Dual-kernel Sparse Approximation Of UWB Signals And Noise Suppression With Windowing

Ultra-Wideband (UWB) wireless communication is one of the key technologies for the next generation of wireless communication networks. It has many advantages such as strong anti-interference performance, high bandwidth, low power consumption, high security, etc, and hence is a research hotspot in wireless communications. UWB communication systems are sensitive to the channel fading. Therefore, channel estimation plays a very important role in the UWB technology. According to the Nyquist sampling theory, UWB digital receivers need to sample signals at high rate before performing the channel estimation. In practice, the sampling of UWB signals is a big challenge for analog-to-digital converters (ADCs) and thus one of the bottlenecks of UWB technology development.To reduce the sampling rate of UWB digital receivers, J. Paredes, G. R. Arce and Z. Wang proposed to solve the problem of channel estimation in the UWB systems using compressed sensing (CS) theory. According to CS, if a signal shows sparsity in some dictionary or basis, then we can reconstruct the original signal by using some nonlinear reconstruction algorithm with a number of random measurements far less than that of Nyquist sampling. By using CS, the sampling rate of UWB signals can be reduced greatly.In this thesis, we focus on CS-based UWB channel estimation. The main contributions of this thesis include the following two aspects:1. Based on the consistent sampling theory, we investigate the sparse approximation of UWB signals and propose a new method of dual-kernel sparse approximation of UWB signals. Compared with the traditional method of single-kernel sparse approximation, the newly proposed captures the energy of UWB signals more efficiently and thus reduce the approximation error effectively. Based on the idea of dual-kernel sparse approximation, we present some dual-kernel dictionaries, including dual-kernel Hilbert dictionary (DKHD), dual-kernel cross-Hilbert dictionary (DKCHD) and dual-kernel accessorial-Hilbert dictionary (DKAHD). These dictionaries are used in CS-based UWB channel estimation and evaluated via simulations. Simulation results show that:1) Among single-kernel dictionaries, the dictionary generated by Hilbert generating kernel performs as well as the conventional dictionary generated by UWB pulse generating kernel, and outperforms the dictionary generated by derivative generating kernel.2) Compared with single-kernel dictionaries, the newly proposed dual-kernel dictionaries can be used to obtain higher channel estimation precision. As a result, the bit error rate (BER) of the UWB receiver can be improved effectively.2. In order to reduce the influence of noise folding on CS-UWB channel estimation, we propose a new method of time domain windowing. In the proposed method, the characteristics of UWB multipath channels, e.g., the averaged power delay profile (APDP) and root-mean-squared (RMS) delay spread, are taken as prior knowledge. A time domain window is designed according to the priors of channel and used to adjust the measurement duration of CS-based UWB channel estimation. Due to the fact that the noise outside the window is suppressed, the influence of noise on channel estimation is reduced significantly. Simulation results show that:1) For the same number of measurements, the newly proposed method of time domain windowing can be used to gain higher channel estimation precision and better BER performance than the traditional method of CS channel estimation; 2) For the same requirement of BER, the method of time domain windowing requires much lower sampling rate than the traditional method of CS channel estimation; 3) For the method of time domain windowing, the proposed dual-kernel dictionaries show better BER performance than single-kernel dictionaries.