A real-time weighted-eigenvector MUSIC method for time-frequency analysis of electrogastrogram slow wave

The surface electrogastrogram (EGG) records the electrical slow wave of the stomach noninvasively, whose frequency is a useful clinical indicator of the state of gastric motility. Estimators based on the periodogram method are widely adopted to obtain this parameter. But they are with a poor frequency domain resolution when the data window is short in time-frequency analysis, and have not taken full advantage of the slow wave model. We present a modified multiple signal classification (MUSIC) method for computing the frequency from surface EGG records, developing it into a real-time time-frequency analysis algorithm. Simulations indicate that the modified MUSIC method has better performance in resolution and precision in the sinusoid-like resultant signal frequency detecting than periodogram. Volunteer data tests show that the modified MUSIC method is stable and efficient for clinical applications, and reduces the danger of pseudo peaks for the diagnosis.