Cyclostationary Signal Processing and Its Blind Equalization Ap- plications

• Main Authors: Meng-Fu Shih, 施孟甫
• Other Authors: Soo-Chang Pei
• Format: Others
• Language: en_US
• Published: 1996
• Online Access: http://ndltd.ncl.edu.tw/handle/44006928176007747763

碩士 === 國立臺灣大學 === 電機工程研究所 === 84 === For most manmade signals encountered in communication, tele- metry, radar, and sonar systems, cyclostationary attributes are exhibited. These attributes are reflected in periodicities of moments of the signals. There are spectral correlations in cyclo- stationary signals. The inherent spectral redundancy can be ex- ploited to improve the performance of several signal processing tasks. For cyclostationary signals, the optimum polyperiodic linear filter is called cyclic Wiener filter, which is a generalization of the traditional Wiener filter. Actually, it is the optimum filter for cyclostationary signals in the sense that minimum mean square error is achieved. This optimum filtering uses not only time translates but frequency translates of the signals and can achieve lower mean-square- error. Baseband fractionally spaced sampled communication signals are cyclostationary signals. The spectral redundancy of the signal can be used to carry out blind equalization as well. Two algo- rithms are introduced in this thesis: One is what using the com- plex cepstrum of the cyclic autocorrelation, which is actually a extension of the well known tricepstrum based method. The other one is what using polyperiodic linear filtering, and we can esti- mate the desired response and update the filter coefficient by many well-developed Bussgang equalization methods. Both methods outperform corresponding traditional ones. The simulation results and performance comparisons are all depicted.