Statistical sensor fusion of ECG data using automotive-grade sensors

[摘要] Driver states such as fatigue, stress, aggression, distraction or evenmedical emergencies continue to be yield to severe mistakes in driving andpromote accidents. A pathway towards improving driver state assessment can befound in psycho-physiological measures to directly quantify the driver'sstate from physiological recordings. Although heart rate is awell-established physiological variable that reflects cognitive stress,obtaining heart rate contactless and reliably is a challenging task in anautomotive environment. Our aim was to investigate, how sensory fusion of twoautomotive grade sensors would influence the accuracy of automaticclassification of cognitive stress levels. We induced cognitive stress insubjects and estimated levels from their heart rate signals, acquired fromautomotive ready ECG sensors. Using signal quality indices and Kalmanfilters, we were able to decrease Root Mean Squared Error (RMSE) of heartrate recordings by 10 beats per minute. We then trained a neural network toclassify the cognitive workload state of subjects from heart rate andcompared classification performance for ground truth, the individual sensorsand the fused heart rate signal. We obtained an increase of 5 % highercorrect classification by fusing signals as compared to individual sensors,staying only 4 % below the maximally possible classification accuracyfrom ground truth. These results are a first step towards real worldapplications of psycho-physiological measurements in vehicle settings. Futureimplementations of driver state modeling will be able to draw from a largerpool of data sources, such as additional physiological values or vehiclerelated data, which can be expected to drive classification to significantlyhigher values.

[学科分类] 电子、光学、磁材料