Spatio-Temporal Progressive Attention Model for EEG Classification in Rapid Serial Visual Presentation Task

TL;DR

This work tackles RSVP EEG classification by leveraging spatiotemporal dependencies with a novel STPAM model that progressively attends to informative electrodes and time slices. It assembles three spatial experts and three temporal experts via PSL and PTL, using graph-based spatial convolutions and temporal graph convolutions, augmented with gradient-based attention and KL-divergence penalties to promote diverse, complementary focus. A new infrared-based RSVP EEG dataset, IRED, is introduced to extend RSVP stimulus modalities to low-light conditions. Across two RSVP EEG datasets, STPAM achieves state-of-the-art accuracy, with strong statistical significance, demonstrating robust electrode/time-slice selection and improved generalization to diverse stimuli.

Abstract

As a type of multi-dimensional sequential data, the spatial and temporal dependencies of electroencephalogram (EEG) signals should be further investigated. Thus, in this paper, we propose a novel spatial-temporal progressive attention model (STPAM) to improve EEG classification in rapid serial visual presentation (RSVP) tasks. STPAM first adopts three distinct spatial experts to learn the spatial topological information of brain regions progressively, which is used to minimize the interference of irrelevant brain regions. Concretely, the former expert filters out EEG electrodes in the relative brain regions to be used as prior knowledge for the next expert, ensuring that the subsequent experts gradually focus their attention on information from significant EEG electrodes. This process strengthens the effect of the important brain regions. Then, based on the above-obtained feature sequence with spatial information, three temporal experts are adopted to capture the temporal dependence by progressively assigning attention to the crucial EEG slices. Except for the above EEG classification method, in this paper, we build a novel Infrared RSVP EEG Dataset (IRED) which is based on dim infrared images with small targets for the first time, and conduct extensive experiments on it. The results show that our STPAM can achieve better performance than all the compared methods.

Figures (8)

• Figure 1: Examples of small target images. There are five types of targets in total including airplanes, animals, helicopters, cars, and persons.
• Figure 2: One trial in the RSVP paradigm. The target type of the current trial is shown for 1000ms first and then the fixation is presented for 1000ms to help the subjects to focus their attention on the center of the screen. Subquently each image is displayed for 200ms until the end.
• Figure 3: Framework of the proposed STPAM model. STPAM consists of two major blocks, i.e., progressive spatial and temporal learning blocks, which can pay attention to the important electrodes and time slices from spatial and temporal dimensions adaptively.
• Figure 4: Structure of the spatial expert.
• Figure 5: Structure of the temporal expert.