Research On CS Based UWB Channel Estimation Methods

IR-UWB is a promising technology for short-range wireless communication systems. It has the advantages of low power consuming, high bandwidth, low complexity, anti-multipath interference and high security, etc. With the developing of computer technology and digital signal processing technology, we prefer to process IR-UWB signal in digital domain. Nyquist sampling theory shows that in order to keep the whole information of the target signal, the sampling rate must be at least twice of the maximum frequency present in the signal. However, due to the huge bandwidth of the IR-UWB signal, signal sampling must be one of the challenges for the realization of IR-UWB systems.Compressed sensing (CS) is a new signal processing framework. It can sample and compress the signal at the same time. According to this theory, if the signal is sparse or compressible in some dictionary or basis, then we can reconstruct or approximate the original signal with a limited number of linear projection measurements which is far less than that of the Nyquist theory. One of the key issues to the success of CS theory is the measurement matrix. It must be incoherent with the sparse dictionary of the original signal. Thus, CS theory provides a feasible way to solve the sampling problems of the IR-UWB system and many researches has been done in the field in recent years. Aiming at improving the channel estimation performance, reducing the computation complexity and improving the efficiency of the channel reconstruction, we do some research of the CS based IR-UWB channel estimation methods in this thesis. The main contributions of this thesis include the following two aspects:1. For the CS theory, the presence of AWGN noise can make the OMP algorithm inevitably selects some false atoms in the iterative reconstruction procedure and this can degrade the reconstruction performance greatly. In order to improve the atom selection precision of this algorithm, we propose a weighted OMP (WOMP) algorithm. For a given sparse dictionary, a weighting factor is assigned to each of the atoms according to prior information about the UWB channels and a weighted matching process is performed by WOMP. For the multipath dictionary, the averaged power delay profile (APDP) can be a good choice to decide the weighting factor of each atom. For the eigen based dictionary, the corresponding eigenvalues (i.e., eigenvectors) can be used as weighting factors for each atom. Simulation results show that no matter adopting which sparse dictionary, using the WOMP can get better correlation coefficients and bit error rate (BER) performance than that of using OMP algorithm.2. For the traditional CS based channel estimation method, we need to estimate the whole channel waveform (the correlation template) at a time. However, for the whole channel waveform, the corresponding sparsity and sampling numbers must be very large. So the computation complexity is relatively high. In order to further reduce the computation costs, we proposed a segmented CS based channel estimation methods, In this way, the channel waveform is gotten segment by segment. For each segment, the sparse level and sampling numbers is much lower than the whole channel waveform and thus can reduce the computation complexity greatly. Simulation results show that besides lower computation complexity, employing the new method can also get better BER performance than the traditional CS UWB channel estimation methods.