Non-parametric linear time-invariant system identification by discrete wavelet transforms

摘要: We describe the use of discrete wavelet transform (DWT) for non-parametric linear time-invariant system identification. Identification is achieved by using a test excitation to under (SUT) that also acts as analyzing function DWT SUT's output, so recover impulse response. The method uses any signal gives an orthogonal inner product in at some step size (that cannot be 1). favor scaling coefficients excitations, with 2. However, or frequency response can then only estimated half available number points sampled output sequence, introducing multirate problem means we have 'oversample' SUT output. has several advantages over existing techniques, e.g., it simple, easy generate excitation, and avoids singularity problems (unbounded) accumulation round-off errors occur standard techniques. In extensive simulations, identification variety finite infinite systems shown considerably better than conventional methods.