An OFDM Signal Identification Method for Wireless Communications Systems

TL;DR

This paper presents a novel OFDM signal identification method using fourth order cumulants and a chi-square Gaussianity test, which is robust against channel impairments and outperforms existing statistical tests.

Contribution

It introduces a new OFDM signal identification technique based on cumulant estimates and a specialized Gaussianity test, simplifying covariance estimation and improving robustness.

Findings

The method effectively distinguishes OFDM signals under various channel conditions.

It outperforms fixed-threshold and empirical statistical tests in accuracy.

The approach requires lower complexity than coherent identification methods.

Abstract

Distinction of OFDM signals from single carrier signals is highly important for adaptive receiver algorithms and signal identification applications. OFDM signals exhibit Gaussian characteristics in time domain and fourth order cumulants of Gaussian distributed signals vanish in contrary to the cumulants of other signals. Thus fourth order cumulants can be utilized for OFDM signal identification. In this paper, first, formulations of the estimates of the fourth order cumulants for OFDM signals are provided. Then it is shown these estimates are affected significantly from the wireless channel impairments, frequency offset, phase offset and sampling mismatch. To overcome these problems, a general chi-square constant false alarm rate Gaussianity test which employs estimates of cumulants and their covariances is adapted to the specific case of wireless OFDM signals. Estimation of the covariance matrix of the fourth order cumulants are greatly simplified peculiar to the OFDM signals. A measurement setup is developed to analyze the performance of the identification method and for comparison purposes. A parametric measurement analysis is provided depending on modulation order, signal to noise ratio, number of symbols, and degree of freedom of the underlying test. The proposed method outperforms statistical tests which are based on fixed thresholds or empirical values, while a priori information requirement and complexity of the proposed method are lower than the coherent identification techniques.