Telecommunications instrument employing variable criteria speech recognition

• US 5,842,161 A
• Filed: 06/25/1996
• Issued: 11/24/1998
• Est. Priority Date: 06/25/1996
• Status: Expired due to Term

• Current Assignee: Nokia of America Corporation (Nokia Corporation)
• Original Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Inventors: Keen, Ellen Anne, Keen, Donald Marion, Deldar, Mitra P., Cohrs, Paul Wesley
• Primary Examiner(s): Dorvil, Richemond
• Application Number:
  US08/668,660
• Time in Patent Office: 882 Days
• Field of Search: 395/2.4, 395/2.53, 395/2.54, 395/2.55, 395/2.56, 395/2.6, 395/2.61, 395/2.84, 395/2.79, 704/231, 704/244, 704/245, 704/246, 704/251, 704/250, 704/252, 704/255, 704/256, 704/257, 704/233, 704/243, 704/200, 704/241, 704/239, 704/275
• US Class Current: 704/251
• CPC Class Codes: G10L 15/10 using distance or distortion measures between unknown speech and reference templates
• CPC Class Codes: G10L 2015/0631 Creating reference templates; Clustering

Abstract

A recognition criterion or set of recognition criteria are updated automatically, over time, in accordance with the speech input of the user(s). Each input utterance is compared to one or more models of speech to determine a similarity metric for each such comparison. A model of speech which most closely matches the utterance is determined based on the one or more similarity metrics. The similarity metric corresponding to the most closely matching model of speech is analyzed to determine whether the similarity metric satisfies the selected set of recognition criteria. The recognition criteria are automatically altered during use or "on-the-fly", so that more appropriate criteria (and associated thresholds) may be used to either increase the probability of recognition or decrease the incidence of false positive results. Illustratively, if a voice sample results in a near miss of a template, a more liberal criterion is thereafter employed to increase the probability of recognition for subsequent input. Parametric histories of recognition and near misses followed by recognition are kept with periodic alteration of the criteria values to correspond to these histories. Additionally, parametric histories of false alarms are maintained and used to update criteria values in combination with recognition histories.

Claims

1. A speech recognizer apparatus for recognizing a phrase including at least one word, based upon an utterance by a user, the apparatus comprising:

means for comparing the utterance to at least one speech model to determine at least one similarity metric, wherein said at least one similarity metric corresponds to said at least one speech model;

means for selecting a particular model of speech from said at least one speech model, that most closely matches the utterance based on said at least one similarity metric;

means, responsive to the means for selecting, for determining whether the similarity metric corresponding to said most closely matching model of speech satisfies at least one recognition criterion;

means for recognizing the utterance as a phrase corresponding to said most closely matching model of speech when said at least one recognition criterion is satisfied;

means for storing voice data representative of the utterance, when said at least one recognition criterion is satisfied; and

means for automatically adjusting said at least one recognition criterion in accordance with said stored voice data, to thereby increase a probability of recognition for the utterance upon subsequent repetition thereof.

2. The apparatus of claim 1, wherein a model of speech reflects one or more predetermined words.

3. The apparatus of claim 2, wherein a predetermined word comprises a command word for a utilization device.

4. The apparatus of claim 2, further including a utilization device.

5. The apparatus of claim 4, wherein the utilization device is a telephone.

6. The apparatus of claim 4, wherein the utilization device is an answering machine.