Joint Parameter And Time-Delay Estimation For Multi-Input Fir Systems Based On Compressed Sensing Recovery

The finite impulse response systems are a class of typical linear systems and widely exist in industrial field. When the systems have unknown time-delay, it is required to estimate not only the unknown parameters but also the time-delay, which increase the difficulty and complexity of the systems identification. Therefore, the joint parameter and time-delay estimation for multiple-input single-output finite impulse response(MISOFIR) systems with unknown time-delay have important theoretical and practical value.By combining the compressed sensing recovery, the least squares principle and gradient optimization principle, this thesis study the parameter and time-delay estimation for MISO-FIR systems with unknown time-delay. The main contents are listed as follows:1. For the parameter and time-delay estimation of the MISO-FIR systems with unknown time-delay, by combining the compressed sensing recovery and the least squares method, the threshold orthogonal matching pursuit(TH-OMP) algorithm is proposed for simultaneously estimating the parameters and time-delay with limited sampling data. In order to improve the computational effectiveness of the TH-OMP algorithm,the gradient search principle is applied to replace the least squares optimization in the TH-OMP algorithm, so that a gradient pursuit(GP) algorithm is proposed. The computation analysis shows GP algorithm has less computational burden compared with the TH-OMP algorithm, and the simulation results show the effectiveness of proposed algorithms.2. For the multiple-input single-output finite impulse response moving average systems with unknown time-delay, the information vector contains unknown white noise variables. In order to solve that problem, by combining iterative identification principle and interaction estimation principle, the threshold orthogonal matching pursuit iterative(TH-OMPI) algorithm and gradient pursuit iterative(GPI) algorithm are proposed respectively for simultaneously estimating the parameter and time-delay with limited sampling data. The computational analysis result shows that GPI algorithm has less computational cost compared with TH-OMPI algorithm. The simulation results demonstrate the effectiveness of the proposed algorithms.3. For the multiple-input single-output finite impulse response autoregressive systems, information vector contains more complex color noise variables compared with moving average systems. By combining iterative method and interaction estimation method, the identification model of the system is rebuilt, and the threshold orthogonal matching pursuit iterative(TH-OMPI) algorithm and gradient pursuit iterative(GPI)algorithm are proposed for the finite impulse response autoregressive systems whose colored noise more complex than moving average systems. In order to show the excellence of proposed algorithms, the least squares iterative(LSI) algorithm is used for comparison. The simulation results and computational analysis show that in the case of identification accuracy approaching the TH-OMPI algorithm has less computational cost compared with LSI algorithm as well as GPI algorithm compared with TH-OMPI algorithm.In summary, this thesis focuses on joint parameter and time-delay estimation for multiple-input single-output finite impulse response systems, and the parameter and timedelay estimation algorithms are derived for each systems respectively. Through computational analysis demonstrate that the proposed algorithms have a better identification efficiency. The numerical simulation results verify the effectiveness of the proposed algorithms.