Analisis Akurasi dan Efisiensi Komputasi Arsitektur Deep Learning untuk Deteksi Jatuh Berbasis Pose Estimation: Accuracy and Computational Efficiency Analysis of Deep Learning Architectures for Pose Estimation-Based Fall Detection

Ruben Emmanuel Mahardo; Andreas Handojo

doi:10.57152/malcom.v6i4.3007

Authors

Ruben Emmanuel Mahardo Universitas Kristen Petra
Andreas Handojo Universitas Kristen Petra

DOI:

https://doi.org/10.57152/malcom.v6i4.3007

Keywords:

Deep Learning, Deteksi Jatuh, LSTM, Pose Estimation, ST-GCN, TCN

Abstract

Kejadian jatuh pada lansia menyebabkan lebih dari 684.000 kematian setiap tahun, sehingga diperlukan sistem deteksi otomatis yang menyeimbangkan akurasi dan efisiensi komputasi. Penelitian ini mengevaluasi kinerja tiga arsitektur deep learning berbasis sekuens untuk deteksi jatuh berbasis estimasi pose, yaitu Long Short-Term Memory (LSTM), Temporal Convolutional Network (TCN), dan Spatio-Temporal Graph Convolutional Network (ST-GCN). Kebaruan penelitian terletak pada pemisahan pengaruh kualitas backbone estimasi pose, stabilisasi post-processing, serta identifikasi ekstraksi pose sebagai bottleneck latensi real-time. Estimasi pose menggunakan YOLOv11-pose pada dataset Le2i yang terdiri atas 190 video dari empat skenario ruangan. Eksperimen menerapkan sliding window 25 frame, evaluasi lintas lingkungan, tuning hyperparameter melalui grid search, serta stabilisasi prediksi. Hasil menunjukkan TCN memperoleh F1-Score 1,000 pada Lecture_room serta Recall 0,824 dan F1-Score 0,875 pada Office, dengan waktu 2,97 detik per epoch. ST-GCN menyamai TCN pada Office, tetapi menghasilkan F1-Score 0,923 pada Lecture_room dengan biaya komputasi 1,6 kali lebih tinggi. LSTM menawarkan inferensi 0,276 milidetik per sekuens dan 17,9 FPS, tetapi Recall Office lebih rendah, yaitu 0,706. Ekstraksi pose menyumbang sekitar 90% latensi pipeline. Dengan demikian, TCN direkomendasikan untuk deteksi jatuh real-time, sedangkan LSTM lebih sesuai untuk perangkat dengan sumber daya terbatas.

Downloads

Download data is not yet available.

References

United Nations, “World Population Ageing 2019: Highlights,” UN DESA, 2019.

World Health Organization, “Falls,” WHO, Apr. 2021. https://www.who.int/news-room/fact-sheets/detail/falls

R. Kakara, G. Bergen, E. Burns, and M. Stevens, “Nonfatal and Fatal Falls Among Adults Aged ?65 Years, United States, 2020–2021,” MMWR, vol. 72, no. 35, pp. 938–943, 2023, doi: 10.15585/mmwr.mm7235a1.

Y. Wang and T. Deng, “Enhancing elderly care: Efficient and reliable real-time fall detection algorithm,” Digital Health, vol. 10, pp. 1–11, 2024, doi: 10.1177/20552076241233690.

E. Alam, A. Sufian, P. Dutta, and M. Leo, “Vision-based human fall detection systems using deep learning: A review,” Comput. Biol. Med., vol. 146, p. 105626, 2022, doi: 10.1016/j.compbiomed.2022.105626.

E. Tîrziu, A.-M. Vasilevschi, A. Alexandru, and E. Tudora, “Real-Time Fall Monitoring for Seniors via YOLO and Voice Interaction,” Future Internet, vol. 17, no. 8, p. 324, 2025, doi: 10.3390/fi17080324.

F. Shu and J. Shu, “An eight-camera fall detection system using human fall pattern recognition via machine learning by a low-cost android box,” Sci. Rep., vol. 11, no. 1, p. 2471, 2021, doi: 10.1038/s41598-021-81115-9.

A. Sherstinsky, “Fundamentals of Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) network,” Physica D, vol. 404, p. 132306, 2020, doi: 10.1016/j.physd.2019.132306.

S. Bai, J. Z. Kolter, and V. Koltun, “An empirical evaluation of generic convolutional and recurrent networks for sequence modeling,” arXiv:1803.01271, 2018. [Online]. Available: https://arxiv.org/abs/1803.01271

S. Yan, Y. Xiong, and D. Lin, “Spatial temporal graph convolutional networks for skeleton-based action recognition,” in Proc. AAAI, vol. 32, no. 1, pp. 7444–7451, 2018, doi: 10.1609/aaai.v32i1.12328.

M. Salimi, J. J. M. Machado, and J. M. R. S. Tavares, “Using deep neural networks for human fall detection based on pose estimation,” Sensors, vol. 22, no. 12, p. 4544, 2022, doi: 10.3390/s22124544.

A. Raza, M. H. Yousaf, W. Ahmad, S. A. Velastin, and S. Viriri, “Human fall detection using pose estimation: From traditional machine learning to vision transformers,” Eng. Appl. Artif. Intell., vol. 143, p. 109809, 2025, doi: 10.1016/j.engappai.2024.109809.

I. Charfi, J. Miteran, J. Dubois, M. Atri, and R. Tourki, “Optimized spatio-temporal descriptors for real-time fall detection: comparison of support vector machine and Adaboost-based classification,” J. Electron. Imaging, vol. 22, no. 4, p. 041106, 2013, doi: 10.1117/1.JEI.22.4.041106.

R. Khanam and M. Hussain, “YOLOv11: An overview of the key architectural enhancements,” arXiv:2410.17725, 2024. [Online]. Available: https://arxiv.org/abs/2410.17725

T.-Y. Lin et al., “Microsoft COCO: Common objects in context,” in ECCV 2014, pp. 740–755, Springer, 2014, doi: 10.1007/978-3-319-10602-1_48.

O. Rainio, J. Teuho, and R. Klén, “Evaluation metrics and statistical tests for machine learning,” Sci. Rep., vol. 14, p. 6086, 2024, doi: 10.1038/s41598-024-56706-x.

L. Qi and H. Sun, “Multistage fall detection framework via 3D pose sequences and TCN integration,” Sci. Rep., vol. 15, p. 27832, 2025, doi: 10.1038/s41598-025-11325-y.

F. Castro, V. Dentamaro, V. Gattulli, and D. Impedovo, “Fall detection with LSTM and attention mechanism,” in Proc. WAMWB@MobileHCI 2023, CEUR Workshop Proceedings, vol. 3517, pp. 37–50, 2023.

J. Shin, A. S. M. Miah, R. Egawa, K. Hirooka, M. A. M. Hasan, Y. Tomioka, and Y. S. Hwang, “Fall recognition using a three stream spatio temporal GCN model with adaptive feature aggregation,” Sci. Rep., vol. 15, p. 10635, 2025, doi: 10.1038/s41598-025-95508-7.

Y. Cho, G. Solak, O. Nocentini, M. Lorenzini, A. Fortuna, and A. Ajoudani, “Anticipatory Fall Detection in Humans with Hybrid Directed Graph Neural Networks and Long Short-Term Memory,” in Proc. IEEE RO-MAN, 2025, arXiv:2509.05337.

M. Zobi, L. Bolzani, Y. Tabii, and R. Oulad Haj Thami, “Robust 3D Skeletal Joint Fall Detection in Occluded and Rotated Views Using Data Augmentation and Inference-Time Aggregation,” Sensors, vol. 25, no. 21, p. 6783, 2025, doi: 10.3390/s25216783.

J. Yan, X. Wang, J. Shi, and S. Hu, “Skeleton-Based Fall Detection with Multiple Inertial Sensors Using Spatial-Temporal Graph Convolutional Networks,” Sensors, vol. 23, no. 4, p. 2153, 2023, doi: 10.3390/s23042153.