The sleep spindles are special patterns that appear at different stages of sleep.We analyze them in a time-frequency (TF) plane which is a general representation to treat stationary and non-stationary signals.The time-frequency representations,based on the Fourier transform (Spectrogram,Wigner-ville) can not be interpreted correctly because of a lack of samples and frequency interferences.So,we prefer to use a recent 2D time-frequency AR representation,called TFAR,which has a better frequency resolution without any interferences.The TFAR method is based on a windowed data matrix around a entral time,and that matrix is then transformed into a complex one.Afterwards,we apply to it a 2D-AR method in order to produce high-resolution spectral analysis without cross term artifacts.We compare the Short-Time Fourier Transform (STFT) and the TFAR method applied to sleep spindles to show the relevant features of the TFAR method.
展开▼
