Synthesis method for broadband beamformer based on FIR filters designed by neural networks

A approach to design a novel constant beamwidth beamformer used in underwater sonar system was proposed. In the proposed algorithm, the FIR digital filters based on a batch back-propagation neural network with a variable learning rate mode and combined with digital-delay-lines were used to weight the received signals. Simulation of the proposed algorithm was done on an array of 16-sensors equally spaced line array. The results show that the broadband beamformer based on FIR filter designed by BP neural networks had better performance. Meanwhile, the computation of designed broadband beamformer can be effectively reduced since the design of FIR filter were symmetrical structure.