Analysis of emotional processes in EEG signals by multidimensional clustering decomposition using wavelet transform

This paper presents a signal analysis of electroencephalogram (EEG), focused on characterizing emotional stimulus from audio -visual evocations records. Wavelet analysis was used as features extraction methodology, implementing discrete extraction methodology and analysis of signal characteristics by frequency bands (rhythms) delta, theta, alpha and beta in order to reduce computational burden of the classification space. Joy, anger and sadness, were taken as main references from the Ekman model, representation on the arousal and valence space(AVS) representation, providing this way, spaces of reference with equivalently distributed emotions, which in turn provides a more adequate to interpret and calculate distances between emotions, due that many of the previous models works with well defined emotional states (i.e. with clear ranges where emotion becomes other); This technique allows to implement techniques for calculating statistical metrics on a more simplest way, which in turn provide a framework to allow detection that could serve as a basis for recognition tasks. In general this work generates a methodology of analysis that would provide a way to establish sets of features to be analyzed and the possibility of generating more compact sets for classification tasks in the emotion detection.