EEG 신호분석을 통한 행동 예측: LSTM과 Transformer 비교 연구
Predicting Behavior through EEG Signal Analysis: A Comparative Study of LSTM and Transformer Models
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
본 연구는 EEG기반 인간 행동 식별을 위해 시계열 데이터에 효과적인 LSTM(Long Short-Term Memory) 모델과 Transformer 모델 비교 실험을 수행했다. 뇌파 데이터는 19~60세 건강한 성인으로 남성 11명, 여성 14명으로 총 25명의 피실험자 대상으로 세 가지 행동(눈 뜨기, 눈 감기, 손 흔들기)을 0.002초 간격으로 수집되었다. 수집된 데이터를 Sampling Rate 500Hz로 하여 1초의 길이를 1개의 데이터로 사용할 수 있게 변환했고, 뇌파 신호에 적용한 전처리 과정은 Notch- Filter 및 Band-Pass Filter를 사용했다. LSTM 모델은 3개의 층으로 구조로 설계하여 모델을 학습했고, Transformer 모델은 5개의 층 구조로 설계하여 모델을 학습했다. LSTM 모델은 0.9095의 정확도를 달성한 반면 Transformer 모델은 0.8827의 정확도를 나타내며, LSTM 모델이 더 높은 성능을 나타냈다.
Abstract
This study conducted a comparative experiment between the Long Short-Term Memory (LSTM) model and the Transformer model, both of which are effective for time-series data, to identify behaviors based on EEG. The EEG data was collected from 25 healthy adults aged 19 to 60, consisting of 11 men and 14 women, and included three behaviors (eyes open, eyes closed, waving hand) recorded at 0.002-second intervals. The collected data was converted to a sampling rate of 500Hz, allowing 1 second of data to be used as a single data point. The preprocessing applied to the EEG signals involved a Notch Filter and a Band-Pass Filter. The LSTM model was trained with a 3-layer structure, while the Transformer model was trained with a 5-layer structure. The LSTM model achieved an accuracy of 0.9095, whereas the Transformer model achieved an accuracy of 0.8827, indicating that the LSTM model outperformed the Transformer model.
참고문헌 (11)
- D. Y. Kim, "Trends in brain wave signal and application technology", Electronics and telecommunications trends, 32(2), pp.19~28 (2017)
- M. . K. Ahn, "Principles and technological trends of brain-computer interface systems", Communications of the Korean Institute of Information Scientists and Engineers, 29(4), pp.42~53 (2011)
- B. H. . Brinkmann, "Seizure diaries and forecasting with wearables: epilepsy monitoring outside the clinic", Frontiers in Neurology, 12, pp.690404~ (2021)
- S. Hochreiter, "Long short-term memory", Neural computation, 9(8), pp.1735~1780 (1997)
- A. Vaswani, "Attention is all you need", Advances in neural information processing systems, 30 (2017)
- J. H. Hwang, "Motor Imagery EEG Signal Classification Using LSTM", The Journal of Next-generation Convergence Technology Association, 8(2), pp.223~229 (2022)
- Z. Wang, "Transformers for EEG-based emotion recognition: A hierarchical spatial information learning model", IEEE Sensors Journal, 22(5), pp.4359~4368 (2022)
- S. Y. Kim, "A Multiple Linear Kalman Notch Filter-based Mitigation Method for Multiple Vibration Frequencies", Journal of Institute of Control, Robotics and Systems, 24(2), pp.182~188 (2018)
- S. H. Kang, "Analysis of EEG signals for attention training game contents", The Journal of The Institute of Internet, Broadcasting and Communication, 19(3), pp.83~90 (2019)