Title :
Matrix formulation: fast filter bank
Author :
Ching, Lim Yong ; Wei, Lee Jun
Author_Institution :
Sch. of Electr. & Electron. Eng., Nanyang Technol. Univ., Singapore, Singapore
Abstract :
The fast filter bank (FFB) describes a class of tree-structured filter banks that operate on a frequency response masking principle. Although the structure is highly regular and conveniently implemented in hardware designs, real-time software implementations lead to inefficiencies due to its branching structure. In this paper, an alternative formulation of the FFB is proposed in terms of matrix computations. This allows an efficient approach in its implementation, and significantly reduces the overall buffer memory size required. The matrix operations can be carried out using easily available highly-optimized mathematical software packages, resulting in improvements in computational speed. Savings of up to a factor of 3 in the computer time have been observed during tests on a Pentium 4 computational platform workstation.
Keywords :
channel bank filters; digital filters; frequency response; matrix algebra; buffer memory size reduction; fast filter banks; frequency response masking principle; frequency-selective channels; matrix computations; matrix operations; real-time FFB software implementations; single-rate filter bank; tree-structured filter banks; Channel bank filters; Computer buffers; Filter bank; Frequency response; Hardware; Packaging; Software design; Software packages; Testing; Workstations;
Conference_Titel :
Acoustics, Speech, and Signal Processing, 2004. Proceedings. (ICASSP '04). IEEE International Conference on
Print_ISBN :
0-7803-8484-9
DOI :
10.1109/ICASSP.2004.1327065