Using N-Grams of Quantized EEG Values for Happiness Detection

When applying classification methods for the automatic detection of happiness in human beings using electroencephalographic signals, the major research works in literature report the employment of power spectral density as the main feature. However, the aim of this paper is to explore wheter or not the use of N-grams of quantized EEG values as new features may help to improve the classification process. N-grams is a standard method of data representation in the area of natural language processing which usually reports good results. In this type of input data make sense to employ this kind of representation because the happiness signal is made up of a sequence of values which naturally matches the N-grams paradigm. The results obtained show that this kind of representation obtains better results than others reported in literature.