Abstract
The wide variety of waveform in EEG signals and the high non-stationary nature of many of them is one of the main difficulties to develop automatic detection system for them. In sleep stage classification a relevant transient wave is the K-complex. This paper comprehend the developing of two algorithms in order to achieve an automatic K-complex detection from EEG raw data. These algorithms are based on a time-frequency analysis and two time-frequency techniques, the Short Time Fourier Transform (STFT) and the Continuous Wavelet Transform (CWT), are tested in order to find out which one is the best for our purpose, being of two wavelet functions to measure the capability of them to detect K-complex and to choose one to be employed in the algorithms. The first algorithm is based on the energy distribution of the CWT detecting the spectral component of the K-complex. The second algorithm is focused on the morphology of the K-complex / sleep spindle waveform after the CWT. Evaluating the algorithms results reveals that a false K-complex detection is as important as real K-complex detection.