EdgeClap: Implementasi Sistem Otentikasi Pintu Pintar Berbasis Tepuk Tangan Menggunakan Model LSTM yang Dioptimalkan dengan Knowledge Distillation pada Perangkat IoT
Abstract
Several studies have shown that biometric-based authentication offers advantages and increased security. However, the performance of clap-based authentication with voice sensors can be improved by artificial intelligence (AI) methods such as long short-term memory (LSTM). This study presents EdgeClap, a smart door lock authentication system based on LSTM-based voice-clap pattern recognition. First, we designed the architecture of the voice-based smart lock system and implemented it. Then, we collected a dataset with several voice labels and used it to train the LSTM. This model was then trained and optimized using the Knowledge Distillation (KD) technique. We then compared the prediction performance and model size between LSTM and LSTM + KD using a confusion matrix, number of parameters, model size, and inference time. The test results showed that the KD method provides efficiency while maintaining prediction performance. The student model achieved a validation accuracy of 0.9731, higher than the teacher model's 0.9676. In addition, KD improved efficiency by reducing the model size by 57.1% and the number of parameters by 25.8%. This optimization enables deep learning models to be implemented directly on the ESP32, enabling real-time inference 68ms without reliance on the internet, cloud computing, or wearable devices. Keywords: acoustic authentication, smart lock, long short-term memory, knowledge distillation, Internet of Things, edge computing
Downloads
References
[1] Y. B. Zikria, R. Ali, M. K. Afzal, and S. W. Kim, “Next-generation internet of things (iot): Opportunities, challenges, and solutions,” Sensors, vol. 21, no. 4, p. 1174, 2021.
[2] H. Touqeer, S. Zaman, R. Amin, M. Hussain, F. Al-Turjman, and M. Bilal, “Smart home security: challenges, issues and solutions at different iot layers,” The Journal of Supercomputing, vol. 77, pp. 14053– 14089, 2021.
[3] K. Luaha, S. Ismail, A. Rahman, N. N. Siagian, L. R. Amalia, and R. F. Wijaya, “Design of a clap detection light control system based on arduino uno,” Zero: Jurnal Sains, Matematika, dan Terapan, vol. 6, no. 2, pp. 142–146, 2022.
[4] Q. Banafsa, Implementasi Sistem Smart Door Lock Berbasis Pengenalan Wajah dengan Kamera Huskylens, N. Hendrarini and D. N. Ramadan, Eds. Bandung: Universitas Telkom, D3 Teknologi Komputer, 2025, [Online]. Available: https://openlibrary.telkomuniversity.ac.id/home/catalog /id/238057/.
[5] A. Altayeva, M. Omarov, and H.-N. Lee, “Convolutional lstm network for real-time impulsive sound detection and classification in urban environments,” Applied Sciences, vol. 13, no. 2, p. 721, 2023.
[6] A. Alkhulaifi, F. Alsahli, and I. Ahmad, “Knowledge distillation in deep learning and its applications,” PeerJ Computer Science, vol. 7, p. e474, 2021.
[7] A. G. Putrada, M. Abdurohman, D. Perdana, and H. H. Nuha, “Edgesl: Edge-computing architecture on smart lighting control with distilled knn for optimum processing time,” IEEE Access, vol. 11, pp. 64697– 64712, 2023.
[8] Y. Wang, Z. Yu, J. Wu, C. Wang, Q. Zhou, and J. Hu, “Adaptive knowledge distillation-based lightweight intelligent fault diagnosis framework in iot edge computing,” IEEE Internet of Things Journal, vol. 11, no. 13, pp. 23156–23169, 2024.
[9] M. A. Al-Shannaq, A. Nasrawi, A. A.-R. K. Bsoul, and A. A. Saifan, “Abnormal heart sound recognition using svm and lstm models in real-time mode,” Scientific Reports, vol. 15, p. 9129, 2025.
[10] S. Ruvinga, G. J. Hunter, O. Duran, and J.-C. Nebel, “Use of lstm networks to identify ”queenlessness” in honeybee hives from audio signals,” in 2021 17th International Conference on Intelligent Environments (IE), pp. 1–6, 2021.
[11] A. Buriro and F. Ricci, “Clapauth: A gesture-based user-friendly authentication scheme to access a secure infrastructure,” in Emerging Technologies for Authorization and Authentication: 5th International Workshop, ETAA 2022, Copenhagen, Denmark, September 30, 2022, Revised Selected Papers, (Berlin, Heidelberg), p. 15–30, Springer-Verlag, 2022.
[12] M. Syafrizal, V. Ikawati, A. Siswanto, N. Lestari et al., "Sistem keamanan door lock berbasis voice recognition dengan natural language processing," Infotronik: Jurnal Teknologi Informasi Dan Elektronika, vol. 9, no. 1, pp. 1-11, 2024.
[13] A. Jylh¨a, C. Erkut, U. Simsekli, and A. Cemgil, “Sonic handprints: Person identification with hand clapping sounds by a model-based method,” 01 2012.
[14] D. Y. Z. Chan and B. S. Tan, "Pentap: A novel rhythm game with innovative acoustic soundwave-based input and audio recognition," in 2024 3rd International Conference on Digital Transformation and Applications (ICDXA). IEEE, 2024, pp. 45-50.
[15] T. A. Onisha, J. Kim, and J. Seol, "Multi label sound classification using deep learning models," in 2024 IEEE/ACIS 22nd International Conference on Software Engineering Research, Management and Applications (SERA). IEEE, 2024, pp. 129-134.
[16] S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural Computation, vol. 9, pp. 1735–1780, 11 1997.
[17] J. Zhang, Y. Gao, R. Liu, X. Cheng, H. Zhang, and S. Chen, “Can students beyond the teacher? distilling knowledge from teacher’s bias,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 39, no. 21, 2025, pp. 22 434–22 442.
