Method for the Distributed Construction of a Voice Recognition Model, and Device, Server and Computer Programs Used to Implement Same

• US 20080103771A1
• Filed: 10/27/2005
• Published: 05/01/2008
• Est. Priority Date: 11/08/2004
• Status: Abandoned Application

• Current Assignee: Orange S.A.
• Original Assignee: Orange S.A.
• Inventors: Monne, Jean, Jouvet, Denis
• Application Number:
  US11/667,184
• Publication Number:
  US 20080103771A1
• Time in Patent Office: Days
• US Class Current: 704/250
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications

Abstract

A method for the distributed construction of a voice recognition model that is intended to be used by a device comprising a model base and a reference base in which the modeling elements are stored. The method includes the steps of obtaining the entity to be modeled, transmitting data representative of the entity over a communication link to a server, determining a set of modeling parameters indicating the modeling elements, transmitting the modeling parameters to the device, determining the voice recognition model of the entity to be modeled as a function of at least the modeling parameters received and at least one modeling element that is stored in the reference base and indicated in the transmitted parameters, and subsequently saving the voice recognition model in the model base.

Claims

1. A method of constructing a voice recognition model of an entity to be modeled, distributed between a device comprising a base of constructed models and a reference base in which modeling elements are stored, said device being able to communicate with a server via a communication link, said method comprising at least the following steps:

obtaining by the device the entity to be modeled;

transmitting by the device data representative of said entity over the communication link to the server;

receiving by the server said data to be modeled and performing by the server a processing to determine a set of modeling parameters indicating modeling elements from said data;

transmitting by the server said modeling parameters over the communication link to the device;

receiving by the device the modeling parameters and determining by the device the voice recognition model of the entity to be modeled as a function of at least the modeling parameters and at least one modeling element stored in the reference base and indicated in the received modeling parameters; and

storing by the device the voice recognition model of the entity to be modeled in the base of constructed models.

2. The method as claimed in claim 1, wherein said device is a user terminal with embedded voice recognition, the model being intended to be used by the user terminal.

3. The method as claimed in claim 1, wherein the processing performed by the server comprises a step for determining a set of phonetic description parameters of the entity to be modeled.

4. The method as claimed in claim 1, wherein the modeling parameters transmitted to the device comprise at least one of said phonetic description parameters, an acoustic model of said phonetic description parameter being stored in the reference base of the device.

5. The method as claimed in claim 1, wherein the processing performed by the server comprises at least one acoustic modeling step, according to which the server determines a Markov model comprising a set of acoustic description parameters associated with the entity to be modeled.

6. The method as claimed in claim 5, wherein the modeling parameters transmitted to the device comprise at least one acoustic probability density identifier, the description of said identified density, comprising a weighted sum of Gaussian functions, being stored in the device reference base.

7. The method as claimed in claim 5, wherein the modeling parameters transmitted to the device comprise at least one weighting coefficient associated with a Gaussian function identifier, the duly indicated Gaussian function being defined in the reference base of the device.

8. The method as claimed in claim 1 wherein, when at least one model of an entity to be modeled has been previously stored in the base of constructed models of the device, and comprising the step of, after determining the model corresponding to a new entity to be modeled, performing by the device a model factorizing step by analyzing said previously stored model and the model corresponding to the new entity, in order to identify common characteristics.

9. The method as claimed in claim 1 comprising the step of performing by the server also a step for factorizing the models of a list of entities comprising said entity to be modeled, by analyzing said models, in order to identify common characteristics.

10. The method as claimed in claim 1 comprising the step of, when a modeling element indicated by at least one received modeling parameter is not in the reference base of the device, sending by the device sends a request to the server via the communication link, to determine the associated modeling element and recover the corresponding parameters in order to add to the reference base.

11. A device able to communicate with a server via a communication link and comprising:

a base of constructed models;

a reference base in which modeling elements are stored;

means for obtaining an the entity to be modeled;

means for transmitting data representative of said entity over the communication link to the server;

means for receiving modeling parameters from the server, corresponding to said entity to be modeled and indicating modeling elements;

means for determining the voice recognition model of the entity to be modeled as a function of at least the received modeling parameters and at least one modeling element indicated in said modeling parameters and stored in the reference base; and

means for storing the voice recognition model of the entity to be modeled in the base of constructed models.

12. A server for performing some of the tasks for building voice recognition models intended to be stored and used by a device with embedded voice recognition, the server being able to communicate with the device via a communication link and comprising:

means for receiving data to be modeled, transmitted by the device, via the communication link;

means for performing a processing to determine a set of modeling parameters indicating modeling elements from said data;

means for transmitting said modeling parameters over the communication link to the device.

13. A computer program for constructing voice recognition models from an entity to be modeled, executable by a processing unit of a device intended to perform the embedded voice recognition, said device being able to communicate with a server via a communication link and comprising a base of constructed models and a reference base in which modeling elements are stored and, said computer program comprising instructions for executing the following steps, when the program is executed by said processing unit;

obtaining an entity to be modeled;

transmitting data representative of said entity over the communication link to the server;

receiving modeling parameters from the server corresponding to said entity to be modeled and indicating modeling elements;

determining the voice recognition model of the entity to be modeled as a function of at least the received modeling parameters and at least one modeling element indicated in said modeling parameters and stored in the reference base; and

storing the voice recognition model of the entity to be modeled in the base of constructed models.

14. A computer program for constructing voice recognition models, executable by a processing unit of a server for performing some of the tasks for building voice recognition models intended to be stored and used by a device with embedded voice recognition, the server being able to communicate with the device via a communication link, comprising instructions for executing the following steps, when the program is executed by said processing unit:

receiving data to be modeled, transmitted by the device, via the communication link;

performing a processing to determine a set of modeling parameters indicating modeling elements from said data;

transmitting said modeling parameters over the communication link to the device.