A sparse system identification method of variable step size lp norm LMS algorithm

Abstract

The invention discloses a sparse system identification method of variable step size lp norm LMS algorithm. The convergence of the algorithm is analyzed. The square root of the error generated in the iterative process is introduced into the step size control, and the normal number Vth is introduced to balance the convergence rate and steady-state error of the system. The step size of the system canbe adjusted effectively in real-time during the iterative process. In order to mine the sparsity of unknown channel, we introduce the lp norm into the cost function, and the algorithm can identify the sparse system accurately On the basis of keeping the excellent convergence rate and steady-state performance of the traditional LMS algorithm, the algorithm further improves the convergence rate andestimation accuracy of the system. Several new sparse adaptive algorithms are compared in experimental simulation, and the superiority of the proposed algorithm is verified by theory and experiment.

Description

technical field [0001] The invention belongs to the field of sparse signal processing and relates to a variable step size l p The sparse system identification method of the norm LMS algorithm is a sparse system identification method in a noisy environment, and can be used for channel estimation in broadband wireless communication systems. Background technique [0002] Adaptive filtering has strong adaptability and better filtering performance. This algorithm has been widely used in channel equalization, linear prediction, spectrum analysis and system identification, radar, echo cancellation and other fields. The purpose of adaptive filtering is to deal with uncertain systems or information. The "uncertainty" here means that the mathematical model of the studied information processing process and its environment has not been fully determined, it contains some unknown and random factors, for example, some noise signals will affect information processing in different ways, the...

AI Technical Summary

Problems solved by technology

The LMS / F algorithm is an adaptive filtering algorithm composed of the LMS algorithm and the LMF algorithm proposed by SJ Lim and JG Harris. It can also be regarded as an adaptive filtering algorithm with variable step size. When the weight vector is far from the optimal value, the convergence speed and stability are better than the LMS algorithm, and when the weight vector is close to the optimal solution, the stability is better than the LMF algorithm, but because the LMS / F algorithm is in the weight vector update A third-order function of the error is used, so it is very sensitive to both additive noise and signal-to-noise ratio

[0004] For sparse unknown systems, the above adaptive algorithm cannot effectively utilize the sparse structural information in the system, and the performance efficiency of parameter estimation is not high

Images