Human affection is associated with neurophysiological changes, variously coupled with thoughts, feelings, and behavioral responses of a human being. Psychologists and cognitive scientists are trying to recognize and discriminate different emotional states to unleash true potential of emotional intelligence. Recently, emotion recognition systems gained much attention for evaluating various neurological disorders, for creating better interfaces to human–machine interaction; to identify behavior and personality traits in a social setting. The advancements in the realm of brain–computer interface (BCI) and machine learning models yield better performance in affective computing. This paper proposes an algorithmic framework for affect recognition in three strategies, viz., valence-arousal classification, emotion intensity recognition, and rhythm wise classification using a Deep Convolutional Neural Network (DCNN). The proposed novel algorithm is implemented on DEAP benchmark dataset, which is a multimodal dataset designed for analyzing human affective states, featuring 32-channel EEG data along with various other physiological signals. The model achieved a highest precision of 95.66% for valence-arousal classification scheme. The emotion intensity recognition scheme yields a highest precision of 96.82% for dominance class; rhythm wise classification scheme achieved a highest precision of 95.86% in high-gamma band, demonstrating the efficacy of the proposed model.