Extracting the parameters of the multipath with high accuracy can be achieved by using high-resolution algorithm for time-domain ultra wideband (UWB) channel modeling.The CLEAN algorithm has been used as such a high-resolution algorithm for UWB time-domain characterization.This paper presents a compressed sensing (CS) based high-resolution deconvolution algorithm for time-domain UWB channel modeling.UWB wireless channels are a prime example of long and sparse channel impulse response (CIR).Furthermore,the dictionary of parameterized waveforms that closely matches the waveform of multipath leads to that the UWB channel measurement signal is more compactly represented.By adjusting the parameter of dictionary,CIRs of different resolutions can be obtained.The matching pursuit (MP) algorithm is used as the signal reconstruction method for CS and outputs the CIR directly.We also demonstrated that if the dictionary of CS is designed specifically,MP is an equivalent of single template CLEAN.Finally,the computation complexity of CS-MP is analyzed and comparison of MP and CLEAN is performed.Simulation results show that compared to CLEAN,the proposed CS-MP deconvolution algorithm can achieve a comparable performance with much fewer samplings. Extracting the parameters of the multipath with high accuracy can be achieved by using high-resolution algorithm for time-domain ultra wideband (UWB) channel modeling. The CLEAN algorithm has been used as such a high-resolution algorithm for UWB time-domain characterization. This paper presents a compressed sensing (CS) based high-resolution deconvolution algorithm for time-domain UWB channel modeling. UWB wireless channels are a prime example of long and sparse channel impulse response (CIR) .Furthermore, the dictionary of parameterized waveforms that closely matches the waveform of multipath leads to that the UWB channel measurement signal is more compactly represented. By the way, the matching pursuit (MP) algorithm is used as the signal reconstruction method for CS and outputs the CIR directly. We also demonstrated that if the dictionary of CS is designed specifically, MP is an equivalent of single template CLEAN.Fi nally, the computation complexity of CS-MP is analyzed and comparison of MP and CLEAN is performed. Simulation results show that compared to CLEAN, the proposed CS-MP deconvolution algorithm can achieve a comparable performance with much fewer samplings.