Human Activity Recognition Using CNN-LSTM-GRU Model

Abstract

Human Activity Recognition (HAR) is a fundamental task in the field of computer vision and machine learning, with applications spanning from healthcare monitoring to human- computer interaction. This research paper presents a novel approach to HAR utilizing a hybrid model combining Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks, referred to as the VGG-LSTM model. The proposed VGG-LSTM model leverages the power of deep learning to address the challenges associated with HAR, including capturing spatial features and modeling temporal dependencies in complex human activities. In this research, we employ the VGG architecture as the feature extractor to capture discriminative spatial information from input sensor data, such as images or videos. Furthermore, the LSTM layer is integrated to model the temporal dynamics of human activities. This enables the model to effectively recognize and differentiate between a wide range of human activities, such as walking, running, sitting, and more, in real-world scenarios. The research demonstrates the effectiveness of the VGG-LSTM model on benchmark datasets, achieving state-of-the-art performance in human activity recognition tasks. The model’s accuracy, robustness, and ability to generalize to diverse scenarios make it a promising solution for applications in healthcare, sports analytics, security, and beyond. The contributions of this paper lie in the development of a powerful hybrid model that combines spatial and temporal information seamlessly, improving the accuracy and applicability of HAR systems. The results underscore the potential of the VGG-LSTM model in advancing human activity recognition technology, with implications for improving the quality of life and safety in various domains.