Selection of decoys for non-vocabulary utterances rejection

• US 6,195,634 B1
• Filed: 12/24/1997
• Issued: 02/27/2001
• Est. Priority Date: 12/24/1997
• Status: Expired due to Fees

• Current Assignee: Nortel Networks Limited (Nortel Networks Corporation)
• Original Assignee: Nortel Networks Corporation
• Inventors: Pelletier, Claude, Dudemaine, Martin
• Primary Examiner(s): Dorvil, Richemond
• Assistant Examiner(s): Azad, Abul K.
• Application Number:
  US08/997,823
• Time in Patent Office: 1,161 Days
• Field of Search: 704/231, 704/246, 704/247, 704/248, 704/249, 704/250, 704/270, 704/256, 704/240, 704/251, 704/243, 704/257, 704/239
• US Class Current: 704/231
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams
• CPC Class Codes: G10L 2015/0631 Creating reference templates; Clustering
• CPC Class Codes: G10L 2015/0638 Interactive procedures

Abstract

Assessing decoys for use in an audio recognition process for identifying predetermined sounds in an unknown input audio signal, involves a test recognition process for matching known training audio signals to models representing the predetermined sounds and the decoys and determining for each of the decoys, from the results of the test recognition process, a score representing the effect of the respective decoy in the recognition of any of the known training audio signals. An advantage arising from generating scores for decoys is that the chance of a poor selection of decoys can be reduced. Thus the possibility of poor recognition performance arising from poorly selected decoys can be reduced. Furthermore, the requirement for expert input into the decoy creation process, which may be time consuming, can be reduced. This can make it easier, or quicker, or less expensive to install.

Claims

1. A method of assessing decoys for use in an audio recognition process for identifying predetermined sounds in an unknown input audio signal, the method comprising the steps of:

carrying out a test recognition process by matching known training audio signals to models representing the predetermined sounds and the decoys; and

determining for each of the decoys, from the results of the test recognition process, a score representing the effect of the respective decoy in the recognition of any of the known training audio signals,

wherein the known training audio signals comprise known non-vocabulary utterances, and the score additionally represents the effect of the respective decoy on the rejection of any of the non-vocabulary utterances.

2. The method of claim 1 further comprising the preliminary step of generating at least some of the decoys from a random combination of component sounds.

3. The method of claim 1 further comprising the steps of determining for the respective decoy, further scores relating to further ones of the known training audio signals, and determining an accumulated score from the score, and the further scores.

4. The method of claim 1 further comprising the step of selecting which decoys to use according to the determined scores.

5. The method of claim 1 wherein the audio recognition process comprises a speech recognition process, and the predetermined sounds comprise words.

6. The method of claim 1 wherein the step of determining the score comprises the steps of:

determining whether the respective decoy is a closer match to a given one of the known training audio signals than the best matching predetermined sound; and

determining the score according to the result.

7. The method of claim 6 wherein the step of determining the score comprises the steps of:

determining how many other decoys are a closer match to the given one of the known training audio signals than the best matching predetermined sound; and

determining the score according to the result.

8. The method of claim 6 wherein the step of determining the score comprises the steps of:

determining how close a match the respective decoy is relative to any other decoys which are a closer match to the given one of the known training audio signals than the best matching predetermined sound; and

determining the score according to the result.

9. A method of audio recognition for identifying predetermined sounds in an unknown input audio signal, using decoys having associated scores determined by carrying out a trial recognition process by matching known training audio signals to models representing the predetermined sounds and the decoys, the scores representing the effect of a respective decoy in the recognition of any of the known training audio signals, the method comprising the step of:

performing the audio recognition process for identifying predetermined sounds in an unknown input audio signal by matching the unknown input audio signal to models of the predetermined sounds and the decoys, according to the scores associated with the decoys, wherein the matching is carried out using decoys weighted according to their score.