Time-series classification (TSC) has been widely utilized across various domains, including brain-computer interfaces (BCI) for emotion recognition through electroencephalogram (EEG) signals. However, traditional methods often struggle to capture the complex emotional patterns present in EEG data. Recent advancements in encoding techniques have provided promising avenues for improving emotion recognition. This study introduces asymmetric windowing recurrence plots (AWRP) as a novel encoding technique to efficiently encapsulate the dynamic characteristics of EEG signals into texture-rich image representations. This study systematically compares the impact of conventional thresholded and unthresholded recurrence plots (RP) versus the proposed AWRP in emotion recognition tasks. Empirical validations conducted across benchmark datasets, such as DEAP and SEED, demonstrate that the AWRP method achieves classification accuracies of 99.84% and 99.69%, respectively, outperforming existing state-of-the-art methodologies. This study emphasizes the significance of input formulation, highlighting that richer input textures, as provided by AWRP, significantly enhance emotion recognition performance while ensuring computational memory usage efficiency. These findings have significant implications in the domain of EEG-based emotion recognition and offer a novel perspective that can guide future research.