The emerging theory of compressed sensing (CS) enables the reconstruction of sparse signals from a small set of random measurements. Since CS samples the signals at sub-Nyquist rate, it is particularly suitable for impulse ultra-wideband (I-UWB) communications where Nyquist sampling of the signal is a formidable challenge. In our previous work, CS has been successfully applied for I-UWB channel estimation and symbol detection. In this paper, we show that the performance of the I-UWB detector based on compressive measurements can be improved by exploiting the signal sparsity model. The matching pursuit (MP) algorithm is used to estimate the signal sparsity model from compressive measurements and then a more effective measurement matrix is designed for unknown signal detection. Performance of the proposed detector is analyzed. Simulation results show that the proposed detector has comparable performance to the digital receiver sampling at Nyquist rate.