Brain Activity of Visual Working Memory Predicted from Image Target Numbers — An EEG Study

The Electroencephologram (EEG) experiments were designed in this paper to emulate our daily visual working memory task and demonstrate the effect of image target numbers and background textures on our brains’ visual working memory. This paper discovered that there is a possibly of implementing a neural network to predict memoryrelated brain activity in a visual working memory experiment, where participants were presented with images of different target item numbers and asked to remember as many target objects as possible. Both the Multi-Layer Perceptron (MLP) network and support vector machine (SVM) were used as training methods for the prediction. The prediction results are consistent with the actual EEG power variation observed in the experiment, which demonstrate the effect of target item number and background texture on the level of difficulty in image memorization.