Classification of mental tasks using different spectral estimation methods

Abstract

In this chapter, parametric (Burg) and non parametric (standard and Welch) spectral methods were utilized in order to estimate the spectral content of EEG signals for different mental tasks. Two parameters were utilized to analyze the behaviour of every spectral estimation methods: the Pm and the RMS of different frequency bands. These methods were tested in two different databases. We found that the use of the RMS allows higher classification accuracies with any spectral estimation technique. The Welch periodogram and Burg method are preferable in front of the standard periodogram. The use of Welch or Burg methods seems to be indistinct due to they performed similar, although in some subjects performed better one than other.