A k-nearest-neighbor classifier with heart rate variability feature-based transformation algorithm for driving stress recognition

This paper presents a k-nearest-neighbor classifier with HRV feature-based transformation algorithm for driving stress recognition. The proposed feature-based transformation algorithm consists of feature generation, feature selection, and feature dimension reduction. In order to generate significant features from ECG signals, two feature generation approaches: trend-based and parameter-based methods are proposed in this study. The trend-based method computes statistical features from long-term HRV variations, while the parameter-based method calculates features from five-minute HRV analysis. The kernel-based class separability (KBCS) is employed as the selection criterion for feature selection. To reduce computational load of the algorithm, principal component analysis (PCA) and linear discriminant analysis (LDA) are adopted for feature dimension reduction. Our experimental results show that the combination of KBCS, LDA, and PCA can achieve satisfactory recognition rates for the features generated by both trend-based and parameter-based methods. The main contribution of this study is that our proposed approach can use only ECG signals to effectively recognize driving stress conditions with very good recognition performance.