Autonomous emotion development using incremental modified adaptive neuro-fuzzy inference system

In this paper, we propose an autonomous emotion development system with incremental learning for interacting with human subjects, and autonomously understanding the emotional status of humans. For the understanding of human emotion the proposed system needs human-like visual senses perceiving natural scenes as stimuli. According to the relationship between emotional factors and characteristics of an image, we incorporate the fuzzy concept to extract emotional features using L*C*H* color and orientation information. Additionally, it can sense inputs that have no analog in human senses-reading brain signals in human subjects. We consider the electroencephalography (EEG) signals which are stimulated by natural stimuli to form the semantic emotional features as well. We develop a novel adaptive neuro-fuzzy inference system (ANFIS) based on an incremental learning algorithm to autonomously develop the capability of understanding complex emotions. The proposed incremental modified ANFIS needs only the newly arrived data to adjust the shape of Gaussian membership functions with full covariance matrix, to generate new membership functions or new rules for labeling emotion according to the characteristics of the new data. Utilizing the developmental process, the proposed system can autonomously develop the mental ability to understand more complex human emotions by mining the characteristics of emotional features and interacting with human subjects.