Junichiro Suzuki

A Simple Method to Stop an Adaptive Process for the Multistage Wiener Filter

SUMMARY The multistage Wiener filter (MWF) outperforms the full rank Wiener filter in low sample support environments. However, the MWF adaptive process should be stopped at an optimum stage to get the best performance. There are two methods to stop the MWF adaptive process. One method is to calculate until the final full-stage, and the second method is to terminate at r-stage less than full-stage. The computational load is smaller in the latter method, however, a performance degradation is caused by an additional or subtractive stage calculation. Therefore, it is very important for the r-stage calculation to stop an adaptive process at the optimum stage. In this paper, we propose a simple method based on a cross-correlation coefficient to stop the MWF adaptive process. Because its coefficient is calculated by the MWF forward recursion, the optimum stage is determined automatically and additional calculations are avoided. The performance was evaluated by simulation examples, demonstrating the superiority of the proposed method.

The bank-overlapped post-Doppler method and its system design for self-developed parallel processing boards

Post-Doppler methods are rank reduction techniques to reduce temporal dimensions by applying the Doppler filter and can reduce the processing time than full degrees-of-freedom (DOF) space-time adaptive processing (STAP). In post-Doppler methods, multi-bin post-Doppler methods are generally used to improve the performance of signal-to-interference-plus-noise ratio (SINR). However, it is necessary to solve the issue of different adaptation for each Doppler bin by using the covariance matrices which formed by multiple filter banks. Therefore, it is very effective to calculate each optimum weight in parallel. For further improvements of processing time, we propose an efficient approach to calculate not one but multiple Doppler bins. When a uniform filter weight is applied to the proposed method, the performance of SINR deteriorate in the vicinity of bank edges. Therefore, we improved the performance by overlapping in the filter bank of both ends. In addition, we show its system design and benchmark results by using self-developed digital signal processor (DSP) boards.

An Efficient Dimensional Reduction of the Blocking Matrix for the Multistage Wiener Filter

The multistage Wiener filter (MWF) is a powerful adaptive processing in low sample support environments. Because the MWF solution is provided by a stage-by-stage decomposition, the computational load increases depending on degrees-of-freedom (DOF). In this paper, we propose two efficient approaches by reducing dimensions of the blocking matrix. One approach is to delete some rows of the blocking matrix at the 1st stage, and following stages are calculated by the normal method. The second approach is to delete some rows at all stages consistently. Although the MWF adaptive process of the proposed approaches must be stopped at the optimum stage to avoid a performance degradation, it was solved by applying a simple stopping criterion based on a cross-correlation coefficient. The performance was evaluated by simulation examples, examining the effectiveness.