Search

This paper investigates the problem for the phonocardiographic cardiographic signal processing by employing the third-order cumulant. The non-Gaussian AR model is proposed to model the signals. The cumulant-based third-order recursion is employed to decide the model parameters for understanding the properties of PCG signals in the bifrequency domain. Moreover, to test the performance of the PCG signal with the background noise, some signals with different SNR are analyzed for the bispectral estimation. The results are exhibited and analyzed to show the effectiveness of the methods based on higher-order spectra.

Wavelet transformation is employed to investigate the nonstationarity of clinical EEG signals and other medical signals. In order to detect 4 kinds of EEG rhythms, wavelet packet analysis is used for designing filters with different frequency characteristics. Two kinds of EEG data with different brain function states are analyzed and compared. It is indicated from the experimental results that the dynamic characteristics of clinical brain electrical activities can be demonstrated by using a wavelet packet technique. The method presented in this paper also proposes a new way for the analysis of other biomedical signals.

In many compression techniques the m.s.e (mean square error) criterion is used, a choice normally made because of mathematical convenience. However, for the ECG application this error criterion is unsatisfactory since it does not reveal whether or not an algorithm can preserve diagnostically significant features of the waveforms. There are many defects with it. For this reason the authors present a new error criterion-attribute correlation criterion. Instead of calculating auto- or cross-correlation function from individual sample values, the authors determine correlation between significant attributes of the waveforms. The authors select the shape (up, down, horizontal), amplitude, and position among P-Q-R-S-T points as the attributes to be correlated.