Method and apparatus for continuous speech recognition using a layered, self-adjusting decoder network

• US 6,006,181 A
• Filed: 09/12/1997
• Issued: 12/21/1999
• Est. Priority Date: 09/12/1997
• Status: Expired due to Term

• Current Assignee: WSOU Investments, LLC (WSOU Holdings, LLC)
• Original Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Inventors: Chou, Wu, Buhrke, Eric Rolse
• Primary Examiner(s): Hudspeth, David R.
• Assistant Examiner(s): SAX, ROBERT L
• Application Number:
  US08/928,372
• Time in Patent Office: 830 Days
• Field of Search: 704/238, 704/240, 704/242, 704/255, 704/256
• US Class Current: 704/231
• CPC Class Codes: G10L 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications

Abstract

A continuous speech decoder that is built up of multiple layers. Each of the layers uses independent knowledge sources and rules, but all the layers cooperate to quickly decode the speech input into words. A first layer is concerned with acoustic data, a second layer with phone data of speech and a third layer concerns word data and word sequences. By separating these layers, the higher layers can be made time independent and asynchronous. Thus the asynchronous layers can process data quickly and give fast support to the first layer which keeps a dynamic record called a dynamic network of the most likely continuous speech results. The speed and separation of this decoder allows better memory efficiency and better decoder results compared to previously known continuous speech decoders.

Claims

1. A method for continuous speech recognition, comprising the steps of:

calculating a soft beam based on existing arcs;

predicting a new arc before any new arc is created;

said predicting step includes calculating by means of an arc predication process a path score entering the first state of each respective arc before allocating memory space for each new arc;

comparing each calculated path score with a present soft beam;

if a calculated path score is within the present soft beam, creating the predicted arc; and

updating the soft beam.

2. The method of claim 1, wherein the updated soft beam cut-off score is lower than the cut-off score determined by the standard beam used in beam searching.

3. The method of claim 1, wherein searching with the updated soft beam causes a mortality rate of newly created arcs to be reduced and an efficiency of a dynamic decoder using the updated soft beam for searching to be improved over a dynamic decoder using standard beam searching.