Smartwatch-based Human Activity Recognition Using Hybrid LSTM Network

As a result of the rapid development of wearable sensor technology, the use of smartwatch sensors for human activity recognition (HAR) has recently become a popular area of research. Currently, a large number of mobile applications, such as healthcare monitoring, sport performance tracking, etc., are applying the results of major HAR research studies. In this paper, an HAR framework that employs spatial-temporal features that are automatically extracted from data obtained from smartwatch sensors is proposed. The hybrid deep learning approach is used in the framework through the employment of Long Short-Term Memory Networks and the Convolutional Neural Network, eliminating the need for the manual extraction of features. The advantage of tuning the hyperparameters of each of the considered networks by Bayesian optimization is also utilized. It was indicated by the results that the baseline models are outperformed by the proposed hybrid deep learning model, which has an average accuracy of 96.2% and an F-measure of 96.3%.