ResNet과 멀티헤드 어텐션을 활용한 타이핑 소리 기반 문자 예측

Vol. 35, No. 2, pp. 253-264, 4월. 2025
10.13089/JKIISC.2025.35.2.253, Full Text:
Keywords: Deep Learning, ResNet, Multi-head attention, Acoustic Signal Processing, IoT
Abstract

This study proposes a typing sound-based character prediction model using ResNet and multi-head attention techniques. Compared to the existing CoAtNet model, the proposed hybrid model effectively processes frequency-domain data transformed by FFT and enhances prediction performance by combining ResNet with multi-head attention. Experimental results show that the proposed model achieves a higher accuracy of 96.81% compared to the ResNet-only model at 96.66%. The incorporation of multi-head attention allows for precise learning of relationships between keystroke sounds, thereby improving prediction capability. This research introduces a novel approach for handling complex sequential data and demonstrates its potential in various application areas.

Statistics
Show / Hide Statistics

Statistics (Cumulative Counts from December 1st, 2017)
Multiple requests among the same browser session are counted as one view.
If you mouse over a chart, the values of data points will be shown.


Cite this article
[IEEE Style]
이규범, 김성진, 홍성관, "Typing Sound-Based Character Prediction Using ResNet and Multi-Head Attention," Journal of The Korea Institute of Information Security and Cryptology, vol. 35, no. 2, pp. 253-264, 2025. DOI: 10.13089/JKIISC.2025.35.2.253.

[ACM Style]
이규범, 김성진, and 홍성관. 2025. Typing Sound-Based Character Prediction Using ResNet and Multi-Head Attention. Journal of The Korea Institute of Information Security and Cryptology, 35, 2, (2025), 253-264. DOI: 10.13089/JKIISC.2025.35.2.253.