Principles, System Identification and Applications of Volterra Nonlinear Filter

• Main Authors: Jiunn-Jong Pan, 潘俊忠
• Other Authors: Soo-Chang Pei
• Format: Others
• Language: en_US
• Published: 1996
• Online Access: http://ndltd.ncl.edu.tw/handle/01245293592972697614

碩士 === 國立臺灣大學 === 電機工程研究所 === 84 === In the chapter 2, we have introduced the direct form method bas- ed on LMS algorithm to achieve the system identification of one- dimensional Volterra filter. We also use the orthogonal approach to perform the system identification of Volterra filter. Last, the approximation form is also proved to be able to be applied to approximate the second-order Volterra filter coefficients. The one-dimensional algorithms then has been extended to two- dimensional one in the chapter 3, and the concept of imposing isotropic constraints to reduce the number of coefficients has also been introduced. In our simulation of this chapter, that imposing the isotropic constraints not only lowers the computat- ional load but also increases the convergence speed is found. In the chapter 4, we have tried to present some applications of Volterra filter.Teager's algorithm can be used to demodulate the digital modulation signals and amplitude modulation signals without losing any imformation. Three digital demodulation types (ASK,PSK, PSK) are proved be able to be demodulated successfully. We also provide two example of base band signals (the music signal and the human voice signal) which have been modulated to AM signal for simulation, and they have been demodulated by the Teager's algorithm without using the local oscillator.Since the interpolation is used to upsampled the signals, some errors are generated in the meanwhile.So it should be noted that the value of SNR of our method would be a little larger than that shown in the figures. Last in the chapter 4, since the orthogonal approach is robust against the input statistical character, it can be used to design a Volterra filter with only the knowledge of the input and desired output signals. The Volterra filter is designed to remove the Gaussian noise of images and a satisfactory result is obtained in our simulation.