| Title |
Important Feature Analysis of HRV and PRV for Cognitive Load Detection |
| Authors |
김민아(Kim Min-A) ; 홍상기(Hong Sang Gi) ; 이강복(Lee Kangbok) ; 김정균(Kim Jeong-Kyun) |
| DOI |
https://doi.org/10.5573/ieie.2023.60.5.70 |
| Keywords |
Cognitive load; SHAP; Features; HRV; PRV |
| Abstract |
Analyzing and understanding cognitive load is an important skill for improving HCI (Human-Computer Interaction) skills and preventing accidents. Recently, heart rate variability analysis has been widely used for cognitive load analysis, and as the use of wearable devices becomes more common, the pulse rate variability of photoplethysmogram has been used as a substitute for heart rate variability. However, pulse rate variability and heart rate variability are not compatible in all cases, and analyzing them with the same algorithm increases the accuracy error. Therefore, the goal of this study is to reduce the error between the two waveforms and increase the overall accuracy by deriving the important parameters of each waveform. A total of 37 features were extracted from three domains, frequency domain, time domain, and nonlinear domain, and XAI (explainable artificial interllingence) was used. As a result, an accuracy of up to 73.8% was obtained, and the error was reduced to less than 1% at maximum. In addition, as a result of analyzing important features for each waveform by conducting a t-test, it was found that SD1/SD2 are important parameters for heart rate variability analysis, and NN20, pNN20 are important parameters for pulse rate variability analysis. |