In Affective Computing, where research continually advances rapidly, affect analysis is crucial in influencing human behavior, decision-making, and task performance. This study proposes an architecture for emotion analysis using a Deep Convolutional Generative Adversarial Network based on Inception-V3, a well-known Deep Convolutional Neural Network. We have used Sca-lograms, a time-frequency representation of Electroencephalogram (EEG) signals generated from the DEAP database, a benchmark database for spontaneous emotion recognition. EEG signals are preferred for Affect analysis due to their non-invasive nature, cost-effectiveness, and high temporal resolution. While most researchers considered four classes for experimentation, we analyzed six classes of emotions. The results obtained with six classes of emotions are even better compared to the four classes of emotions reported in the existing literature. Hence, the proposed architecture is performing well with many classes.