The difference in statistical characteristics between physiological signals and noise creates the possibility of designing a selective circuit to separate signals from noise. Analysis shows that by an extremely low signal-to-noise ratio before the pulse conditioner in a biotelemetry system, the mixture of normal and spurious pulses are present at the demodulator input. The system of noise destruction analyzes the sequential distances between two following pulses and blanks all spurious pulses using the criteria of high autocorrelation of physiological signals in short intervals. Evaluation of the circuit demonstrates a dramatic increase in accuracy in the presence of strong interference because the system of noise destruction performs a linear extrapolation of free-of-noise parts of pulse flow into destroyed spaces. The equations to calculate selective thresholds and output errors are delineated. The noise destructor is especially useful in long-term monitoring and automatic data analysis because it prevents a formation of sharp noisy spikes at the biotelemetry system's output.