Efficient Real-Time Human Activity Recognition with Accelerometer Using CNN Models

This paper presents a real-time Human Activity Recognition (HAR) system that utilizes accelerometer sensor data processed through deep learning (DL) and machine learning (ML) techniques. We employ Convolutional Neural Networks (CNN) to capture spatial dependencies within accelerometer signals and Support Vector Machines (SVM) for robust classification. We have proposed an enhanced model, Adaptive Gated Neural Network (AGNN), which integrates CNN's feature extraction with SVM's classification capabilities. Experimental results show that AGNN achieves an accuracy of 98%, outperforming conventional HAR models. This system has applications in health monitoring, fitness tracking, and smart home automation.