Abstract
We describe here an approach for identifying the transfer function for the continuous-time model of a system. The correlation method in this paper is as follows; the cross-correlation functions (CCFs) between the M-sequence Gaussian (shape) pulse (MG) signal and the n-th differenciated value (nD) of the M-sequence input signal, and the CCFs between the MG signal and the nD of the system output are obtained by the CCFs between the nD of the MG (nD-MG) signal and the M-sequence input signal, and the CCFs between the nD-MG signal and the system output respectively. By this method; (1) we can get an un-biased estimation in a white observation noise system. (2) The method can be applied off-line or on-line (non-recursively or recursively). (3) In comparison to the two-stage least-squares method and the instrumental variables method about the correction of the bias caused by the noise, the described method filters off the noise in the first place using the MG signals; then the estimation is obtained by the least-square method. In addition, it is shown by the experimental results that even if the transfer function of the model is different from that of the system, it is possible to obtain the system parameters in a reasonable form.
Our experimental results of digital simulation demonstrate that the proposed method can identify the system fairly well.
