Abstarct: Emotion recognition and analysis, given its profound impact on physiological and psychological dimensions, represents a key focus in human-computer interaction. This study seeks to improve the performance of emotion recognition systems by leveraging electroencephalography (EEG) and photoplethysmography (PPG) signals. The data were sourced from the DEAP dataset, comprising 32 EEG channels and 8 physiological channels. From these, 5 EEG channels and 1 PPG channel were selected to evaluate their impact on emotion recognition. In this research, deep neural networks, including Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) networks, ResNet18, and VGG19, were utilized for feature extraction and emotion classification. To enhance the accuracy of emotion recognition, advanced preprocessing techniques such as Fourier Transform, Short-Time Fourier Transform (STFT), Discrete Wavelet Transform (DWT), Continuous Wavelet Transform (CWT), and Stockwell Transform were assessed for their effects on neural network performance. The findings indicate that these preprocessing methods, especially those involving signal domain transformation, significantly boost the accuracy and efficiency of emotion recognition systems. Statistical analyses specifically reveal that combining EEG and PPG signals with the ResNet18 network, while applying STFT preprocessing, achieves an emotion recognition accuracy of 99%. This outcome highlights a substantial enhancement in accuracy and performance compared to conventional methods, underscoring the high potential of combining EEG and PPG signals to advance the effectiveness of such systems. Overall, this study demonstrates the considerable promise of physiological signal-baxsed approaches for improving emotion recognition systems. The application of signal domain transformation techniques, in particular, has shown a pronounced impact on accuracy and efficiency, marking a significant step toward the development of intelligent systems responsive to human emotions.