High resolution multipath channel estimation method of gradient-type least square in real number field

In order to solve the multi-path channel high-resolution estimation problem in underwater localization, a multi-path channel high-resolution estimation method in real number field based on gradient-type least squares is presented. The method reduces the computational effort by removing the matrix inversion from the time-domain iterative solution and directly obtains the high-resolution real domain channel impulse response, which is different from the method in frequency field. A comparative analysis among several typical time-domain high-resolution channel estimation algorithms, such as MUSIC, Richardson-Lucy (RL) and sparse Bayesian learning (SBL) is performed. The simulation and experimental results show that the gradient-type LS and SBL algorithms have significantly better time-delay resolution than the MUSIC and RL algorithms. The pool experimental results show that the time-delay resolution can reach 1/(5B) (B is the signal bandwidth) by using the proposed method. The amplitude attenuation estimation accuracy of the proposed method is slightly worse than that of SBL algorithm. However, the proposed method does not require the number of multipaths in advance, which makes it more suitable for large time-delay multipath extensions, and has significantly less amount of calculation than the SBL algorithm.