Estimation of signal-to-noise ratio of a quasiperiodic cardiovascular signal using coherency and correlation techniques

Two algorithms, based on coherency and correlation functions, are proposed for estimating the signal-to-noise ratio of physiological signals including the effect of flicker noise. New sampling techniques based on converting a single continuous signal into two time series that satisfy the requirements of the algorithms are proposed. The algorithms are applied to a computer-generated signal and noise composite of known SNR, as well as to quasiperiodic left ventricular pressure (LVP) waveforms transduced from ten patients. Both coherency and correlation methods were used to estimate the known SNR of the computer-generated compsite, signal and noise to within ±5 per cent, for SNR values between 10 and 30 dB. When applied to LVP waveforms, the two methods gave SNR values which were almost identical. Flicker noise associated with the LVP waveforms was computed to be 14 dB higher than the white noise. It is concluded that flicker, or 1/f noise, which has been heretofore ignored in SNR calculations of pressure and flow signals, must now be included.