Comparison of various time/frequency distributions (classical and signal-dependent) applied to synthetic uterine EMG signals

Synthetic signals simulating uterine EHG are created in order to compare and validate several time/frequency distributions (TFD): we tested autoregressive spectra, short time Fourier transform, smooth pseudo Wigner Ville, exponential and cone shaped kernel distributions, as well as a signal-dependent method. Specific criteria were especially designed in order to quantify either the overall quality of the TFDs or their capability of tracking efficiently the evolution of the main spectral peak. Both the autoregressive and the signal-dependent methods yielded good results. We then tested these two transformations in terms of robustness and selectivity.<
>