Generalized Correntropy Induced Metric Based Block Sparse Memory Improved Proportionate Affine Projection Sign Algorithm

In this paper, by combining the block sparse memory improved proportionate affine projection sign algorithm (BS-MIP- APSA) and the generalized correntropy induced metric (GCIM), we have proposed two block sparse adaptive filtering algorithms, named as GCIM-BS-MIP-APSA-I and GCIM-BS-MIP-APSA-II, by directly using GCIM to measure the sparse information and treating GCIM as a function of block weight vector, respectively, for identification of sparse systems. Based on the good characteristic of GCIM approximating the $\ell_{0}$-norm, the proposed algorithms, especially GCIM-BS-MIP-APSA-II, can achieve improved filtering accuracy and faster convergence rate when the identified system is block sparse. We also have analyzed the computational complexity of proposed algorithms. Some simulation results are conducted to demonstrate the efficiency of GCIM-BS-MIP-APSAs compared with other competing algorithms.