METHOD AND APPARATUS FOR RECOGNIZING A SPEAKER IN LAWFUL INTERCEPTION SYSTEMS
First Claim
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 Assignments
0 Petitions
Accused Products
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.
78 Citations
34 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. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14)
-
-
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. - View Dependent Claims (16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33)
-
-
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.
-
Specification