FPGA implementation of normalized correlation function

Abstract

Correlation analysis is a frequently used tool in signal detection and classification tasks. This paper presents the design and FPGA implementations of a hardware module for calculating the Pearson correlation coefficient. This module is designed for use in signal template matching, where a measurement signal is correlated with a template. It has been described in Verilog and implemented on Intel Cyclone V FPGA. The module consists of two main parts, which are: a correlation filter and normalization modules. Correlation filters performing the calculation in the time domain and in the frequency domain are described. The project has been verified in simulation using ModelSim and checked on hardware. As a result of this work, hardware IP cores are developed enabling parametrization and programming in data word-lengths, filter size, calculation speed, FFT/IFFT size, length, and number of processing templates. Developed resources are intended to be used in FPGA-based hardware, e.g. DAQ systems, working with sampling frequencies from kHz to above 130 MHz.

Keywords

• correlation
• hardware algorithms
• fpga
• embedded systems
• time series analysis
• pulse recognition