Arrangement of speaker-independent speech recognition

• US 7,392,184 B2
• Filed: 04/15/2002
• Issued: 06/24/2008
• Est. Priority Date: 04/17/2001
• Status: Expired due to Term

• Current Assignee: HMD Global Oy
• Original Assignee: Nokia Corporation
• Inventors: Laurila, Kari, Viikki, Olli
• Primary Examiner(s): Vo, Huyen X.
• Application Number:
  US10/122,730
• Publication Number:
  US 20020152067A1
• Time in Patent Office: 2,262 Days
• Field of Search: 704/256, 704/277, 704/260, 704/258, 704/207, 704/8, 704/9, 704/270, 704/1, 704/5, 704/7, 704/236, 704/250, 704/243, 704/244, 704/231, 704/246, 709/203, 709/260
• US Class Current: 704/236
• 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 needed in speech recognition for forming a pronunciation model in a telecommunications system comprising at least one portable electronic device and server. The electronic device is arranged to compare the user'"'"'s speech information with pronunciation models comprising acoustic units and stored in the electronic device. A character sequence is transferred from the electronic device to the server. In the server, the character device is converted into a sequence of acoustic units. A sequence of acoustic units is sent from the server to the electronic device.

Claims

1. A method of forming a pronunciation model for speech recognition in a telecommunications system comprising at least one portable electronic device and a server, the electronic device being configured to compare the user'"'"'s speech information with pronunciation models comprising acoustic units and stored in the electronic device, the method comprising:

transferring a character sequence from the electronic device to the server, comparing the character sequence with a language resource to determine the language of the character sequence in a language selector of the electronic device,

selecting a text to phenome resource, stored in the server, according to the determined language;

converting the character sequence in the server into at least one phoneme sequence in text format in accordance with the determined language; and

transferring said at least one phoneme sequence in text format from the server to the electronic device;

storing the phoneme sequence in association with the transferred character sequence in the electronic device;

forming an audio model of the phoneme sequence at the electronic device;

storing the audio model in the electronic device in association with the character sequence; and

comparing received speech information with the stored phoneme sequence and if a match results, selecting the character sequence associated with the matching phoneme sequence.

2. A method according to claim 1, the method further comprising:

selecting a character sequence according to said phoneme sequence from contact information; and

activating a service in accordance with said character sequence.

3. A method according to claim 1, the method further comprising:

searching in the server for information related to the character sequence, e.g. telephone numbers, on the basis of the received character sequence; and

sending said information in addition to the phoneme sequence to the electronic device.

4. A method according to claim 1, the method further comprising:

playing the audio model to the user of the electronic device as a response to the user'"'"'s speech command being substantially matching the phoneme sequence received from the character sequence.

5. A method according to claim 1, wherein the electronic device is a mobile station and the data transmission between the server and the electronic device is configured by messaging through a mobile network.

6. A method according to claim 1, wherein the language is determined by means of decision trees.

7. A telecommunications system comprising:

at least one electronic device and a server, wherein the electronic device is configured to send a character sequence intended for speech recognition to the server;

a language selector of the electronic device, adapted to compare the character sequence with a language resource to determine the language of the character sequence;

wherein the server is configured to convert the character sequence into at least one phoneme sequence in accordance with the determined language and the server is configured to send said at least one phoneme sequence to the electronic device;

wherein the electronic device is configured to form an audio model of the phoneme sequence and to store said audio model in association with the character sequence;

a memory in the electronic device for storing the phoneme sequence in association with the transferred character sequence in the electronic device; and

further wherein the electronic device is configured to compare the user'"'"'s speech information with the stored phoneme sequences stored in the electronic device and if a match results, selecting the character sequence associated with the matching phoneme sequence.

8. An electronic device comprising:

a language selector adapted to compare a character sequence with a language resource to determine a language of the character sequence intended for speech recognition;

a transmitter adapted to send the character sequence intended for speech recognition and information on the language of the character sequence to a server;

a receiver adapted to receive a phoneme sequence formed of the character sequence in accordance with the determined language from the server;

a processor in the electronic device configured to form an audio model of the phoneme sequence and to store said audio model in association with said character sequence;

a memory adapted to store the phoneme sequence in association with the character sequence; and

wherein the processor is adapted to compare user'"'"'s speech information with stored phoneme sequence and if a match results, select the character sequence associated with the matching phoneme sequence.

9. An electronic device according to claim 8, wherein the processor is further adapted to:

associate a phoneme sequence received from the server to the character sequence stored in the memory of the electronic device or its tag;

select a phoneme sequence substantially according to the user'"'"'s speech information and further a character sequence according to said phoneme sequence; and

activate a service in accordance with said character sequence.

10. An electronic device according to claim 8, wherein the language selector is arranged to determine the language by means of decision trees.

11. A server comprising:

a communication connection for receiving a character sequence from at least one electronic device;

a language selector adapted to compare the character sequence with a language resource to determine a language for the character sequence intended for speech recognition;

a processor for converting the character sequence into at least one phoneme sequence in accordance with the determined language; and

wherein said processor is adapted to send the at least one phoneme sequence to the electronic device over the communication connection; and

and wherein the phoneme sequence is received and further processed and stored in association with the character sequence in the electronic device; and

wherein a processor in the electronic device is adapted to compare a user'"'"'s speech information with the stored phoneme sequence and if a match results, select the character sequence associated with the matching phoneme sequence; and

further wherein the processor in the electronic device is configured to form a audio model of the phoneme sequence and to store said speech model in association with the character sequence.

12. A server according to claim 11, wherein the language selector is arranged to determine the language by means of decision trees.

13. A computer program product comprising

a computer readable program code stored in a memory for causing a computer to perform speech recognition, said computer program product further comprising:

computer readable program code means for causing a computer server to receive a character sequence for speech recognition from an electronic device;

computer readable program code means for causing the computer server to compare the character sequence with a language resource to determine a language of the character sequence;

computer readable program code means for causing the computer server to convert the character sequence into at least one phoneme sequence in text format in accordance with the determined language; and

computer readable program code means for causing a processor in said electronic device to receive the at least one phoneme sequence in text format from the server;

computer readable program code means for causing the processor to store the phoneme sequence in association with the character sequence; and

compare received speech information with the stored phoneme sequence and, if a match results, select the character sequence associated with the matching phoneme sequence; and

computer readable program code means for causing the processor to form a audio model of the phoneme sequence and to store said speech model in association with the character sequence.