Title:
INFERENCE PROGRAM, INFERENCE DEVICE, AND INFERENCE METHOD OF HIDDEN MARKOV MODEL
Document Type and Number:
WIPO Patent Application WO/2012/008184
Kind Code:
A1
Abstract:
Disclosed is a hidden Markov model inference algorithm that can calculate unknown parameters even outside specialized conditions. An inference means (14) is provided with: an update/setting means (24) that, as the unknown parameters of a hidden Markov model, updates/sets the probabilities that are a state transition probability (a), an output probability (b), an initial state probability (π) and a likelihood (P (y|θ)), and the expected value (Na) of state transition and the expected value (Nb) of output; and a computation means (26) that calculates new probabilities and expected values using infinitesimal approximation by means of Taylor expansion, using not only the probabilities and expected values immediately before updating/setting by means of the update/setting means (24), but also the previous time-shifted probabilities and expected values, and using observation data read from a register (1) and a speed-up parameter value (β) set by an initial setting means (22).
More Like This:
Inventors:
MATSUYAMA YASUO (JP)
HAYASHI RYUNOSUKE (JP)
HAYASHI RYUNOSUKE (JP)
Application Number:
PCT/JP2011/058312
Publication Date:
January 19, 2012
Filing Date:
March 31, 2011
Export Citation:
Assignee:
UNIV WASEDA (JP)
MATSUYAMA YASUO (JP)
HAYASHI RYUNOSUKE (JP)
MATSUYAMA YASUO (JP)
HAYASHI RYUNOSUKE (JP)
International Classes:
G10L15/06; G10L15/14
Foreign References:
| JP2006285881A | 2006-10-19 |
Other References:
MATSUYAMA Y. ET AL.: "Alpha-EM gives fast Hidden Markov Model estimation: Derivation and evaluation of alpha-HMM, Neural Networks (IJCNN)", THE 2010 INTERNATIONAL JOINT CONFERENCE ON, July 2010 (2010-07-01), pages 1 - 8
MOLINA C.: "Maximum Entropy-Based Reinforcement Learning Using a Confidence Measure in Speech Recognition for Telephone Speech, Audio, Speech, and Language Processing", IEEE TRANSACTIONS ON, July 2010 (2010-07-01), pages 1041 - 1052
MATSUYAMA Y.: "The a-EM algorithm: surrogate likelihood maximization using a-logarithmic information measures", IEEE TRANSACTIONS ON INFORMATION THEORY, March 2003 (2003-03-01), pages 692 - 706
MATSUYAMA Y. ET AL.: "Fast learning by the a-ECME algorithm, Neural Information Processing", PROCEEDINGS. ICONIP '99. 6TH INTERNATIONAL CONFERENCE, 1999, pages 1184 - 1190
HIROYUKI SHIOYA ET AL.: "Transformation of Convex Functions and Inequalities of Divergences", THE IEICE TRANSACTIONS ON FUNDAMENTALS OF ELECTRONICS, COMMUNICATIONS AND COMPUTER SCIENCES (JAPANESE EDITION). A, March 1997 (1997-03-01), pages 509 - 515
MOLINA C.: "Maximum Entropy-Based Reinforcement Learning Using a Confidence Measure in Speech Recognition for Telephone Speech, Audio, Speech, and Language Processing", IEEE TRANSACTIONS ON, July 2010 (2010-07-01), pages 1041 - 1052
MATSUYAMA Y.: "The a-EM algorithm: surrogate likelihood maximization using a-logarithmic information measures", IEEE TRANSACTIONS ON INFORMATION THEORY, March 2003 (2003-03-01), pages 692 - 706
MATSUYAMA Y. ET AL.: "Fast learning by the a-ECME algorithm, Neural Information Processing", PROCEEDINGS. ICONIP '99. 6TH INTERNATIONAL CONFERENCE, 1999, pages 1184 - 1190
HIROYUKI SHIOYA ET AL.: "Transformation of Convex Functions and Inequalities of Divergences", THE IEICE TRANSACTIONS ON FUNDAMENTALS OF ELECTRONICS, COMMUNICATIONS AND COMPUTER SCIENCES (JAPANESE EDITION). A, March 1997 (1997-03-01), pages 509 - 515
Attorney, Agent or Firm:
USHIKI, Mamoru (JP)
牛木 護 (JP)
牛木 護 (JP)
Download PDF:
Claims:
Previous Patent: LAMINATED BODY AND PNEUMATIC TYRE USING SAME
Next Patent: SURFACE COATING CUTTING TOOL
Next Patent: SURFACE COATING CUTTING TOOL