Direction Of Arrival Estimation Based On RBFFNN And Beamforming

Space spectrum estimation, beamforming and nulling are important technique in smart antenna of TD-SCDMA communication standard. Space spectrum estimation concerns direction of arrival(DOA) of signals. Classical algorithm of DOA estimation is hard to meet real time system's requirement due to its very large computation. Radial-basis function neural network(RBFNN) method has better real-time performance for needing less computation. Mutual coupling between the elements of the circular array severely affects output signal of the array. Otherwise, output of RBFNN depends on its basis functions' centers, so compensation of training data for RBFNN is necessary for correct centers. First, generalized impedance matrix of array is calculated by Method of Moment; then, the training data for RBFNN is compensated by Direct Data Domain algorithm using generalized impedance matrix; after RBFNN's training stage finished, it can estimate DOA. At last, pattern pointing to desired users and nulling interferes can be got with technique of beamforming and nulling. Simulation results show the feasibility of our method.