Speech recognition employing key word modeling and non-key word modeling

• US 5,649,057 A
• Filed: 01/16/1996
• Issued: 07/15/1997
• Est. Priority Date: 05/17/1989
• Status: Expired due to Term

• Current Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Original Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Inventors: Wilpon, Jay Gordon, Rabiner, Lawrence Richard, Lee, Chin-Hui
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): Chawan, Vijay B.
• Application Number:
  US08/586,413
• Time in Patent Office: 546 Days
• Field of Search: 395/2.4, 395/2.45, 395/2.54, 395/2.6-2.66, 381/41-45
• US Class Current: 704/256
• CPC Class Codes: G10L 15/142 Hidden Markov Models [HMMs]
• CPC Class Codes: G10L 2015/088 Word spotting

Abstract

Speaker independent recognition of small vocabularies, spoken over the long distance telephone network, is achieved using two types of models, one type for defined vocabulary words (e.g., collect, calling-card, person, third-number and operator), and one type for extraneous input which ranges from non-speech sounds to groups of non-vocabulary words (e.g. `I want to make a collect call please`). For this type of key word spotting, modifications are made to a connected word speech recognition algorithm based on state-transitional (hidden Markov) models which allow it to recognize words from a pre-defined vocabulary list spoken in an unconstrained fashion. Statistical models of both the actual vocabulary words and the extraneous speech and background noises are created. A syntax-driven connected word recognition system is then used to find the best sequence of extraneous input and vocabulary word models for matching the actual input speech.

Claims

1. A method for generating a sink model for use in a speech recognition system, said method comprising the steps of

obtaining a first extraneous sound training token,

obtaining a second extraneous sound training token, said first and second extraneous sound training tokens being other than repetitions of a particular one vocabulary item, and

generating said sink model in response to a plurality of extraneous sound training tokens which includes said first and second extraneous sound training tokens.

2. The method of claim 1 wherein said first and second extraneous sound training tokens are respective different vocabulary items.

3. The method of claim 1 wherein said first and second extraneous sound training tokens are a background sound and a vocabulary item, respectively.

4. The method of claim 3 wherein said background sound includes a silence component.

5. The method of any one of claims 1 through 4 comprising the further step of combining said sink speech recognition model with a plurality of key word speech recognition models into a grammar which defines expected sequences of keywords and extraneous sounds.

6. The invention of any one of claims 1 through 4 wherein each of said plurality of speech recognition models is a Hidden Markov Model.

7. A sink model stored in a storage medium generated by the method of any one of claims 1 through 4.