In this thesis, we propose a cost-based CLEAN algorithm to accurately find dense multi-path parameters and improve the performance of selective RAKE receiver in indoor UWB systems. RAKE receiver can resolve the dense multi-path interference problems with the multi-path parameters. Because the weak paths are of lower valuable for system performance improvement, selective RAKE receiver combines only the strongest multi-path components and reduce the number of fingers to lower the complexity of RAKE receiver. However, selective RAKE receiver needs accurate multi-path detection to decide the suitable number and parameters of fingers. In order to improve the performance of selective RAKE receiver, the main issue in this thesis is to detect the best paths of channel with the CLEAN algorithm. CLEAN algorithm uses the correlation of the received signal and the template signal as the basis for searching paths. If there are closely adjacent paths, or if one of signal paths is relatively stronger, the detection error of paths may occur and thus affects the performance of the receiver. EP-based CLEAN algorithm uses the cost function and the evolutionary programming (EP) to search the multi-path delay times and gain coefficients for minimizing the cost function. Accurate multi-path detection and high resolution of adjacent paths can be obtained. However, EP-based CLEAN algorithm makes a time-consuming blind search. In the thesis, a CLEAN algorithm based on the cost function is proposed. The proposed cost-based CLEAN algorithm searches the delay times near the peaks of the cross-correlation for local minimum of the cost function, and then uses CLEAN algorithm to extract autocorrelation components and obtain the accurate multi-path detection. By testing the IEEE802.15.3a UWB channel models, and comparing with CLEAN algorithm, the cost-based CLEAN algorithm in the thesis can achieve better detection accuracy in multi-path searching, and improve the performance of selective RAKE receiver.