Fault detection based on spectral estimation of nonuniformly sampled signals

Two methods for signal discrete sample value fault detection are considered. While both of them are based on the sample value comparison with their predicted values, the sample value prediction and the fault detection, in the first case, is performed on the basis of signal short-time analysis-synthesis and, in the second case, a special filter is used for that. The application specifics of both methods are compared and illustrated on an example of frequency modulated digital signal fault detection. This work was focused on achieving two functional advantages. First, the predicted signal values should be obtained by processing data sequences that are considerably shorter than those usually needed at applications of the traditional spectral estimation based prediction methods. Second, the fault detection methods should be equally well applicable both for processing periodically and nonuniformly sampled signals. Therefore it should be possible to use them, under certain conditions, for alias-flee signal processing at frequencies exceeding the Nyquist limit