Title :
Fast algorithm for pyramid vector quantization
Author :
J. Mohorko;P. Planinsic; Zarko CuCej
Author_Institution :
Fac. of Electr. Eng. & Comput. Sci., Maribor Univ., Slovenia
Abstract :
This paper presents an efficient implementation of a pyramid vector quantization coding/decoding algorithm developed by Filip and Ruf (see Proc. IEEE Conf. GLOBECOM, Orlando, FL, USA, p.240-4, 1992), where the original classical calculation of pyramid numbers V/sub L,K/ is replaced by a sliding calculation. The sliding calculation allows movement in any direction of a lattice so that the calculation of V/sub L,K/ in each algorithm iteration can be continued from the previous pyramid number. These properties can, under certain conditions, significantly (up to 25 times) accelerate the coding/decoding process. The condition under which the sliding calculation is more effective than the original one is also presented. The simulation results confirm the expected improvement over the original algorithm.
Keywords :
"Vector quantization","Lattices","Iterative decoding","Signal processing algorithms","Acceleration","Performance gain","Performance loss","Measurement","Audio coding","Pattern recognition"
Journal_Title :
IEEE Signal Processing Letters