TriAngle-GaitSense: A Hybrid Approach for Biomechanical Gait Anomaly Detection through Multivariate Joint Angles

Human gait is a critical indicator of both physical and mental well-being, with abnormalities often signaling underlying health conditions. This research introduces "TriAngle-GaitSense," a hybrid deep learning framework that leverages multivariate joint angles data (hip, knee, and ankle) to detect biomechanical gait anomalies. Using Bidirectional Long Short-Term Memory (BiLSTM) networks optimized by a Genetic Algorithm (GA), we classify gait conditions across three distinct categories: unbraced, knee-braced, and ankle-braced. The Multivariate Gait Data dataset provides rich temporal insights into gait dynamics, enabling precise identification of movement patterns under various support conditions. By comparing our model against traditional classifiers like Logistic Regression, Random Forest, and CNN-LSTM, we demonstrate superior performance, achieving an accuracy of 84% in training and 82% in testing. These results highlight the model's robustness and potential to enhance clinical gait assessments and personalized rehabilitation strategies.