Detection of certain heart arrhythmia by digital spectral analysis of ECGs

Certain dangerous heart arrhythmia manifest themselves as late potentials in ECGs. Their detection from non-invasively collected, averaged ECG signals is highly desirable. Present measurements define the part of the time axis to be analyzed, apply a hopping-window DFT, search for the window with the most significant increase of higher-frequency components, and determine, as a diagnostic feature, the quotient of the energies in the upper and the lower part of the spectrum. In a detailed study, the parameters of such a procedure are examined and optimized. Then, alternative methods-based both on DFT and parametric spectrum descriptions-are investigated. A novel approach is shown to improve the diagnostic certainty conceivably