Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{188897,
        author = {Y Vineeth and Tejas and Rekha Krishnan and Chandana K R},
        title = {Deep Learning Approach for Gait Phase Classification: A Multi-Model Comparison},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {12},
        number = {7},
        pages = {4079-4084},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=188897},
        abstract = {Accurate gait phase classification is essential for rehabilitation engineering, prosthetics control, and clinical move- ment assessment. Conventional machine learning techniques like RF have a moderate level of accuracy, while deep learning ap- proaches like CNNs and BiLSTMs reach approximately 80-83% but struggle with long-range dependencies and interpretability. This study proposes a Temporal Fusion Transformer (TFT) for gait phase classification, leveraging attention mechanisms to model temporal patterns while providing clinically interpretable insights. Using a private dataset of 10 subjects and over 90 gait cycles, the TFT achieved 92% accuracy and 0.92 F1-score, outperforming CNN and BiLSTM baselines by 12% and 9%, respectively. Attention analysis revealed the model prioritizes heel-strike and toe-off events as critical temporal markers, consistent with clinical knowledge. These results demonstrate TFT’s potential as a robust, interpretable framework for real- time gait analysis in wearable healthcare systems.},
        keywords = {deep learning, temporal fusion transformer, gait phase classification, wearable sensors, human activity recognition, biomechanics},
        month = {December},
        }