I want to use Spatial Temporal Discriminant Analysis which is introduced here to classify motor imagery movement from EEG signal.
But to apply this method, as the paper suggests, i have to construct spatial-temporal two-way samples as feature matrices by keeping the spatial features in the first way (rows) while the temporal features in the second way (columns).
I think it is similar to time-frequency matrix. But i do not know how to construct such matrix. (The given dataset is already in spatial-temporal like channel x time. But is there any other merhod?).
Hi Rusho. It sounds like you need to extract some features from the data and enter them into a matrix. If you are using time-frequency features, then you would need to select a certain frequency range. You might start by computing a channels X time X frequency matrix and then average over a frequency range to get a channels X time matrix. I certainly discuss ways to do that in my books and youtube videos.