Emotion recognition based on electroencephalography (EEG) is an increasingly important tool for diagnosing emotional disorders in neurology and psychiatry, despite the challenges posed by EEG's low signal-to-noise ratio and spatial resolution due to volumetric conduction. Multiple kernel learning (MKL) has shown great potential for managing such complex and non-linear data. Inspired by this, we propose a Support Vector Machine (SVM) classifier using the EasyMKL algorithm combined with Multitaper Spectral Estimation to address the limitations of EEG-based emotion recognition. In our approach, we extract power spectral density (PSD) and differential entropy (DE) features from four rhythmic EEG signals. The proposed method was validated using random division and evaluated on the DEAP benchmarking dataset for two-category classification along the arousal-valence dimension. Five-fold cross-validation results show that the method improves emotion recognition accuracy by approximately 3.8% compared to single-kernel SVM, achieving high mean accuracies of 91.71% for valence and 91.59% for arousal classification.