MVDR based adaptive beamformer design and its FPGA implementation for ultrasonic imaging

Adaptive beamforming outperforms its non-adaptive counterpart in that it is capable of achieving a narrower main lobe and suppressing interference. It, however, calls for intensive computations to obtain the weighting factors. In this paper, we present an MVDR (minimum variance directionless response) based adaptive beamformer design for ultrasound imaging and implement the design in an FPGA platform. An efficient matrix inversion computing scheme, which can reduce the complexity by an order, is first developed to facilitate real time operations. Simulation results indicate the performance degradation due to this simplification measure is negligible. A linear systolic array design is further devised to support highly concurrent and pipelined computations. Finally, the design is implemented in a Xilinx Vertex XC7z045ffg900-2 FPGA platform. The working frequency can reach 98 MHz, and the MVDR coefficients can be updated every 64 clock cycles. For the entire ultrasound system, the proposed design can support a 64 lines by 512 pixels scanning at a frame rate of 20fps.