Our article Feature Extraction and Selection for Emotion Recognition from Electrodermal Activity got published in prestigious IEEE Transactions on Affective Computing. Individuals with a wide range of mental health concerns such as children with autism, individuals with intellectual disability etc. have limited ability in recognition and expression of their emotional states. In such cases, analysis of human emotions expressed in the pure unaltered form of physiological signals such as Electrodermal activity (EDA) emerges as the most useful method for monitoring their emotions. However, extracting information about the emotional state from EDA data can be challenging, especially if the processing is to be done online. Hence, it is important to automatically identify meaningful smaller subsets of various EDA features to achieve efficient emotion recognition from EDA signals. In this research, we found that approximately the same numbers of features are required from EDA signals to obtain the optimal accuracy for the arousal recognition and valence recognition. In addition, our research also showed that statistical features related to the Mel-Frequency Cepstral Coefficients (MFCC) give better classification than commonly used Skin Conductance Response (SCR) related features. Our research has opened venues for the future development of new emotion recognition systems based on EDA with higher accuracy and minimizing its computational cost, which is key for the development of emotion detection applications that may work in real time.
The article can be accessed online at https://ieeexplore.ieee.org/abstract/document/8653316
Congratulations to all the authors, Jainendra Shukla (Asst. Professor, IIIT-D, India), Miguel Barreda-Ángeles (Sr. Researcher, EURECAT, Spain), Joan Oliver (Researcher, IRD, Spain), G. C. Nandi (Professor, IIIT-A, India), Domènec Puig (Professor, URV, Spain)