Improving spectral parameters estimation for non stationary biological signals by means of a real time adaptive system

Abstract

A novel spectrum analyzer is presented which implements a real-time fast Fourier transform up to a frequency of 40 kHz and uses the adaptive algorithm to estimate maximum frequency with a high immunity from noise. Starting with this adaptive filtering, the other spectral parameter can be easily estimated. The system has been evaluated both with simulated and real signals, showing performances uniformly better than those of commercially available systems. Some applications to Doppler and myoelectric signals are reported and discussed. Preliminary results show that this system can be profitably used when nonstationary signals are to be studied.<>