Identification of physiological systems using pseudo-inverse based deconvolution

Abstract

The identification of nonparametric impulse response functions (IRFs) from noisy, finite-length data records is analyzed using the techniques of matrix perturbation analysis. Based on these findings, we develop a new method for IRF estimation which is expected to be more robust than existing techniques, particularly when the input is non-white. An application to the identification of human ankle dynamics is presented which demonstrates the superiority of this new method over classical techniques.