Sensor-based Complex Human Activity Recognition from Smartwatch Data using Hybrid Deep Learning Network

Sensor-based human activity recognition (HAR) research is being used for tasks like healthcare tracking, fall detection, and misbehavior prevention. Because of the sophistication of hand gesture signals, complex human activity (CHA) recognition is a difficult task in HAR research. Compared to simple human behavior (SHA), the CHA has more input and long sequential information to deal with. To solve the CHA problem, we proposed a hybrid deep learning model that combines a CNN network and an LSTM network in this paper. The proposed model is an end-to-end model that automatically extracts high features. The model is tested on the daily human activity (DHA) dataset, which is a publicly accessible dataset of complex human activities. The results of the experiments show that the proposed hybrid deep learning model outperforms the current state-of-the-art recognition model.