Human Activity Recognition using Smartphone Sensors: A Deep Learning Approach with LSTM and CNN

The proliferation of Internet of Things (IoT) devices and wearable technology has generated vast amounts of sensor data, enabling advanced health and fitness monitoring applications. This paper presents a robust framework for Human Activity Recognition (HAR) utilizing data from smartphone accelerometers and gyroscopes. The objective is to classify distinct physical activities—such as walking, sitting, standing, and running—by analyzing time-series sensor data. The methodology employs a hybrid Deep Learning architecture combining Convolutional Neural Networks (CNN) for spatial feature extraction and Long Short-Term Memory (LSTM) networks for temporal sequence modeling. Performance is evaluated using the UCI HAR dataset. Experimental results demonstrate that the proposed deep learning model outperforms traditional statistical feature extraction methods, achieving high classification accuracy. This study contributes to the development of unobtrusive, real-time health monitoring systems.