A structure based on the recurrent wavelet neural networks(RWNNs) trained with unscented Kalman filter (UKF) algorithm is proposed for the time-varying fading channel equalization in wireless communication system. Compared with traditional neural networks based equalization, the main features of the proposed recurrent wavelet neural networks equalization algorithm are fast convergence and good performance using relatively short training symbols, provided with better performance of equalization. The simulation results for various time-varying channels are presented to show that the proposed equalization algorithm is fit for Wavelet packet transform-based multicarrier modulation communication system.

Date Published: 8 September 2011
PDF: 10 pages
Proc. SPIE 8193, International Symposium on Photoelectronic Detection and Imaging 2011: Advances in Infrared Imaging and Applications, 81930E (8 September 2011); doi: 10.1117/12.897348

Published in SPIE Proceedings Vol. 8193:
International Symposium on Photoelectronic Detection and Imaging 2011: Advances in Infrared Imaging and Applications
Jeffery J. Puschell; Junhao Chu; Haimei Gong; Jin Lu, Editor(s)