基于并行结构的DOA聚类算法及FPGA实现

脉冲到达角（DOA）是脉冲信号分选中可利用的重要参数。目前,利用DOA进行的脉冲分选都是基于传统的串行聚类算法,实时性能差。本文针对阈值分割的聚类方式,设计了一种基于并行流水结构的实时聚类算法,使单个DOA的聚类可在单周期内完成,并通过对聚类数目分裂过多的情况进行控制,保证了算法的稳定性和有效性。文章还介绍了算法在FPGA上的实现方法,以及应用在XilinxV2P芯片上的实时性能,并对其聚类性能进行了比较分析。

FPGA Implementation for a DOA clustering algorithm based on parallel architecture

Direction of Arrival (DOA) is an important parameter in pulse sorting. Current pulse sorting techniques which concerned with DOA are almost based on traditional serial clustering algorithms, hardly achieving a real-time performance. In this paper, a real-time threshold partition algorithm is proposed based on single-cycle clustering for every DOA sample by a parallel and pipeline architecture. Stability and effectiveness are well sustained by strategy of over-split control. The implementation of the algorithm on FPGA is also introduced, providing its real-time performance with Xilinx V2P series chips.