Revisiting Electrode Importance: Quantity Outweighs Location in 3D-CNN Emotion Classification

Previous studies have often emphasized the special contribution of specific brain regions to emotion recognition. However, using the DEAP dataset and a three-dimensional convolutional neural network (3D-CNN), our study found that the number of electrodes is more critical for emotion recognition performance, while the specific electrode locations have minimal impact. We designed three electrode configuration schemes: (A) Full Electrode Scheme (all 32 electrodes); (B) Hemispheric-Region Scheme (13 or 14 electrodes per sub-scheme); and (C) Four-Region Division Scheme (6 electrodes per sub-scheme). The model achieved four-class classification accuracies of 93.53%, 91.26%, and 88.49% for schemes A, B, and C, respectively. The results show that recognition accuracy decreases progressively as the number of electrodes is reduced. However, under the same electrode count, even with significant spatial distribution differences, accuracy variations among sub-schemes remain below 1%, indicating that electrode location has a negligible effect on performance. These findings suggest that, given a sufficient number of electrodes, the layout can be optimized for user comfort, offering theoretical guidance for designing lightweight wearable emotion recognition devices.