Nonlinear Signal Processing for Antenna Array and DSP Implementation

• Main Authors: Zhi-Wei Shi, 施志偉
• Other Authors: Ying Li
• Format: Others
• Language: zh-TW
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/35591048019427950293

碩士 === 元智大學 === 電機工程研究所 === 89 === This thesis investigates the multichannel equalization and direction finding (DOA estimation) problems for antenna arrays. For GSM-based multichannel equalization, we implement the simplified Decision Feedback Maximum Likelihood Sequence Estimation algorithm including DF(1)_RS_MLSE and DF(2)_RS_MLSE. Past experiments show that incorporating feedback decisions can reduce the amount of computations due to the increased number of antenna elements, and performance improvement can still be retained for the simplified algorithms. To exploit the feasibility of DF_RS_MLSE for real time operation, we implement the algorithms in TMS320C6201 DSP chip. Our results show that the computation time can conform to the GSM requirement (4.615ms/burst). In DOA estimation, we use Normalize Radial Basis Function Network (NRBFN) to directly compute the DOA estimate, and compare its performance with RBFN. Simulations show that NRBFN can largely reduce the number of basis functions while achieving higher accuracy. We also propose a new way to set the initial values of network weights. Simulations show that the new method can notably improve the performance of RBFN.