The possibility of increasing the signal-to-noise ratio by the method of echo-signal spectrum splitting jointly with spectrum extrapolation is considered. The essence of the method proposed is that the echo-signal spectrum known within a given frequency range (f min, f max) is divided into several subranges (f min i , f max i ). The construction of the AR model of the echo-signal spectrum allows the spectrum extrapolation from each subrange by an interval (f min e , f max e that far exceeds the initial interval. This means that, with the use of one echo signal, a set of signals can be obtained, the adding together of which increases the signal-to-noise ratio and ensures an ultrafine beam resolution for flaw images. The presented results of processing echo signals, which were obtained in numerical and model experiments, confirm the efficiency of the proposed technique for processing echo signals.