Application of nonlinear regression in recognizing distribution of signals in wireless channels

Abstract

In many applications, it is important to recognise the distribution of empirical data in almost real time. One of the specific applications is the identification of statistical models for fading in wireless systems of the base station receivers. This is one of the most important problems in spatial diversity. In this paper, we describe the methodology and the results of a nonlinear regression approach for recognising the distribution of the input signal with the values of its parameters. Furthermore, the proposed approach could be used for the real-time recognition of the probability distributions without any prior knowledge about the input signal. To prove its performance, the Levenberg–Marquardt nonlinear least-squares algorithm is tested on a large set of randomly generated signals with the Gamma, Rayleigh, Rician, Nakagami-m, and Weibull distributions. The experimental results demonstrate that this approach is accurate in recognizing statistical distributions from the signal.

Keywords

• data mining
• classification
• nonlinear regression
• curve fitting
• probability distribution.