A new approach to time dependent AR modeling of multi-frame signals and its application to diagnosis of myocardial infarction

A method for estimating spectrum transition between short-length multiframe signals in low SNR (signal-to-noise ratio) cases is presented. If the transition pattern is complex and/or there are large differences in the transition patterns among the individual sets of multiframe signals, it is difficult to estimate the transition pattern stably by the time-varying AR modeling because the results are considerably dependent on the choice of the basic functions to be used. The proposed approach uses a linear algorithm without any basic functions. Instead, the spectrum transition constraint is used, and the singular-value-decomposition-based technique is applied to obtain more accurate estimates. When this method is applied to the analysis of multiframe signals of the fourth heart sounds significant differences in the transition patterns are clearly detected in the spectra between patients with myocardial infarction and normal persons. The characteristics of these transition patterns may be applied to acoustic diagnosis of heart diseases