Blind and semi-blind channel equalization for GSM wireless communication systems

In this dissertation, applications of blind/semi-blind equalization principles in practical wireless cellular systems are investigated. Specifically, we focus on the popular Global System for Mobile (GSM) communication system in our effort to improve the system efficiency while maintaining the system performance at an acceptable level. We begin by deriving a linear approximation model for the Gaussian Minimum Shift Keying (GMSK) signal used in the GSM standard. A single-input two-output diversity model is obtained through a de-rotation scheme. Conditions for blind identification of the two diversity sub-channels are derived. This model is essential, as virtually all known blind equalization algorithms are based on linear system models. With the model, several second order statistical (SOS) blind channel identification algorithms are adopted for GSM applications. Next, two blind adaptive equalizers based on higher order statistics (HOS) are also adopted for GSM, and their convergence issues are discussed. These direct adaptive equalizers are found to be robust to channel order mismatch and are shown to converge well in the presence of our de-rotation induced diversity channels. A new two-step channel identification algorithm is introduced to improve GSM receiver performance by exploiting the equalizer output for more accurate channel identification. This new algorithm combines one of the HOS blind equalizers with a traditional least squares channel identification method. Lastly, two different semi-blind algorithms are developed for the GSM system. These semi-blind algorithms can identify the channel by combining both SOS information and information from training. They can overcome some limitations of SOS blind algorithms. (Abstract shortened by UMI.)...