TCN-BiLSTM 앙상블 모델 기반 심전도를 이용한 운전자 식별 시스템
Driver identification system using ECG based on TCN-BiLSTM ensemble model
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
최근 스마트 모빌리티 시스템이 운전자 중심 서비스를 위해 위·변조에 강인한 신체 내부의 전기적 신호인 생체신호를 이용하여 운전자 인식 연구가 활발히 진행 중이다. 기존 딥러닝 네트워크로써 CNN은 제한적인 지역적 특징 추출과 LSTM은 반주기적 생체신호의 형태학적 특징 분석 한계로 성능을 개선하는데 한계가 있다. 본 논문은 시간변화에 따른 신호 패턴을 학습하는 TCN(Temporal Convolutional Network)과 장기 의존성을 학습하는 BiLSTM(Bidirectional Long Short Term Memory)을 이용한 운전자 식별 시스템을 제안한다. 제안한 식별 시스템은 신호의 잡음을 제거하는 전처리 과정, 기준점 및 비기준점 기반 특징 분할 과정, TCN과 BiLSTM 앙상블 모델에 의해 운전자를 식별하는 과정으로 구성된다. 실험 결과, PhysioNet 공개 DB인 drivedb와 normal sinus rhythm database를 이용하여 운전자 식별 정확도는 각각 99.76%, 99.97%로 기존 단일 네트워크보다 4.64%, 3.84% 더 우수함을 확인하였다.
Abstract
Recently, driver recognition research is actively underway for smart mobility systems to provide driver-centered services using biosignals, electrical signals inside the body that are resistant to forgery and falsification. As an existing deep learning network, CNN has limited local feature extraction and LSTM has limitations in improving performance due to limitations in analyzing morphological features of semi-periodic biological signals. This paper proposes a driver identification system using a temporal convolutional network (TCN) that learns signal patterns according to time changes and a Bidirectional Long Short Term Memory(BiLSTM) that learns long-term dependencies. The proposed identification system consists of a preprocessing process to remove noise from the signal, a feature segmentation process based on fiducial points and non-fiducial points, and a process to identify the driver using the TCN and BiLSTM ensemble models. Experimental results using the PhysioNet public databases, drivedb and normal sinus rhythm database, showed that the driver identification accuracy was 99.76% and 99.97% respectively, which is 4.64% and 3.84% better than the existing single network models.
참고문헌 (12)
- J. Yang, "Driving assistance system based on data fusion of multisource sensors for autonomous unmanned ground vehicles", Computer Networks, 192(108053) (2021)
- G. H. Choi, "Multi-modal biometrics based implicit driver identification system using multi-TF images of ECG and EMG", Computers in Biology and Medicine, 159(106851) (2023)
- G. Du, "A multimodal fusion fatigue driving detection method based on heart rate and PERCLOS", IEEE Trans. Intell. Transport. Syst., 23, pp.21810~21820 (2022)
- X. Jin, "Prediction for time series with CNN and LSTM", Proceedings of the 11th international conference on modelling (ICMIC2019), 582, pp.631~641 (2020)
- S. S. Abdeldayem, "A novel approach for ECG-based human identification using spectral correlation and deep learning", IEEE Transactions on Biometrics, Behavior, and Identity Science, pp.1~14 (2019)
- M. G. Kim, "Deep learning based on 1-D ensemble networks using ECG for realtime user recognition", IEEE Transactions on Industrial Informatics, pp.5656~5663 (2019)
- S. Bianc, "Biometric recognition using multimodal physiological signals", IEEE Access, pp.83581~83588 (2019)
- J. Y. Pyun, "ECG identification for personal authentication using LSTM-based deep recurrent neural networks", Sensors, 11 (2020)
- S. Dandapat, "An ECG biometric system using hierarchical LSTM with attention mechanism", IEEE Sensors Journal, 22, pp.6052~6061 (2022)
- S. C. Wu, "A scalable open-set ECG identification system based on compressed CNNs", IEEE Transactions on Neural Networks and Learning Systems, pp.4966~4980 (2021)
- D. Jyotishi, "An LSTM-based model for person identification using ECG signal", IEEE Sensors Letters, 4, pp.1~4 (2020)
- J. Pan, "A real-time qrs detection algorithm", IEEE Trans. Biomed. Eng, 32(3), pp.230~236 (1985)