Research on Classification Algorithm of Reduced Support Vector Machine for Driving Fatigue Detection

In order to improve the accuracy and real-time performance of driving fatigue detection, a driving fatigue detection method is proposed based on reduced support vector machine algorithm. The ECG indicators are treated as human fatigue characterization indicators in the method. The human ECG can be obtained through experiment in different state, and PERCLOS value in selecting and controlling environmental conditions is taken as a basis for human fatigue state judging. Thus, the ECG data from experiment can be divided into two categories of the normal and fatigue statue. Then integrating the ECG linear and nonlinear indicators establishes Identifiable eigenvector space, and the computational complexity is reduced by Reduced Kernel clustering method to improve robustness of the algorithm. The results of detection experiment on driving fatigue detection show the effectiveness of this method.