Analisis Kinerja CNN pada Human Activity Recognition yang Efisien dengan Quantization dan Pemilihan Fitur Ridge
Abstract
Human Activity Recognition (HAR) merupakan bidang penting dalam sistem cerdas berbasis sensor, khususnya pada perangkat bergerak yang menuntut model akurat namun tetap efisien. Tantangan utama dalam topik ini adalah banyaknya fitur dari sensor smartphone dan kebutuhan model yang dapat bekerja pada perangkat dengan keterbatasan daya dan komputasi. Data masukan berupa sinyal sensor, sementara keluarannya adalah klasifikasi aktivitas manusia.
Topik ini menarik karena banyak aplikasi modern bergantung pada HAR, seperti pemantauan kesehatan, kebugaran, dan keamanan. Namun, model yang ada sering berukuran besar dan kurang efisien ketika dijalankan pada perangkat edge, sehingga menimbulkan kesenjangan antara performa yang diharapkan dan kemampuan perangkat.
Solusi yang dikembangkan adalah RRQ-CNN, sebuah model yang menggabungkan seleksi fitur berbasis ridge regression dengan kompresi melalui kuantisasi untuk menghasilkan model yang lebih ringan. Pendekatan ini bertujuan mengurangi fitur tidak relevan, menyederhanakan struktur model, dan meningkatkan efisiensi komputasi tanpa mengorbankan performa.
Hasil pengujian menunjukkan bahwa model ini mampu meningkatkan akurasi melalui seleksi fitur, mengurangi ukuran model secara signifikan, serta menghasilkan waktu inferensi yang cepat pada perangkat edge. Dengan demikian, RRQ-CNN memberikan keseimbangan yang baik antara akurasi dan efisiensi, sehingga layak digunakan dalam aplikasi HAR berbasis perangkat bergerak.
Kata kunci: human activity recognition, cnn, ridge regression, quantization, edge computing, efficient model
References
A. Wang, G. Chen, J. Yang, S. Zhao and C.-Y. Chang, "A Comparative Study on Human Activity Recognition Using Inertial Sensors in a Smartphone," IEEE Sensors Journal, vol. 16, no. 11, p. 4566–4578, 2016.
F. Ordóñez and D. Roggen, "Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition," Sensors, vol. 16, no. 1, p. 115, January 2016.
A. G. Howard, M. Zhu, B. Chen, D. Kalenichenko, W. Wang, T. Weyand, M. Andreetto and H. Adam, "MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications," arXiv, 2017.
M. Kimhi, T. Rozen, A. Mendelson and C. Baskin, "AMED: Automatic Mixed-Precision Quantization for Edge Devices," Mathematics, vol. 12, no. 12, p. 1810, 2024.
A. G. Putrada and S. Prabowo, "IMBGAFS: GA Feature Selection for AUC in Bird Strike Prediction," Machine Learning Techniques and NLP, vol. 13, 2023.
S. Fahimifar, K. Mousavi, F. Mozaffari and M. Ausloos, "Identification of the Most Important External Features of Highly Cited Scholarly Papers Through 3 (i.e., Ridge, Lasso, and Boruta) Feature Selection Data Mining Methods," Quality & Quantity, vol. 57, no. 4, p. 3685–3712, September 2022.
H. Zhou, X. Zhang, Y. Feng, T. Zhang and L. Xiong, "Efficient Human Activity Recognition on Edge Devices Using DeepConv LSTM Architectures," Scientific Reports, vol. 15, April 2025.
A. S. Pimpalkar and D. V. Niture, "Towards Contactless Elevators with TinyML Using CNN-Based Person Detection and Keyword Spotting," arXiv, 2024.
A. Ashiq O K, V. B. Semwal, M. B. Rathore and V. Soni, "A Compressed Deep Learning Model for Recognizing Human Activities with Wearable Sensors," SSRN, 2024.
B. Palermo, "Recognizing human activities in a privacy-preserving way," 2024.
T. Dupré la Tour, M. Eickenberg, A. O. Nunez-Elizalde and J. L. Gallant, "Feature-Space Selection with Banded Ridge Regression," NeuroImage, vol. 264, p. 119728, 2022.
A. G. Putrada, M. Abdurohman, D. Perdana and H. H. Nuha, "NoCASC: A Novel Optimized Cost-Complexity Pruning for AdaBoost Model Compression on Edge Computing-Based Smart Lighting," in 2024 IEEE International Conference on Communication, Networks and Satellite (COMNETSAT), 2024.
A. Ignatov, "Real-Time Human Activity Recognition from Accelerometer Data Using Convolutional Neural Networks," Applied Soft Computing, vol. 62, p. 915–922, 2018.
D. Anguita, A. Ghio, L. Oneto, X. Parra and J. L. Reyes-Ortiz, "Human Activity Recognition on Smartphones Using a Multiclass Hardware-Friendly Support Vector Machine," in Ambient Assisted Living and Home Care, Berlin, Heidelberg, 2012.
A. G. Putrada, N. Alamsyah, S. F. Pane, M. N. Fauzan and D. Perdana, "VANET Severity Classification in BSM Messages with a Novel Naïve Bayes Feature Selection," in 2023 International Conference on Advancement in Data Science, E-Learning and Information System (ICADEIS), 2023.
N. Y. Hammerla, S. Halloran and T. Ploetz, "Deep, Convolutional, and Recurrent Models for Human Activity Recognition Using Wearables," arXiv, 2016.
D. P. Kingma and J. Ba, "Adam: A Method for Stochastic Optimization," arXiv, 2017.
N. S. Keskar, D. Mudigere, J. Nocedal, M. Smelyanskiy and P. T. P. Tang, "On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima," arXiv, 2017.
A. G. Putrada, N. Alamsyah, M. N. Fauzan, S. Prabowo and I. D. Oktaviani, "QUIDS: A Novel Edge-Based Botnet Detection with Quantization for IoT Device Pairing," Indonesia Journal on Computing (Indo-JC), vol. 8, no. 3, pp. 29-41, 2023.
A. G. Putrada, N. Alamsyah, M. N. Fauzan and I. D. Oktaviani, "Pearson Correlation for Efficient Network Anomaly Detection with Quantization on the UNSW-NB15 Dataset," in 2024 International Conference on ICT for Smart Society (ICISS), 2024.
M. N. Adiputra, A. G. Putrada and I. D. Oktaviani, "DistilQ-NILM: A Hybrid Quantization-Knowledge Distillation LSTM Model for Edge-Based Load Monitoring," in 2024 International Conference on Intelligent Cybernetics Technology & Applications (ICICyTA), 2024.
S. Gupta, "Deep learning based human activity recognition (HAR) using wearable sensor data," International Journal of Information Management Data Insights, vol. 1, no. 2, p. 100046, 2021.
