Voice-controlled data system

• US 8,666,727 B2
• Filed: 02/21/2006
• Issued: 03/04/2014
• Est. Priority Date: 02/21/2005
• Status: Active Grant

• Current Assignee: Harman Becker Automotive Systems GmbH (Samsung Electronics Co. Ltd.)
• Original Assignee: Harman Becker Automotive Systems GmbH (Samsung Electronics Co. Ltd.)
• Inventors: Hennecke, Marcus
• Primary Examiner(s): JACKSON, JAKIEDA R
• Application Number:
  US11/359,755
• Publication Number:
  US 20070198273A1
• Time in Patent Office: 2,933 Days
• Field of Search: 704/8
• US Class Current: 704/8
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

A voice-controlled data system may include a data storage unit including media files having associated file identification data, and a vocabulary generating unit generating phonetic data corresponding to the file identification data, the phonetic data being supplied to a speech recognition unit as a recognition vocabulary, where one of the media files may be selected according to a recognized speech control command on the basis of the generated phonetic data, where the file identification data include a language identification part for identifying the language of the file identification data, and where the vocabulary generating unit generates the phonetic data for the file identification data of a media file based on its language identification part.

Claims

1. A voice-controlled data system comprising:

a data storage unit configured to store a plurality of media files;

a vocabulary generating unit configured to generate multilingual phonetic data corresponding to each of the media files;

a speech recognition unit configured to recognize a speech control command;

a processor configured to select a media file from the media files based on a comparison of a part of the speech control command and the generated multilingual phonetic data corresponding to the media files, where the multilingual phonetic data corresponding to any one of the media files comprises a phonetic representation of each data field of multilingual file identification data in a header section of the media file, and where the phonetic representation of each data field is based on a language identifier associated with said each data field of the multilingual file identification data in the header section; and

a media playback unit configured to play the selected media file.

2. The data system of claim 1, further comprising:

at least one language database including phonetic rules and phonetic transcriptions of terms of a plurality of languages.

3. The data system of claim 1, further comprising a language identification unit configured to extract the language identifier from a data field of the multilingual file identification data in the header section of the media file.

4. The data system of claim 3, further comprising a language database selecting unit, executable by the processor, configured to select a language database from a plurality of language databases in accordance with the extracted language identifier.

5. The data system of claim 1, where the multilingual file identification data includes at least one natural language term.

6. The data system of claim 1, where the media files include audio or video files, and the multilingual file identification data of the media file comprises a title of the media file, a composer of the media file, an author of the media file, an artist of the media file, and a genre of the media file.

7. The data system of claim 1, where the media file player is further configured to operate based on the speech control command.

8. The data system of claim 1, where the multilingual file identification data include data sets of different languages.

9. A multimedia system comprising the data system of claim 1.

10. The data system of claim 1, where the multilingual phonetic data comprises a first phonetic representation and a second phonetic representation, where,

the first phonetic representation is generated according to a first language indicated by a first language identifier associated with a first data field of the multilingual file identification;

and

the second phonetic representation is generated according to a second language indicated by a second language identifier associated with a second data field of the multilingual file identification.

11. A method for the voice-controlled selection of a media file stored on a data storage unit, the data storage unit including a plurality of media files, the media files including respective file identification data, the method comprising:

receiving voice data indicative of selection of the media file from among the media files, and supplying the voice data to a speech recognition unit;

extracting, by a processor, a first language identification tag included in the respective file identification data of each of the media files, the first language identification tag indicating a first language associated with a first data field of the respective file identification data, where the respective file identification data is in a header section of each of the respective media files;

extracting, by the processor, a second language identification tag included in the respective file identification data of each of the media file, the second language identification tag indicating a second language associated with a second data field of the respective file identification data in the header section of each of the respective media files;

generating, by the processor, phonetic data corresponding to the file identification data for each of the media files, the generated phonetic data comprising phonetic representations of the first data field and the second data field that are generated based on the first language identification tag and the second language identification tag, respectively;

comparing, by the processor, the generated phonetic data to the received voice data by the speech recognition unit and generating a corresponding speech control command; and

selecting, by the processor, the media file from the data storage unit in accordance with the generated speech control command.

12. The method of claim 11, where generating the phonetic data includes accessing a plurality of language databases that include phonetic transcriptions of a respective language, where the language databases are accessed in accordance with at least one of the first language identification tag or the second language identification tag.

13. The method of claim 11, where generating the phonetic data includes accessing at least one language database including phonetic rules of a language, and the language database is accessed in accordance with at least one of the first language identification tag or the second language identification tag.

14. The method of claim 11, where the file identification data of the media file includes a name of an author of the media file, and the generated phonetic data being a phonetic representation of the name of the author.

15. The method of claim 11, where the file identification data of the media file includes a title of the media file, and the generated phonetic data being a phonetic representation of the title.

16. The method of claim 11, further comprising transferring the selected media file to a media file player where the selected media file is reproducible.

17. The method of claim 11, where generating the phonetic representation of the first data field includes applying phonetic rules of the first language identified by the first language identification tag on natural language text stored in the file identification data.

18. The method of claim 11, further comprising converting the generated phonetic data to the speech control command by a pattern matching technique applied in the speech recognition unit.

19. The method of claim 18, where the pattern matching technique includes a technique selected from the group consisting of dynamic time warping, dynamic programming, a statistical modeling technique, a hidden Markov Model, and combinations of two or more of these.

20. A system comprising:

a data storage unit configured to store a plurality of media files, where each media file includes file identification data in a header section, the file identification data includes a first data entry containing a first natural language term and the file identification data further includes a second data entry containing a second natural language term;

a language estimation unit configured to identify a first language of the first natural language term in the first data entry and further configured to add an association between a first language tag and the first data entry, where the first language tag is an indication of the first language;

the language estimation unit further configured to identify a second language of the second natural language term in the second data entry and further configured to add an association between a second language tag and the second data entry, where the second language tag is an indication of the second language;

a vocabulary generating unit configured to generate phonetic data associated with each respective media file of the media files, where the phonetic data contains phonetic representations of the first natural language term and the second natural language term based on the first language tag and the second language tag, respectively;

a speech recognition unit configured to receive a natural language voice input representative of a request to select a media file;

a processor configured to select the media file based on a comparison of the natural language voice input with the phonetic data generated for each of the media files based on the respective first and second language tags.

21. The system of claim 20, where data entries in the file identification data of each media file include information selected from the group consisting of an artist of the media file, a title of the media file, a composer of the media file, a recording year of the media file, and a genre of the media file, and combinations of two or more of these.

22. The system of claim 20, where the language estimation unit is configured to identify the first language of the first natural language term in the first data entry based on a statistical model and file identification data of other media files.

23. The system of claim 20, where the language estimation unit is configured to identify the first language of the first natural language term in the first data entry based on a statistical model and content of the second data entry of the file identification data of the media file.

24. A method for automatically detecting the language of file identification data of a media file, the file identification data including a plurality of data fields in a header section of the media file allowing identification of the media file, the method comprising:

receiving, with a processor, a selection of at least one media file out of a group of media files;

retrieving, with the processor, a first data field and a second data field of the data fields in the file identification data;

estimating a first language of the first data field of the file identification data of the media file with the processor based on a statistical model;

estimating a second language of the second data field of the file identification data of the media file with the processor based on the statistical model;

adding, with the processor, the estimated first language and the estimated second language to the file identification data;

linking, with the processor, the estimated first language with the first data field and the estimated second language with the second data field; and

identifying, by the processor, the media file in response to receipt of a voice control command by comparing the voice control command to phonetic data comprising a first phonetic representation of the first data field in the estimated first language and a second phonetic representation of the second data field in the estimated second language.

25. The method of claim 24, where the file identification data include different data sets, at least one data set includes the at least one data entry, and where the natural language is estimated for more than one of the data entries of the file identification data.

26. The method of claim 24, where the natural language of the at least one data entry is determined taking into account file identification data of other media files.

27. The method of claim 26, further comprising:

estimating the natural language for that at least one data entry of each of the other media files;

calculating a score for each estimated natural language, the score indicating a degree to which the file identification data matches the respective estimated natural language.

28. The method of claim 27, where the estimated natural language of the at least one data entry of the media file is estimated based on the calculated score.

29. The method of claim 24, where the natural language of the at least one data entry is determined taking into account file identification data of other data entries of the media file.

30. The method of claim 24, where estimating the natural language includes utilizing a confidence based approach including determining a confidence value for each estimated language, the confidence value being a measure of assuredness that the estimated language of the data entry is the correct language.

31. The method of claim 30, where the natural language of the data entry is estimated based on the determined confidence value.

32. The method of claim 31, where the data files are arranged in accordance with the determined confidence value.

33. The method of claim 24, where the natural language is determined for information included in the language identification data, the information selected from the group consisting of an artist of the media file, a title of the media file, a composer of the media file, a recording year of the media file, a genre of the media file, and combinations of two or more of these.

34. The method of claim 24, where estimating the natural language includes estimating a language of a name of an artist of the media file by determining a language of a title of the media file, and where the determined language of the title is utilized as the language of the name of the artist.

35. The method of claim 24, where estimating the natural language includes determining a language of a name of an artist by:

identifying other media files of the artist;

estimating languages of titles of the other identified media files; and

determining the language of the name of the artist based on the estimated languages of the other media files.

36. The method of claim 24, where estimating the natural language includes determining a language of a title of the media file by:

identifying an artist of the media file;

identifying other media files of the artist;

estimating languages of titles of the other media files of the artist; and

estimating the language of the title of the media file in correspondence with the estimated languages of the other media files.

37. The method of claim 36, where estimating the languages of the titles of the other media files of the artist includes combining the titles of the other media files into a new title string and estimating the natural language for the new title string, and where the estimated language is utilized as language of the title of the media file.

38. The method of claim 36, where the language most estimated among the languages of the titles of the other media files is utilized as language of the title of the media file.

39. The method of claim 36, where determining the language of the title of the first media file includes determining a score of each estimated language for the titles of the other media files and combining the scores of each estimated language, and where the language with the highest combined score is utilized as language of the title of the media file.

40. The method of claim 24, where the media file is a first media file, the method further comprising:

determining a second media file having a particular title and different artists.

41. The method of claim 24, where the estimated natural language is a first estimated natural language, and the at least one data entry is a first data entry, the method further comprising:

associating, with the processor, the first estimated language with the first data entry;

retrieving, with the processor, a second data entry of the file identification data;

estimating a second natural language of the second data entry with the processor based on the statistical model;

adding the second estimated natural language to the file identification data; and

associating the second estimated natural language with the second data entry.