Time and Frequency Domain Description of Gilbert-Elliott Data Loss Models

Abstract

Nowadays, the usage radio channel or Internet connection grows for real-time transmission of measurement data or signals. For these applications, not reliable communication needs to be analyzed from a signal processing point of view. Our paper contributes to this topic with the analysis of data loss as a signal distorting phenomenon. The paper contains a possible mathematical description of the data loss via the indicator function. Moreover, the Gilbert-Elliott data loss model class is presented, which gives a framework to analyse numerous data loss phenomena in general. In order to analyse the data loss in the frequency domain, the spectrum of the indicator function is required. This paper presents the derivation of the power spectral density of the Gilbert-Elliott model. The paper reviews a possible identification method for the simpler models derived from the Gilbert-Elliott model. Theoretical analysis and simulations are supported by measurements. The measurement results are used to show how the data loss models function in practice.