Human Activity Recognition in Smart Cities from Smart Watch Data using LSTM Recurrent Neural Networks

Abstract

Due to the convergence of healthcare and smart cities, information and technology are employed in health and medical procedures worldwide. Residents of smart cities now enjoy better lives and healthier bodies because to integration. The advent of smart watch technology and deep learning techniques has engendered a potential increase in performance and accuracy of such tasks. Our paper explores the application of recurrent neural networks particularly LSTM for action recognition task using smart watch sensor data, i.e., gyroscope and accelerometer data. We have taken into consideration some general activities like standing, walking, sitting, etc. and implemented a LSTM network for the action recognition job using the time sequences sensor data. This model can be useful for other health-based applications which can monitor the health conditions of a user by keeping record of his activities.