Detection of Sinusoids with Frequency Drift in White Gaussian Noise

Abstract

In this study, 4 methods of detecting sinusoid targets in white Gaussian noise are investigated. Averaging FFTs coherently and non-coherently is examined. A brute force method is introduced and investigated. Using a single large FFT spanning the observation window is also considered. Detection performance is studied on targets whose frequencies drift over time. It is determined that averaging non-coherent FFTs is a safe choice in a detection system. The preferred size of the FFT used for tone detection as well as the number of spectra to average is driven by the nature of the target tone frequency drift. This information is especially useful when high sample rate detectors are used to find tones from low quality transmitters because making FFTs too large or taking too many averages may hinder detection.