스티어링 휠로부터 측정된 생체신호 데이터에 기계 학습을 적용한 운전자 피로도 모니터링 시스템 개발

Abstract

Driver fatigue is one of the major causes of traffic accidents. Therefore, in order to prevent traffic accidents and ensure traffic safety, it is necessary to analyze the fatigue characteristics of the driver. In order to monitor driving fatigue, physiological signals such as heart rate or skin conductivity can be used. Signal changes measured while driving were clinically proven for quantitative analysis of stress and mental fatigue. In this study, we developed a real-time driving fatigue identification and classification system using physiological signals from the hand on the steering wheel for the minimum restraint on the driver. Experimental procedures were designed and constructed to induce driving conditions similar to those in real life. Driving simulation program is used to measure fatigue, and physiological signals such as skin conductance (GSR), electrocardiogram (ECG) and photoplethysmography (PPG) are recorded during the experiment. A total of 18 subjects participated in this experiment. In addition, thirteen features were extracted from the statistical and time-frequency changes of the physiological signals in relation to fatigue and heart rate variability induced in long-term driving. In the signal analysis, the proposed classification algorithm is designed based on the Time Delay Neural Network (TDNN) using the extracted features. The proposed system can classify fatigue into three levels. The system also identifies persistent fatigue situations or instantaneous stress depending on the activity of the subject. The experimental results show that the accuracy is 77.27% in Support Vector Machine (SVM). The proposed system can detect and classify fatigue stages based on two non-invasive signals. These signals can be preconfigured and collected during simulation and processed in real time. These driving safety systems for drivers are expected to reduce traffic accidents by providing drivers with fast and reliable real-time notifications. For example, our system can be applied to fatigue monitoring in air traffic.