A VLSI Architecture with Multiple Fast Store-Based Block Parallel Processing for Output Probability and Likelihood Score Computations in HMM-Based Isolated Word Recognition

Kazuhiro NAKAMURA; Ryo SHIMAZAKI; Masatoshi YAMAMOTO; Kazuyoshi TAKAGI; Naofumi TAKAGI

doi:10.1587/transele.E95.C.456

A VLSI Architecture with Multiple Fast Store-Based Block Parallel Processing for Output Probability and Likelihood Score Computations in HMM-Based Isolated Word Recognition

Kazuhiro NAKAMURA, Ryo SHIMAZAKI, Masatoshi YAMAMOTO, Kazuyoshi TAKAGI, Naofumi TAKAGI

Full Text Views

0

Share
Cite this

Summary :

This paper presents a memory-efficient VLSI architecture for output probability computations (OPCs) of continuous hidden Markov models (HMMs) and likelihood score computations (LSCs). These computations are the most time consuming part of HMM-based isolated word recognition systems. We demonstrate multiple fast store-based block parallel processing (MultipleFastStoreBPP) for OPCs and LSCs and present a VLSI architecture that supports it. Compared with conventional fast store-based block parallel processing (FastStoreBPP) and stream-based block parallel processing (StreamBPP) architectures, the proposed architecture requires fewer registers and less processing time. The processing elements (PEs) used in the FastStoreBPP and StreamBPP architectures are identical to those used in the MultipleFastStoreBPP architecture. From a VLSI architectural viewpoint, a comparison shows that the proposed architecture is an improvement over the others, through efficient use of PEs and registers for storing input feature vectors.

Publication: IEICE TRANSACTIONS on Electronics Vol.E95-C No.4 pp.456-467

Publication Date: 2012/04/01

Publicized

Online ISSN: 1745-1353

DOI: 10.1587/transele.E95.C.456

Type of Manuscript: Special Section PAPER (Special Section on Solid-State Circuit Design – Architecture, Circuit, Device and Design Methodology)

Category

Cite this

Copy

Kazuhiro NAKAMURA, Ryo SHIMAZAKI, Masatoshi YAMAMOTO, Kazuyoshi TAKAGI, Naofumi TAKAGI, "A VLSI Architecture with Multiple Fast Store-Based Block Parallel Processing for Output Probability and Likelihood Score Computations in HMM-Based Isolated Word Recognition" in IEICE TRANSACTIONS on Electronics, vol. E95-C, no. 4, pp. 456-467, April 2012, doi: 10.1587/transele.E95.C.456.
Abstract: This paper presents a memory-efficient VLSI architecture for output probability computations (OPCs) of continuous hidden Markov models (HMMs) and likelihood score computations (LSCs). These computations are the most time consuming part of HMM-based isolated word recognition systems. We demonstrate multiple fast store-based block parallel processing (MultipleFastStoreBPP) for OPCs and LSCs and present a VLSI architecture that supports it. Compared with conventional fast store-based block parallel processing (FastStoreBPP) and stream-based block parallel processing (StreamBPP) architectures, the proposed architecture requires fewer registers and less processing time. The processing elements (PEs) used in the FastStoreBPP and StreamBPP architectures are identical to those used in the MultipleFastStoreBPP architecture. From a VLSI architectural viewpoint, a comparison shows that the proposed architecture is an improvement over the others, through efficient use of PEs and registers for storing input feature vectors.
URL: https://globals.ieice.org/en_transactions/electronics/10.1587/transele.E95.C.456/_p

Copy

@ARTICLE{e95-c_4_456,
author={Kazuhiro NAKAMURA, Ryo SHIMAZAKI, Masatoshi YAMAMOTO, Kazuyoshi TAKAGI, Naofumi TAKAGI, },
journal={IEICE TRANSACTIONS on Electronics},
title={A VLSI Architecture with Multiple Fast Store-Based Block Parallel Processing for Output Probability and Likelihood Score Computations in HMM-Based Isolated Word Recognition},
year={2012},
volume={E95-C},
number={4},
pages={456-467},
abstract={This paper presents a memory-efficient VLSI architecture for output probability computations (OPCs) of continuous hidden Markov models (HMMs) and likelihood score computations (LSCs). These computations are the most time consuming part of HMM-based isolated word recognition systems. We demonstrate multiple fast store-based block parallel processing (MultipleFastStoreBPP) for OPCs and LSCs and present a VLSI architecture that supports it. Compared with conventional fast store-based block parallel processing (FastStoreBPP) and stream-based block parallel processing (StreamBPP) architectures, the proposed architecture requires fewer registers and less processing time. The processing elements (PEs) used in the FastStoreBPP and StreamBPP architectures are identical to those used in the MultipleFastStoreBPP architecture. From a VLSI architectural viewpoint, a comparison shows that the proposed architecture is an improvement over the others, through efficient use of PEs and registers for storing input feature vectors.},
keywords={},
doi={10.1587/transele.E95.C.456},
ISSN={1745-1353},
month={April},}

Copy

TY - JOUR
TI - A VLSI Architecture with Multiple Fast Store-Based Block Parallel Processing for Output Probability and Likelihood Score Computations in HMM-Based Isolated Word Recognition
T2 - IEICE TRANSACTIONS on Electronics
SP - 456
EP - 467
AU - Kazuhiro NAKAMURA
AU - Ryo SHIMAZAKI
AU - Masatoshi YAMAMOTO
AU - Kazuyoshi TAKAGI
AU - Naofumi TAKAGI
PY - 2012
DO - 10.1587/transele.E95.C.456
JO - IEICE TRANSACTIONS on Electronics
SN - 1745-1353
VL - E95-C
IS - 4
JA - IEICE TRANSACTIONS on Electronics
Y1 - April 2012
AB - This paper presents a memory-efficient VLSI architecture for output probability computations (OPCs) of continuous hidden Markov models (HMMs) and likelihood score computations (LSCs). These computations are the most time consuming part of HMM-based isolated word recognition systems. We demonstrate multiple fast store-based block parallel processing (MultipleFastStoreBPP) for OPCs and LSCs and present a VLSI architecture that supports it. Compared with conventional fast store-based block parallel processing (FastStoreBPP) and stream-based block parallel processing (StreamBPP) architectures, the proposed architecture requires fewer registers and less processing time. The processing elements (PEs) used in the FastStoreBPP and StreamBPP architectures are identical to those used in the MultipleFastStoreBPP architecture. From a VLSI architectural viewpoint, a comparison shows that the proposed architecture is an improvement over the others, through efficient use of PEs and registers for storing input feature vectors.
ER -