Wearable devices often struggle to capture the full range of emotions accurately through peripheral physiological signals like Electrodermal Activity (EDA), Photoplethysmogram (PPG), and Electromyogram (EMG), leading to reduced classification accuracy. While Electroencephalogram (EEG) signals are recognized for their superior emotion detection capabilities, past EEG-based studies have been confined to lab settings with high-end systems using 32-64 channels. There is a growing need for systems that can continuously monitor emotions outside the lab. Modern wearable EEG devices, with configurations from single to eight channels, offer a solution. Our study investigates if integrating EEG data from only two channels in commercial wearable EEG devices with peripheral signals can enhance emotion detection accuracy. Using the publicly available DEAP dataset, which provides EEG and peripheral signals from 32 participants exposed to emotional stimuli, we explore various preprocessing and feature extraction techniques for multimodal emotion detection. The DEAP dataset categorizes emotions based on valence and arousal ratings, converting them into binary class problems: High Valence vs. Low Valence and High Arousal vs. Low Arousal. Intra-class emotion classification is performed using a Random Forest classifier with $\mathbf{5}$-fold cross-validation. Results show that while single-modality peripheral signals achieve classification accuracies of 76-85%, the addition of EEG data significantly boosts accuracy up to $\mathbf{9 5 \%}$. Combining EEG data with peripheral signals also greatly improves individual subject valence and arousal classification accuracy and F1 scores.