Fall Detection System using a single Accelerometer through Machine Learning

Falls pose a critical health risk, often resulting in severe injuries or fatalities and profoundly affecting individuals' quality of life. This study introduces an innovative machine learning-based fall detection system that utilizes a single low-cost ADXL accelerometer sensor to capture body acceleration signals during daily activities. By combining data from three axial accelerations into a Sum Magnitude Vector (SMV) and extracting five distinctive features from the time and frequency domain, this system achieves a remarkable accuracy rate of 99.3% when fed into a Support Vector Machine (SVM) classifier, as confirmed through rigorous comparisons with other classifiers. This research presents a promising solution to minimize the consequences of falls, enhancing the safety and well-being of individuals at risk concisely and cost-effectively.