基于脑电与眨眼频率的可穿戴疲劳驾驶检测系统

发布时间：2018-03-01 15:24

本文关键词： 可穿戴疲劳驾驶检测脑电信号眨眼频率分类算法相关系数　出处：《计算机工程》2017年02期 　论文类型：期刊论文

【摘要】：在小型化、低功耗的可穿戴设备上,针对运行基于脑电信号的驾驶疲劳检测系统的准确率不高的问题,在对被试者左前额脑电信号Attention和Meditation以及Blink的数据进行关系分析的基础上,分别筛选最佳窗口宽度和分类算法,设计适用于可穿戴设备的疲劳驾驶检测算法,并在安卓智能设备上进行系统实现。采用准确率、正样本识别正确率、负样本识别正确率、敏感性与特异性指标,分别测试4种分类算法,即k临近算法、决策树算法、朴素贝叶斯算法、多层人工神经网络算法的性能,并最终选择k NN分类算法进行系统实现。实验结果证明,该系统的准确率达到83.7%,敏感性与特异性分别达到73.8%和88.6%,系统具有无线、实时、准确高效的特点。
[Abstract]:On the miniaturized, low-power wearable devices, aiming at the problem that the accuracy of driving fatigue detection system based on EEG is not high, the relationship between Attention, Meditation and Blink data of left frontal EEG is analyzed. Selecting the best window width and classifying algorithm, designing the fatigue driving detection algorithm suitable for wearable device, and implementing it on the Android intelligent device. The accuracy rate, positive sample recognition accuracy rate, negative sample recognition accuracy rate, and negative sample recognition accuracy rate are adopted. The sensitivity and specificity of four classification algorithms, namely, k-neighborhood algorithm, decision tree algorithm, naive Bayesian algorithm and multi-layer artificial neural network algorithm, are tested respectively. The experimental results show that the accuracy of the system reaches 83.7, the sensitivity and specificity of the system reach 73.8% and 88.6respectively. The system has the characteristics of wireless, real-time, accurate and efficient.
【作者单位】：北京工业大学软件学院;
【基金】：国家自然科学基金(61040039,61201361) 北京市自然科学基金(4102005,4122010)
【分类号】：TP18;TP274

【参考文献】