Time-varying singular value decomposition analysis of electrodermal activity: A novel method of cognitive load estimation

Abstract

The development of reliable and cost effective cognitive load estimation is an important task in the clinical research as it finds widespread application in the fields of brain computer interface. For this purpose, a novel simple but discriminative algorithm was proposed using minimum number of physiological signals and time-varying singular value decomposition (TSVD) approach. The proposed feature could deal with time-varying characteristics and was also able to extract algebraic information of the electrodermal activity (EDA) signals. They were recorded while participants were performing different levels of an arithmetic task.The experimental results on the data set of 35 participants showed high average accuracy rate of 98.3%. The method had much better performance compared to previous techniques, such as Fourier, cepstrum, and wavelet transforms, and conventional tonic measures. It provided a computationally efficient and robust characterization of the signals in the presence of individual differences and noises.