Characterizing the Cortical Processes Related to the Perception of Long-Lasting Warm and Cool Sensations

Extracting relevant information from noisy multidimensional signals has tremendous impacts in numerous applications, ranging from audio separation to electrophysiological recording analysis. Linear filters are often considered to reconstruct and interpret the latent sources generating the data. Known properties of the sources can be used to guide their separation. In neuroscience, the cortical processes underlying perception in different modalities (visual, auditory, ...) is often studied using electroencephalography (EEG) during periodic stimulation, eliciting periodic activity in neural sources, some of which being specific to the considered modality. Whereas current approaches extract sources either periodic or discriminative, none of them accounts for both aspects at once. This paper proposes several methods extracting periodic sources specific between two classes, hence termed as Linear Periodic Discriminant Analysis methods. They are validated on synthetic data and EEG recordings of subjects to whom periodic stimulation from two modalities is applied. The methods highlight modality-specific periodic responses.