Assisted discrimination of similar sounding speakers
View Patent ↗A communications system is provided that includes: (a) a speech discrimination agent 136 operable to generate a speech profile of a first party to a voice call; and (b) a speech modification agent 140 operable to adjust, based on the speech profile, a spectral characteristic of a voice stream from the first party to form a modified voice stream, the modified voice stream being provided to the second party.
1. A method of managing a conference call including at least three conference call participants, comprising:
during a multiparty conference call, a processor receiving from a second conference call participant a request to discriminate between first and third conference call participants, wherein the request identifies the first and third participants;
generating a profile associated with each conference call participant;
during the conference call, refining each profile based on one or more of speech profiling and prosodic analysis;
in response to the request, automatically comparing the profiles for the first and third conference call participants to determine similarities in one or more spectral characteristics;
in response to the comparison, automatically isolating, by the processor, a first spectral characteristic of a received voice stream of one of the first and third conference call participants to form a modified voice stream of the one of the first and third conference call participants;
adjusting the modified voice stream by one or more suitable speech modifiers determined by a speech modification agent; and
providing, by the processor, the modified voice stream to a communication device for audible presentation to the second conference call participant.
2. The method of claim 1 , wherein the received voice stream is received from the first conference call participant, wherein the received voice stream is provided to the third conference call participants, and wherein at least a first spectral characteristic of the received and modified voice streams differs.
3. The method of claim 1 , wherein a common voice stream is received from a common communication device used by the first and third conference call participants, the common voice stream comprising first and third voice streams from the first and third conference call participants, respectively.
4. The method of claim 1 , wherein first and third speech profiles are generated for the first and third conference call participants, wherein the first and third speech profiles are a pitch profile, and wherein the spectral characteristic is at least one of a magnitude of a pitch value, a shape of the pitch distribution, and a location of the pitch distribution.
5. The method of claim 4 , wherein the spectral characteristic is the magnitude of the pitch value and wherein the pitch value of the modified voice stream is either decreased or increased relative to the received voice stream.
6. The method of claim 4 , wherein the spectral characteristic is a shape of the pitch distribution and wherein, in the profile distribution of the modified voice stream, at least one of the median pitch value, mean pitch value, maximum and minimum frequency values, and standard deviation is different from a corresponding at least one of the median pitch value, mean pitch value, maximum and minimum frequency values, and standard deviation of the pitch distribution of the received voice stream.
7. The method of claim 1 , wherein the adjusting step comprises the substeps of:
decomposing the received voice stream into a plurality of subcomponents;
adjusting at least one of the subcomponents; and
recombining the subcomponents to form the modified voice stream, wherein, in the modified voice stream, the adjusted at least one subcomponent replaces the subcomponent from which the adjusted at least one subcomponent was derived.
8. A non-transitory computer readable medium comprising processor executable instructions that, when executed, perform the steps of claim 1 .
9. A system that manages a conference call including at least three conference call participants, comprising:
a processor operable to:
during a multiparty conference call, receive from a second conference call participant, a request to discriminate between first and third conference call participants, wherein the request identifies the first and third participants;
generate a profile associated with each conference call participant;
refine, during the conference call, each profile based on one or more of speech profiling and prosodic analysis;
in response to the request, automatically compare the profiles for the first and third conference call participants;
determine one or more similar spectral characteristics in the profiles for the first and third conference call participants;
execute a speech modification agent, wherein the speech modification agent is operable to determine one or more suitable speech modifiers to modify the one or more similar spectral characteristics in the profiles for the first and third conference call participants;
in response to the determination of the one or more suitable speech modifiers, automatically adjust a first spectral characteristic of a received voice stream of a selected one of the first and third conference call participants to form a modified voice stream with one or more suitable speech modifiers; and
provide the modified voice stream to a communication device for audible presentation to the second conference call participant.
10. The system of claim 9 , wherein the received voice stream is received from the first conference call participant, wherein the received voice stream is provided to the third conference call participants, and wherein at least a first spectral characteristic of the received and modified voice streams differs.
11. The system of claim 9 , wherein a common voice stream is received from a common communication device used by the first and third conference call participants, the common voice stream comprising first and third voice streams from the first and third conference call participants, respectively.
12. The system of claim 9 , wherein first and third speech profiles are generated for the first and third conference call participants, wherein the first and third speech profiles are a pitch profile, and wherein the spectral characteristic is at least one of a magnitude of a pitch value, a shape of the pitch distribution, and a location of the pitch distribution.
13. The system of claim 12 , wherein the spectral characteristic is the magnitude of the pitch value and wherein the pitch value of the modified voice stream is either decreased or increased relative to the received voice stream.
14. The system of claim 12 , wherein the spectral characteristic is a shape of the pitch distribution and wherein, in the profile distribution of the modified voice stream, at least one of the median pitch value, mean pitch value, maximum and minimum frequency values, and standard deviation is different from a corresponding at least one of the median pitch value, mean pitch value, maximum and minimum frequency values, and standard deviation of the pitch distribution of the received voice stream.
15. The system of claim 9 , wherein the adjusting operation comprises the sub-operations:
decompose the received voice stream into a plurality of subcomponents;
adjust at least one of the subcomponents; and
recombining the subcomponents to form the modified voice stream, wherein, in the modified voice stream, the adjusted at least one subcomponent replaces the subcomponent from which the adjusted at least one subcomponent was derived.
16. A method, comprising:
a processor receiving for a disadvantaged conference call participant a voice stream from a first conference call participant and at least one other conference call participant;
automatically creating, by the processor, a speech profile for each of the first conference call participant and the at least one other conference call participant based on one or more of speech profiling and prosodic analysis;
during the multiparty conference call, automatically refining, by the processor, the speech profiles for the first conference call participant and the at least one other conference call participant;
automatically comparing, by the processor, the speech profiles for the first conference call participant and the at least one other conference call participant
automatically determining, based on a comparison of the speech profiles, that the voice stream from the first conference call participant is similar to another conference call participant;
adjusting, by the processor, a spectral characteristic of the received voice stream to form a modified voice stream of the first conference call participant to eliminate the determined similarity; and
providing, by the processor, the modified voice stream to a communication device for audible presentation to the disadvantaged conference call participant while presenting, substantially simultaneously, the received voice stream, unmodified, to a third conference call participant.
17. The method of claim 16 , further comprising:
receiving a request to discriminate between the first and third conference call participants, wherein the request identifies the first and third participants.
18. The method of claim 16 , wherein the modified portion of the received voice stream is associated with interference.
19. The method of claim 17 , wherein first and third speech profiles are generated for the first and third conference call participants, wherein the first and third speech profiles are a pitch profile, and wherein the spectral characteristic is at least one of a magnitude of a pitch value, a shape of the pitch distribution, and a location of the pitch distribution.
20. The method of claim 16 , wherein the adjusting step comprises the substeps of:
decomposing the received voice stream into a plurality of subcomponents;
adjusting at least one of the subcomponents; and
recombining the subcomponents to form the modified voice stream, wherein, in the modified voice stream, the adjusted at least one subcomponent replaces the subcomponent from which the adjusted at least one subcomponent was derived.
21. A non-transitory computer readable comprising processor executable instructions that, when executed, perform the steps of claim 16 .
22. The method of claim 16 , wherein the first and a fourth conference call participant are in a common room and use a common communication device, wherein a common voice stream is received from the first and fourth conference call participants, the common voice stream comprising first and fourth voice streams from the first and fourth conference call participants, respectively, and further comprising:
identifying first and fourth voice components from the first and fourth voice streams, respectively, in the common voice stream, said identifying being by triangulation using a plurality of voice signals received from a plurality of microphones positioned at different locations in proximity to the first and fourth conference call participants to determine a physical location of each of the first and fourth conference call participants and, using the determined physical locations of the first and fourth participants, directional microphone techniques to distinguish the first and fourth voice streams from one another.