Spectral analysis of bioelectric signals by adapted wavelet transforms

In this study we use the adapted wavelet transform methods (wavelet and cosine packets) for spectral analysis of bioelectric signals. These methods have recently been introduced for analysis of non-stationary signals. Using recordings of the heart rate variability in twenty healthy subjects, the estimated power in different frequency bands is compared to results based on the classical methods: fast Fourier transform and autoregressive modelling. The results showed that cosine packets gave similar results to classical methods, and may be preferred to characterise the rhythmic components in the recorded signals. On the other hand, the non-stationary fluctuations, i.e., the "trend", was efficiently decomposed using the wavelet transform method.