based on smartphone data convolutional nerual network and long short time memory

Abstract

A deep learning framework for activity recognition based on smartphone acceleration sensor data, convolutional neural network (CNN) and long short-term memory (LSTM) is proposed in the paper. The proposed framework aims to improve the accuracy of human activity recognition (HAR) by combining the strengths of CNN and LSTM. The CNN is used to extract features from the acceleration data and the LSTM is used to model the temporal dependencies of the data. The proposed framework is evaluated on the publicly available dataset, it includes 6 different actions: walking, walking upstairs, walking downstairs, sitting, standing, and laying. The recognition accuracy has reached 94 %.