作者: Fang Zhou , Jun-Zheng Jiang , Peng-Lang Shui
DOI: 10.1016/J.SIGPRO.2014.12.021
关键词:
摘要: This letter presents a low-complexity algorithm to design two-dimensional (2D) DFT modulated filter bank (DMFB) with nearly perfect reconstruction (NPR). The problem is formulated into an unconstrained optimization whose cost function consists of the overall distortion and stopband energy. By exploiting gradient information, prototype (PF) coefficients are iteratively optimized closed-form formula. At each iteration, computational complexity dramatically reduced by evoking Toeplitz-block Toeplitz matrix inversion lemma, which makes suitable for 2D DMFB large number channels PF coefficients. Numerical example comparison included show effectiveness proposed algorithm. energy block matrix.The largely via inverse lemma..The designing high-complexity DMFB.