System for speech recognition with multi-part recognition

• US 7,392,189 B2
• Filed: 02/21/2003
• Issued: 06/24/2008
• Est. Priority Date: 02/23/2002
• Status: Expired due to Fees

• Current Assignee: Harman Becker Automotive Systems GmbH (Samsung Electronics Co. Ltd.)
• Original Assignee: Harman Becker Automotive Systems GmbH (Samsung Electronics Co. Ltd.)
• Inventors: Koch, Walter, Nüβle, Gerhard, Hennecke, Marcus, Reng, Richard
• Primary Examiner(s): Hudspeth; David
• Assistant Examiner(s): ALBERTALLI, BRIAN LOUIS
• Application Number:
  US10/371,982
• Publication Number:
  US 20040034527A1
• Time in Patent Office: 1,950 Days
• Field of Search: None
• US Class Current: 704/254
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams
• CPC Class Codes: G10L 15/32 Multiple recognisers used in sequence or in parallel; Score combination systems therefor, e.g. voting systems

Abstract

A speech recognition system for processing voice inputs from a user to select a list element from a list or group of list elements. Recognition procedures are carried out on the voice input of the user. One recognition procedure separates the voice input of a whole word into at least one sequence of speech subunits to produce a vocabulary of list elements. Another recognition procedure compares the voice input of the whole word with the vocabulary of list elements.

Claims

1. A method for speech recognition to select a list element from a list of list elements, comprising:

storing a voice input, where the voice input comprises a whole word of a desired list element;

separating the voice input into multiple speech subunits;

forming at least one sequence of speech subunits in accordance with the separated voice input;

configuring a speech recognition unit in response to the at least one sequence of speech subunits;

generating with a first recognition procedure a sub-list of list elements from the list of list elements in response to the at least one sequence of speech subunits wherein upon determination that the representation of the list of list element is different from the representation of the speech subunits, the first recognition procedures comprises mapping the representation of the list of list element to that of the speech subunits, or mapping the representation of the speech subunits to that of the list of list elements;

comparing with a second recognition procedure the list elements in the sub-list with the stored whole word voice input; and

selecting at least one list element of the sub-list as a recognition result.

2. The method according to claim 1 where the recognition result comprises a list element to match the desired list element.

3. The method according to claim 1 further comprising forming at least one character sequence out of at least one sequence of speech subunits.

4. The method according to claim 3 where the characters of the at least one character sequence and the list elements of the whole list are presented identically.

5. The method according to claim 3 where the characters of at least one character sequence and the list elements of the list are presented as at least one of letters and phonetic characters.

6. The method according to claim 3 where representation of the characters of the at least one character sequence is specified in response to the representation of the list elements of the list.

7. The method according to claim 6 where the characters of the at least one character sequence and the list elements of the list represent at least one of letters and phonetic characters.

8. The method according to claim 1 where the first recognition procedure further comprises comparing a hypothesized list composed of a plurality of different character sequences with the list of list elements.

9. The method according to claim 8 further comprising:

forming the at least one sequence of speech subunits using speech subunits having similar characters; and

generating character sequences of the hypothesized list in response to the at least one sequence of speech subunits.

10. The method according to claim 1 further comprising acoustically communicating the at least one selected list element to the user.

11. The method according to claim 1 further comprising optically communicating the at least one selected list element to the user.

12. The method according to claim 1 further comprising:

forming at least one character sequence in response to the at least one sequence of speech subunits, where the character sequence comprises at least one character; and

generating the sub-list of list elements in response to the at least one character sequence.

13. The method according to claim 1 implemented in at least one of a destination guidance system, a navigation system, a directory assistance system and a system for querying a data base.

14. A speech recognition system, comprising:

an input unit;

a first speech recognition unit connected to receive a voice input from the input unit, the first speech recognition unit configured with speech subunits as a first vocabulary for separating the voice input into speech subunits;

a recording unit connected with the input unit, the recording unit to store the voice input where the voice input comprises a whole word;

a database connected with the first speech recognition unit, the database containing a list of list elements for creating a sub-list in response to a match with at least one character sequence generated using the speech subunits, in accordance with the first speech recognition unit

wherein upon determination that the representation of the list of list element is different from the representation of the speech subunits, the first speech recognition unit operates that the representation of the list of list elements is mapped to the representation of the speech subunits, or the representation of the speech subunits is mapped to the representation of the list of list elements;

a second speech recognition unit connected with the recording unit, the second speech recognition unit configured with a second vocabulary created in response to the list elements of the sub-list, the second speech recognition unit for comparing the whole word voice input of the list element with the list elements of the sub-list; and

an output unit connected with the second speech recognition unit, the output unit to transmit at least one of the list elements evaluated as being a match the list element.

15. The speech recognition system according to claim 14 further comprising a mapping unit connected with the first speech recognition unit, the mapping unit to image the speech subunits in response to the characters of the at least one character sequence.

16. The speech recognition system according to claim 14 implemented in one of a destination guidance system, a navigation system, a directory assistance system and a system for querying a data base.

17. A method for determining a list element from a list of list elements, comprising:

inputting a whole word of a list element,

separating the inputted whole word into at least one subunit;

forming at least one sequence of subunits based on the inputted whole word, independently from the list of list elements;

processing one of the at least one sequence of subunits and the list of list elements in a form that can be compared, wherein upon determination that the representation of the list of list element is different from the representation of the subunits, the processing comprises mapping the representation of the list of list elements to that of the subunits, or mapping the representation of the subunits to that of the list of list elements;

matching the at least one subunit with a sub-list of list elements; and

determining at least one list element from the sub-list tat matches the whole word, where the sequence of subunits is determined in a first recognition procedure and where the whole word recognition is performed in a second recognition procedure.

18. A method for speech recognition, comprising:

storing a list of list elements in a database, the list elements comprising data entries in list;

separating a voice input into multiple speech subunits;

forming at least one sequence of speech subunits;

generating, in a first recognition procedure, a hypothesis list of sequences by converting the at least one sequence of speech subunits into the hypothesis list of sequences having a form that can be compared with the list of list elements, wherein upon determination that the representation of the list of list element is different from the representation of the speech subunits, the generating in the first recognition procedure comprises mapping the representation of the list of list elements to that of the speech subunits, or mapping the representation of the speech subunits to that of the list of list elements;

comparing, in the first recognition procedure, the hypothesis list of sequences with at least one list element from the list of list elements;

generating the sub-list of list elements in response to the at least one character sequence;

comparing, in a second recognition procedure, the list elements in the sub-list with the voice input; and

selecting at least one list element of the sub-list as a recognition result.