Study On Compressed Sensing And UWB Channel Modeling

Ultra Wide-band (UWB) wireless channel modeling is a prerequisite for UWB system design. Compressed sensing (CS) has a great potential in the UWB channel estimation and modeling. This thesis is supported by Important National Science&Technology Specific Projects and attempts to make a contribution to the theory and application of UWB system.In this paper, based on UWB channel measurement data, the basic principle of compressed sensing technology and its application in UWB system channel modeling and receiver design are investigated. Several creative algorithms are developed in this thesis.The IEEE802.15.3a/4a channel models are not applicable for the UWB channels of China. Based on the UWB frequency regulation specification of China, large scale channel measurements were performed for indoor, outdoor and industrial12scenarios. A novel aumated cluster identification based on wavelet analysis is proposed. All parameters for modified S-V channel models are obtained by fitting from the channel measurement data. The proposed channel models are verified by tesing with the measurement data and are more effective than the IEEE channel models for China environments.A new piecewise double exponential channel model for office non-line of sight (NLOS)case is proposed based on the measurement data. Compared to the IEEE802.15.4a channel model, the proposed office NLOS channel model for has more similar channel characteristic on delay spread and number of paths with measurement data. Another compressed sensing based average number of multipath model of UWB channel is proposed. It reveals that there exists a functional relation between the number of compressive samples required by CS and the average number of UWB channel.To solve the high sampling rate problem in UWB channel model measurement, this thesis proposed deconvolution algorithm based on compressed sensing. For frequency-domain and time-domain measurement approach, the time-domain window pulse corresponding with the frequency domain window function and the template signal is used to construct a parameterized waveform dictionary respectively. The dictionary can enhance the sparse representation of time-domain measurement data. Greedy restruction algorithms can be used to perform deconvolution. Discrete channel impulse response are obtained. Compared to the traditional CLEAN deconvolution algorithm, the CS based deconvolution algorithm can achieve similar deconvolution performance with much fewer samples required by Nyquist sampling rate. We also demonstrated that with a dictionary designed specifically, MP algorithm is an equivalent of CLEAN algorithm. When the frequency dependent distortion of UWB channel is considered, a multi-template CS-based deconvolution is proposed, which improves deconvolution performance.Consider the high sampling rate of traditional UWB channel estimation, we present a compressed channel estimation algorithm with channel state a priori information. By using the partial channel state information Average Power Delay Profile (APDP) to weight the atoms of dictionary, modifying the atoms selection procedure based on the principle of Probablistic MP (PMP), the proposed reconstruction algorithm not only reduces the requirement of highg sampling rate, but also improves the performance under low Signal-to-Noise Ratio (SNR) using channel a priori information. The simulation results show that the prosed CS-SWMP algorithm achieves higher channel estimation accuracy than other reconstruction algorithm, with almost identical complexity.Considering the high sampling rate of traditional UWB noncoherent detection, we present a compressed channel estimation algorithm with channel state a priori information APDP. To make the CS-based noncoherent detection feasible for real-time application, the block sparse reconstruction algorithm Block Orthogonal Mathcing Pursuit (BOMP) is exploited for the block sparsity characteristic of received UWB-IR signal. The CS-based noncoherent detection algorithm outperforms the energy detection algorithm and reduces sampling rate.Finally, the content of the whole dissertation is summarized, and several valuable research directions of compressed sensing and UWB channel modeling are discussed.