An ensemble classifier for emotion classification from EEG and GSR signals for autism detection

Determining individuals' psycho physiological states is essential in multiple fields, especially for integrating disabled individuals into social environments. Among these pursuits, the domain of autism detection emerges as an important area of concern. In this context, the importance of leveraging distinct physiological signals for emotion recognition becomes evident. Traditional methods relying largely on posture and facial expressions have shown limited success for achieving the desired accuracy in emotion recognition. To address these limitations, recent research has turned towards exploring a diverse array of physiological signals. Electroencephalogram (EEG) and Galvanic Skin Response (GSR) signals have come to the forefront due to their ability to offer more nuanced insights into the emotional states of individuals. However, very few works have been reported on the fusion of multimodal signals in this aspect. Therefore, this paper proposes an Ensemble Classifier for Emotion Classification from EEG and GSR Signals for Autism Detection (ECE-AD). This model called the Emotion Classification Ensemble for Autism Detection (ECE-AD) is intended to capture longer range dependencies and contextual information from data sequences for accurate emotion classification. It is built as a multiclass classifier and evaluated on a public dataset achieving 99.97 % detection accuracy, 99.31 % sensitivity, 99.99 % specificity and 0.9870 % precision. The ECE-AD showcases robustness suitable for clinical integration in emotion detection and enhancing the quality of care within healthcare settings.