Silhouette segmentation for mobility anomalydetection via human movement analysis.

Abstract

In this work, we present a deep learning-based approach for human activity recognition and fall detection. The proposed system uses image processing techniques to extract human silhouettes from video frames, and analyzes movement using artificial intelligence models. We used a CNN-LSTM model to classify human postures and a One-Class SVM to detect falls as abnormal movements. The CNN-LSTM model reached an accuracy of 50.4%, and the One- Class SVM achieved a detection accuracy of 80%. The results show that this method can help detect falls and improve safety in smart home environments.