New parallel implementations for DTW algorithms
Authors:
- G.N. StainhaouerGSG.N. Stainhaouer
- G. CarayannisGCG. Carayannis
- 1 April 1990
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Acoustics, Speech, and Signal Processing
- Vol. 38 (4) , 705-711
Abstract
The parallel realization of a popular dynamic time warping (DTW) algorithm is discussed. Two alternative techniques are proposed, one based on a circular array and the other using a linear array of processing elements (PEs). The architecture of each PE is defined in both cases and computational phases are outlined. The number of PEs is not restricted to be fixed. With small modifications, both of the techniques can implement DTW with any number of PEs available. The performance of the new architectures is superior to that of architectures previously reported in the literatureKeywords
This publication has 7 references indexed in Scilit:
- A digit pipelined dynamic time warp processor (word recognition)IEEE Transactions on Acoustics, Speech, and Signal Processing, 1988
- Parallel partitioning techniques for the DTW algorithm in speech recognition1987
- Systolic architectures for connected speech recognitionIEEE Transactions on Acoustics, Speech, and Signal Processing, 1986
- Array configurations for dynamic time warpingIEEE Transactions on Acoustics, Speech, and Signal Processing, 1984
- Dynamic Time Warp Pattern Matching Using an Integrated Multiprocessing ArrayIEEE Transactions on Computers, 1983
- Dynamic programming algorithm optimization for spoken word recognitionIEEE Transactions on Acoustics, Speech, and Signal Processing, 1978
- Minimum prediction residual principle applied to speech recognitionIEEE Transactions on Acoustics, Speech, and Signal Processing, 1975