A Study On Wavelet Packet Based Algorithm For Representation of Wireless Channel

Abstract

Wireless telecommunication services are growing rapidly both in terms of the underlying technology as well as in the nature of the applications. The trends point to further growth in the foreseeable future. However, the dynamic nature of the wireless channel places fundamental limitations on the performance of wireless communication system. Unlike wired channels, which are by-and-large stationary and predictable, wireless channels are extremely random and do not offer easy analysis. Therefore, in order to gain better system performance it is of great importance to accurately and efficiently map the wireless channels. Existing wireless channel models are based on statistical impulse response methods derived from empirical analysis. Such models employ a lot of channel coefficients which affect the complexity of computation. In this context the theory of wavelets and wavelet packets, which have recently found applicability for signal processing applications, hold the promise for wireless channel modeling. Wavelet packets offer the important feature of localization in both frequency and time domains. This property can be exploited to model channels with reduced complexity. In this thesis report, the applicability of wireless channel representation based on wavelet packet algorithm is addressed. The possibility of one-dimensional and two-dimensional wavelet packet algorithm for time-invariant and time-varying representation, respectively, is investigated. The results illustrating the efficiency of one-dimensional and two-dimensional approach are presented. To improve the performance of the proposed system, two optimization methods, namely coefficient reduction and tree pruning are implemented. Furthermore, the impact of the number of decomposition levels and the type of mother wavelet employed are also investigated. To measure the system performance, first and second order stochastic metrics such as Mean-square Error (MSE), Level-Crossing Rate (LCR) and correlation were employed.