基于并行结构的DOA聚类算法及FPGA实现

FPGA Implementation for a DOA clustering algorithm based on parallel architecture

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。

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.