Research On Vigilance Monitoring Methods For Wearable Scenarios

Vigilance reflects a person’s perception of potential danger in the environment,which gradually decreases with working hours.People often find it difficult to detect and their vigilance falls below the safety threshold range,leading to safety accidents in areas such as transportation,fire rescue,and security.People often find it difficult to detect this process,which leads to vigilance falling below the safety threshold range,leading to safety accidents in areas such as transportation,fire rescue,and security.Therefore,real-time monitoring of personnel’s vigilance and timely warning when vigilance falls below the safety threshold is of great significance and application value.For this reason,this thesis proposes a real-time vigilance monitoring algorithm based on EEG signals,designs EEG signal acquisition hardware and software framework,and finally realizes a low-power wearable vigilance monitoring system.For this reason,this thesis proposes a real-time vigilance monitoring algorithm based on EEG signals,designs EEG signal acquisition hardware and software framework,and finally realizes a low-power wearable vigilance monitoring system.This system can continuously collect EEG signals,analyze personnel’s vigilance in real-time through algorithm models,and timely alert them through sound and vibration when their vigilance is low.The main work of the thesis is as follows:1.In terms of vigilance monitoring algorithms,this thesis proposes a cross-subject vigilance monitoring algorithm based on EEG signals for low computing power platforms,which solves the problem of decreased model accuracy due to individual differences among different personnel.The algorithm synchronously extracts the spatial and frequency band features of EEG,adjusts the feature extractor based on the classification discrepancy of the two classifiers,maps the confusing features of new individuals to the area far away from the decision boundary of the classifier,increases the feature discrimination,and improves the model generalization ability.Meanwhile,lightweight algorithm models are more suitable for deployment on low computing power platforms.The experimental results on three public datasets show that compared to existing algorithms,the algorithm improves accuracy by 2.04%,0.70%,and 1.81%,respectively.2.In terms of signal acquisition hardware,this article designs a low-power wearable EEG signal acquisition hardware platform with signal acquisition,model inference,and human-machine feedback functions.In terms of signal acquisition,an active circuit was designed to solve the problem of high and unstable contact impedance of dry electrodes,while a low noise sampling circuit was designed to improve the quality of signal acquisition.In terms of model inference,an external memory based on Quard SPI was designed,which increased the I/O bandwidth by four times to meet the computational power requirements of alert monitoring model inference.In terms of human-machine feedback,sound and vibration feedback circuits have been designed to achieve early warning prompts when the vigilance is too low.After testing,the input noise of the collection hardware platform designed in this article is less than 2(1₈),the absolute attenuation of average amplitude is less than 10%,and the absolute error of frequency is less than 1.5%,which meets the requirements of vigilance monitoring system.3.In the aspect of vigilance monitoring system,this thesis designs a software framework with signal acquisition control,pre-processing,vigilance analysis and early warning feedback.The software framework collects EEG signals in real time by controlling the hardware platform,and removes the eye electrical interference in EEG signals based on wavelet threshold denoising algorithm;At the same time,integrating the vigilance monitoring algorithm proposed in this thesis,analyzing personnel’s vigilance,and timely warning and prompt through sound and vibration feedback when the vigilance is too low,achieving closed-loop human-machine feedback function.Finally,this thesis implements a low-power wearable vigilance monitoring system based on the EEG signal acquisition hardware platform and vigilance monitoring software.After testing,this system can monitor individual vigilance in real-time,with an accuracy rate of 86.5%.