Radar-Based Human Movement Detection and Classification for Smart Homes Applications

Smart homes, known for their capacity to facilitate seamless environmental control, have garnered increasing attention in recent years. In this paper the potential of Frequency Modulated Continuous Wave Radar for autonomous detection and classification of human movements within indoor environments is investigated. Data from ten subjects were employed to evaluate and distinguish the efficacy of four Machine Learning algorithms - K-Nearest Neighbor, Support Vector Machine, Linear Discriminant Analysis, and Random Forest - in classifying movements into the categories of "Stopped", "Moving", and "Working Out". Each algorithm is compared, not only with performance metrics but also considerations of creation time, memory usage, and results variability are considered. The results, underscored by a 97.96% accuracy rate predominantly attained through the Random Forest algorithm, illustrate the practicality of employing these technologies for movement classification within smart homes. While the K-Nearest Neighbor algorithm consumed the least memory, Linear Discriminant Analysis proved to be the fastest, and the Support Vector Machine exhibited the least favorable performance in terms of both accuracy and resource efficiency.