互相关算法在Cortex-M3平台上的实现和优化  

Implementation and optimization of cross-correlation algorithm on Cortex-M3

作　　者：黄娇郁 HUANG Jiaoyu(Weather Bureau of Jiangxia District,Wuhan 430200,China)

机构地区：[1]江夏区气象局,湖北武汉430200

出　　处：《电子设计工程》2021年第3期93-98,共6页Electronic Design Engineering

摘　　要：互相关算法在多传感器测量仪器中有重要用途,FFT算法可以用于加速互相关计算,尤其在序列点数较大的时候,FFT计算互相关的速度优势明显。然而即便如此,FFT计算对于CPU的速度要求仍然很高。直接计算FFT一般需要使用带有浮点协处理器的CPU,这对于设备成本和功耗提出了要求;CPU执行一次浮点运算所需要的时间一般高于整数计算。在Cortex-M3定点CPU平台上实现和优化了一套计算函数,通过对FFT算法的整数化,使算法适用于定点CPU;将旋转因子参数和倒位序存储为数组,使用查表方式加速计算;根据参与互相关的两个原始信号都是实数序列的特点,进一步优化计算流程,显著降低了计算量。The cross-correlation algorithm has important useage in multi-sensor measuring instruments,and the FFT algorithm can be used to accelerate the cross-correlation calculation,especially when the sequence points are large,the speed advantage of the cross-correlation calculation is obvious.Even so,the FFT calculation still requires high speed CPU.Direct computation of the FFT generally requires the use of a CPU with a floating-point coprocessor,which requires device cost and power consumption.The CPU generally takes more time to perform a floating-point operation than integer computation.This paper implements and optimizes a set of computing functions on the fixed-point CPU platform Cortex-M3,which makes this algorithm suitable for fixed-point CPU by integering the FFT algorithm,calculates the rotation factor parameters and inverted orders array in RAM,and accelerates the calculation by using the look-up table.According to the characteristics that the two original signals involved in the cross-correlation are real number sequences,the calculation flow is further optimized and the computation quantity is reduced markedly.

关 键 词：互相关 延时估计 FFT优化 CORTEX-M3 

分 类 号：TP393.2[自动化与计算机技术—计算机应用技术]