METHOD AND APPARATUS FOR RECOGNIZING A SPEAKER IN LAWFUL INTERCEPTION SYSTEMS

• US 20090043573A1
• Filed: 08/09/2007
• Published: 02/12/2009
• Est. Priority Date: 08/09/2007
• Status: Active Grant

• Current Assignee: Cyberbit Ltd.
• Original Assignee: Nice Systems Limited (Nice Ltd)
• Inventors: GUTMAN, Renan, OPHER, Irit, WEINBERG, Adam, BENAROYA, Eyal
• Granted Patent: US 8,219,404 B2
• Time in Patent Office: Days
• US Class Current: 704/223
• CPC Class Codes: G10L 17/06 Decision making techniques; Pattern matching strategies
• CPC Class Codes: H04L 63/302 gathering intelligence information for situation awareness or reconnaissance

Abstract

A method and apparatus for identifying a speaker within a captured audio signal from a collection of known speakers. The method and apparatus receive or generate voice representations for each known speakers and tag the representations according to meta data related to the known speaker or to the voice. The representations are grouped into one or more groups according to the indices. When a voice to be recognized is introduced, characteristics are determined according to which the groups are prioritized, so that the representations participating only in part of the groups are matched against the o voice to be identified, thus reducing identification time and improving the statistical significance.

Claims

1. A method for associating a voice of a first speaker, the voice extracted from a captured audio signal, with at least one of a multiplicity of speakers, each of the multiplicity of speakers associated with an acoustic model and with data, the method comprising the steps of:

receiving or extracting the data associated with each of the multiplicity of speakers;

tagging the acoustic model associated with each of the multiplicity of speakers according to an at least one parameter associated with the acoustic model or with a second voice sample the acoustic model is associated with or with a speaker of the second voice sample;

constructing according to the tagging an at least one group comprising an acoustic model;

determining an at least one matched group to be matched against the voice of the first speaker;

determining an at least one non-acoustic score between data related to the first speaker, and the at least one matched group or an at least one acoustic model from the at least one matched group;

determining an at least one acoustic score between the voice of the first speaker and an at least one acoustic model from the at least one matched group;

obtaining a total score by combining the non-acoustic score with the acoustic score;

determining according to the total score whether an identification criteria was met; and

if the identification criteria was met, associating the first speaker with the at least one model from the matched group.

2. The method of claim 1 further comprising the step of constructing the acoustic model.

3. The method of claim 1 further comprising the step of determining a relative order between the at least one group and an at least one second group.

4. The method of claim 1 wherein the data associated with each of the multiplicity of speakers is meta data related to any at least one of the multiplicity of speakers.

5. The method of claim 1 wherein the data associated with each of the multiplicity of speakers relates to the acoustic model.

6. The method of claim 1 wherein the at least one parameter relates to the acoustic model.

7. The method of claim 1 wherein the at least one parameter relates to data associated with an at least one of the multiplicity of speakers.

8. The method of claim 1 wherein the step of constructing the at least one group is performed by considering the models or the associated data.

9. The method of claim 1 wherein the tagging is performed according to a level of connectivity between a second speaker the first speaker was communicating with, and an at least one of the multiplicity of speakers.

10. The method of claim 1 wherein the tagging is performed according to a time of communication between a second speaker the first speaker was communicating with, and an at least one of the multiplicity of speakers.

11. The method of claim 1 wherein the tagging is performed according to a predetermined group of speakers.

12. The method of claim 1 wherein the at least one parameter relates to any one or more of the group consisting of:

identity of a speaker in the second voice sample;

age of a speaker in the second voice sample;

accent of a speaker in the second voice sample;

language spoken by a speaker in the second voice sample;

a feature of the at least one voice model;

data extracted from the second voice sample;

level of connectivity between a speaker in the second voice sample and another speaker;

an at least one word used by a speaker in the second voice sample;

an at least one name mentioned by a speaker;

a location associated with a speaker in the second voice sample;

a phone number or part thereof associated with a speaker in the second voice sample;

a pronunciation of an at least one phoneme by a speaker in the second voice sample;

a characteristic of a channel used by a speaker in the second voice sample; and

a time of an at least one communication of a speaker in the second voice sample.

13. The method of claim 1 wherein the data related to the first speaker relates to any one or more of the group consisting of:

identity of the first speaker;

age of the first speaker;

accent of the first speaker;

language spoken by the first speaker;

a characteristic of the at least one voice model;

data extracted from the voice sample;

level of connectivity between the first speaker and a second speaker the first speaker was communicating with;

an at least one word used by the first speaker;

an at least one name mentioned by the first speaker;

a location associated with the first speaker;

a phone number or part thereof associated with first speaker;

a pronunciation of one or more phonemes by the first speaker;

a characteristic of a channel used by the first speaker; and

a time of an at least one communication of the first speaker.

14. The method of claim 1 wherein the audio signal is in a format selected from the group consisting of:

PCM, a-law, mu-law, GSM, CDMA, TDMA, ADPCM and VOIP.

15. An apparatus for associating a voice of a first speaker, the voice extracted from a captured audio signal, with at least one of a multiplicity of speakers, each of the multiplicity of speakers associated with an acoustic model and with data, the apparatus comprising:

a storage device for storing the acoustic model and associated meta data;

a capturing or logging component for receiving a voice sample of the first speaker to be identified;

a tagging component for tagging the acoustic model according to an at least one parameter associated with the acoustic model or with a second voice sample the acoustic model is associated with or with a speaker of the second voice sample;

a selection component for selecting a matched group comprising an at least one matched model or an at least one model for matching with the voice sample of the first speaker to be identified;

a non-acoustic score determination component, for determining a non-acoustic score between data related to the first speaker, and the at least one matched group or an at least one acoustic model from the at least one matched group;

an acoustic score determination component for determining an acoustic score between the voice of the first speaker and an at least one acoustic model from the at least one matched group;

a combining component for combining the acoustic score and the non-acoustic score into a total score; and

a criteria evaluation component for determining whether the total score meets an at least one criteria.

16. The apparatus of claim 15 further comprising a group determination order for determining a matching order between at least two groups.

17. The apparatus of claim 15 further comprising a model determination order for determining a matching order between at least two models belonging to the same group.

18. The apparatus of claim 15 further comprising a model generation component for generating an acoustic model from a voice sample.

19. The apparatus of claim 15 further comprising a data extraction component for extracting data related to a voice sample or to a speaker thereof.

20. The apparatus of claim 15 further comprising an alert generation device for generating an alert when the first speaker is identified as at least one of the multiplicity of speakers.

21. The apparatus of claim 15 further comprising a result reporting component for reporting a result related to matching the first speaker and the at least one matched model.

22. The apparatus of claim 15 wherein the data associated with each of the multiplicity of speakers is meta data related to any at least one of the multiplicity of speakers.

23. The apparatus of claim 15 wherein the data associated with each of the multiplicity of speakers relates to the acoustic model.

24. The apparatus of claim 15 wherein the at least one parameter relates to the acoustic model.

25. The apparatus of claim 15 wherein the at least one parameter relates to data associated with any at least one of the multiplicity of speakers.

26. The apparatus of claim 15 wherein the captured audio signal represents any one or more items selected from the group consisting of:

a phone conversation;

a voice over IP conversation;

an audio part of a video conference;

a radio broadcast;

an audio part of a television broadcast; and

a captured microphone.

27. The apparatus of claim 15 wherein the captured audio signal is in a format selected from the group consisting of:

PCM, a-law, mu-law, GSM, CDMA, TDMA, ADPCM and VOIP.

28. The apparatus of claim 15 wherein the associated meta data relates to an at least one level of connectivity between a second speaker the first speaker was communicating with and an at least one speaker associated with the voice models.

29. The apparatus of claim 15 wherein the at least one parameter relates to any one or more of the group consisting of:

identity of the speaker of the second voice sample;

age of the speaker of the second voice sample;

accent of the speaker of the second voice sample;

language spoken by the speaker of the second voice sample;

a feature of the at least one voice model;

data extracted from the voice sample;

level of connectivity between the speaker of the second voice sample and a second speaker the speaker of the second voice sample was communicating with;

one or more words used by the speaker of the second voice sample;

one or more to names mentioned by the speaker of the second voice sample;

a location associated with the speaker of the second voice sample;

a phone number or part thereof associated with a speaker of the second voice sample;

a pronunciation of one or more phonemes by a speaker of the second voice sample;

a characteristic of a channel used by speaker of the second voice sample; and

a time of an at least one communication of a speaker of the second voice sample.

30. The apparatus of claim 15 wherein the data related to the first speaker relates to any one or more of the group consisting of:

identity of the first speaker;

age of the first speaker;

accent of the first speaker;

language spoken by the first speaker;

a feature of the at least one voice model;

data extracted from the second voice sample;

level of connectivity between the first speaker and another speaker;

an at least one word used by the first speaker;

an at least one name mentioned by the first speaker;

a location associated with the first speaker;

a phone number or part thereof associated with the first speaker;

a pronunciation of an at least one phoneme by the first speaker;

a characteristic of a channel used by the first speaker; and

a time of an at least one communication of the first speaker.

31. The apparatus of claim 15 wherein the tagging is performed according to a level of connectivity between a second speaker the first speaker was communicating with, and an at least one of the multiplicity of speakers.

32. The apparatus of claim 15 wherein the tagging is performed according to a time of communication between a second speaker the first speaker was communicating with, and an at least one of the multiplicity of speakers.

33. The apparatus of claim 15 wherein the tagging is performed according to a predetermined group of speakers.

34. A method for associating a voice of a first speaker, the voice extracted from a captured audio signal, with an at least one of a multiplicity of speakers, each of the multiplicity of speakers associated with an acoustic model and with meta data, the method comprising the steps of:

constructing an at least one group of models, each one of the group of models comprising the acoustic model and the meta data associated with one of a multiplicity of speakers;

matching the voice of the first speaker with all models belonging to the at least one group of models to obtain a score; and

associating the first speaker as a speaker associated with one of the multiplicity of speakers for which the score meets a predetermined criteria.