US 2003/0198195 A1Published application

Speaker tracking on a multi-core in a packet based conferencing system

AssigneeTexas Instruments Incorporated

InventorDunling Li

Life of this patent

Estimating the term…

Loading patent…

Abstract

A distributed intelligence conferencing system having a plurality of conferencing nodes to connect groups of participants to a conference. Each of the conferencing nodes provides for the connection of one or more participants to the conference. Each node includes a DSP for distributed signal processing. The nodes DSP includes: A signal measuring device for measuring features of the signals from each of the participants. A single core speaker tracking algorithm identifies the active speaker on a local node based upon the relative signal features of each of the number of participant input signals. A communication device communicates the measured signal features for the active local participant input signals only to all other conferencing nodes. A multi core speaker tracking algorithm compares the features of the signals from the speakers on the local node with the features of the signals from the speakers on other nodes. A final determination is made to identify the active speakers for the entire conference so that only the signals from a limited number of speakers transmitted across the system bus.

Claims16

Dependent claims

A multiple core speaker tracking system in a distributed conferencing system, comprising:

a system communication bus;
a plurality of nodes connected to said system communication bus, each of said plurality of nodes comprising:
a state machine for receiving a signal from each of a plurality of local participants connected to said node for classification of said signals as voice or noise and for producing a indicator of the classification of each of said signals;
feature extractors connected to said state machine for receiving said indicator and to said local participant signals for measuring and extracting at least one feature of each of said local participant signals classified as voice;
a local speaker tracker connected to said feature extractors for selecting a number of said local participant signals for tracking based upon said extracted features;
a first multiplexer connected to said speaker tracker, said local participant signals and said system communication bus, for passing said selected local participant signals and said extracted features to said system communication bus and for blocking the remaining local participant signals from said system communication bus;
a demultiplexer connected to said system communication bus for receiving selected and passed participant signals from other remote nodes connected to said system bus and for receiving said extracted features of each of said selected and passed signals from said other remote conferencing nodes;
a multi-core speaker tracker connected to said demultiplexer and to said local speaker tracker for comparing said extracted features of each of said significant signals of said local participant input signals and said remote participant input signals and providing a control signal to said first multiplexer to control the transmission of local significant signals to said system bus based upon said comparison;
and
a second multiplexer connected to said multi-core speaker tracker and to said first multiplexer to provide selected signals of said local and remote participants to said local participants.

7 more elements

A method for tracking speakers on a multi-core distributed conferencing system, comprising the steps of:

setting a maximum number of active speakers for the system;
receiving a plurality of extracted features corresponding to a plurality of signals from a plurality of cores on said system, wherein said features include:
determining the voice or noise state of each speaker on said single core;
assigning a valid identifier to each speaker signal on said single core which is identified as a voice signal;
extracting values for the short-term energies, long term energies, and energy variation of each speaker signal identified as voice;
detecting idle channels within said signals from said plurality of cores;
comparing the number of assigned valid speakers to the allowed maximum number of active speakers and designating the valid speakers as current active speakers when the number of valid speakers does not exceed the allowed maximum number of active speakers;
updating the parameters and features for the current active speakers based upon the features extracted from the valid speakers when the identity of the current active speakers is the same as the identity of the previous active speakers.

5 more elements

A method of determining which speakers to mute and which speakers to pass through to other participants on a multi-core packet based speaker conferencing system, comprising the steps of:

setting a maximum number n of active speakers;
identifying the n most active speakers on each core, including:
determining the voice or noise state of each speaker on said single core;
assigning a valid identifier to each speaker signal on said single core which is identified as a voice signal;
extracting values for the short-term energies, long term energies, standard energy and energy variation of each speaker signal identified as voice;
providing the extracted values for the n most active speakers from each core to each of the other cores on the system;
comparing the extracted values for the n most active speakers of all cores to determine the n most active speakers for the entire system;
muting all of said speakers except for said n most active speakers for the entire system.

5 more elements

Dependencies are inferred from the claim text.

Facts

Application: 10/315,188Filed 10 Dec 2002
Priority: 17 Apr 2002
Published: 23 Oct 2003
Term ends: Not estimatedThe source records don’t support an estimate.

Classification7 codes

H04M 7/006
Arrangements for interconnection between switching centres: networks other than PSTN/ISDN providing telephone service, e.g. Voice over Internet…
H04M 3/56
Systems providing special services or facilities to subscribers: arrangements for connecting several subscribers to a common circuit, i.e.…
H04M 3/562
Arrangements for connecting several subscribers to a common circuit, i.e. affording conference facilities: where the conference facilities are…
H04M 3/568
Arrangements for connecting several subscribers to a common circuit, i.e. affording conference facilities: audio processing specific to telephonic…

Family

Application10/123,310Continuation in part · filed 17 Apr 2002
Application10/123,483Continuation in part · filed 17 Apr 2002
Application10/315,188Continuation in part · filed 10 Dec 2002
US 2003/0198193 A1Published application
US 2003/0198328 A1Published application
US 7,020,257 B2Related patent
US 7,145,884 B2Related patent
US 7,292,543 B2Related patent
US 2003/0198195 A1This patent

From the US filing history. Foreign family members aren’t listed.