A generalized approach to the design of variable fractional-delay FIR digital filters

In this paper, a complex-oriented weighted least-squares approach is proposed for the design of arbitrary variable fractional-delay FIR filters. The objective error function is formulated in a quadratic form, such that the filter coefficients can be obtained by proper matrix/vector operation, including matrix inversion. However, all elements of related matrices and vectors can be represented in closed forms, and the matrix to be inversed is a positive-definite Hermitian symmetric matrix, which can be decomposed by the Cholesky factorization, such that computation time can be effectively reduced and the ill condition can be avoided. Comparing with the existing methods, the proposed method can be applied to design arbitrary complex coefficient or real coefficient variable fractional-delay filters. Several examples are presented to demonstrate the flexibility and effectiveness of the proposed method.