System and method of speech recognition for non-native speakers of a language

• US 20060020462A1
• Filed: 07/22/2004
• Published: 01/26/2006
• Est. Priority Date: 07/22/2004
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Reich, David E.
• Granted Patent: US 7,640,159 B2
• Time in Patent Office: Days
• 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 2015/025 Phonemes, fenemes or fenones being the recognition units
• CPC Class Codes: G10L 2015/227 of the speaker; Human-factor methodology

Abstract

An accent compensative speech recognition system and related methods for use with a signal processor generating one or more feature vectors based upon a voice-induced electrical signal are provided. The system includes a first-language acoustic module that determines a first-language phoneme sequence based upon one or more feature vectors, and a second-language lexicon module that determines a second-language speech segment based upon the first-language phoneme sequence. A method aspect includes the steps of generating a first-language phoneme sequence from at least one feature vector based upon a first-language phoneme model, and determining a second-language speech segment from the first-language phoneme sequence based upon a second-language lexicon model.

Claims

1. An accent compensative speech recognition system for use with a signal processor generating at least one feature vector based upon a voice-induced electrical signal, the system comprising:

a first-language acoustic module determining a first-language phoneme sequence based upon the at least one feature vector; and

a second-language lexicon module determining a second-language speech segment based upon the first-language phoneme sequence.

2. The system as defined in claim 1, wherein the first-language acoustic module comprises a plurality of distinct first-language acoustic modules determining distinct language phoneme sequences based upon the at least one feature vector;

and wherein the system pairs one of the plurality of first-language acoustic modules with the second-language lexicon module.

3. The system as defined in claim 2, wherein the pairing is based upon at least one confidence score.

4. The system as defined in claim 3, wherein the second-language lexicon module determines a plurality of speech segments, including the second-language speech segment, each having a corresponding confidence score and being determined based upon a distinct pairing of one of the plurality of first-language acoustic modules and the second-language lexicon module;

and wherein the confidence score associated with the second-language speech segment is greater than the confidence score associated with the confidence scores of each other of the plurality of speech segments.

5. The system as defined in claim 1, wherein the first-language acoustic module comprises a plurality of distinct first-language acoustic modules each determining a distinct language phoneme sequence based upon the at least one feature vector;

wherein the second-language lexicon module comprises a plurality of distinct second-language lexicon modules; and

wherein the system pairs one of the plurality of first-language acoustic modules with one of the plurality of second-language lexicon modules.

6. The system as defined in claim 5, wherein the pairing is based upon at least one confidence score corresponding to a distinct pairing of one the plurality of first-language acoustic modules and one of the plurality of second-language lexicon modules.

7. The system as defined in claim 6, wherein the second-language lexicon module determines a plurality of distinct speech segments, including the second-language speech segment, each distinct speech segment having a corresponding confidence score and being determined based upon a distinct pairing of one of the plurality of first-language acoustic modules and one of the plurality of second-language lexicon modules;

and wherein the confidence score corresponding to the second-language speech segment is greater than the confidence score corresponding to each other of the plurality of speech segments.

8. A method for use with a signal processor generating at least one feature vector based upon a voice-induced electrical signal, the method comprising the steps of:

generating a first-language phoneme sequence from the at least one feature vector based upon a first-language phoneme model; and

determining a second-language speech segment from the first-language phoneme sequence based upon a second-language lexicon model.

9. The method as defined in claim 8, wherein the step of generating comprises generating a plurality of distinct language phoneme sequences based upon a plurality- of first-language models;

and wherein the step of determining comprises determining a second-language speech segment based on a pairing of one of the plurality of first-language phoneme models and the second-language lexicon model.

10. The method as defined in claim 9, wherein the pairing is based upon at least one confidence score.

11. The method as defined in claim 10, wherein the step of determining comprises determining a plurality of distinct speech segments, including the second-language speech segment, each distinct speech segment having a corresponding confidence score and being determined based upon a distinct pairing of one of the plurality of first-language acoustic models and one of the plurality of second-language lexicon models;

and wherein the confidence score corresponding to the second-language speech segment is greater than the confidence score corresponding to the confidence scores of each other of the plurality of speech segments.

12. The method as defined in claim 8, wherein the step of generating comprises generating a plurality of distinct language phoneme sequences based upon a plurality of distinct first-language acoustic models;

and wherein the step of determining comprises determining a second-language speech segment based upon a pairing of one of the plurality of first-language acoustic models and one of a plurality of distinct second-language lexicon models.

13. The method as defined in claim 12, wherein the pairing is based upon a confidence score corresponding to a distinct pairing of one of the plurality of first-language acoustic models and one of the plurality of second-language lexicon modules.

14. The method as defined in claim 13, wherein the second-language lexicon module determines a plurality of speech segments, including the second-language speech segment, each having a corresponding confidence score and being determined based upon a distinct pairing of one of the plurality of first-language acoustic modules and one of the plurality of second-language lexicon modules;

and wherein the confidence score associated with the second-language speech segment is greater than the confidence score associated with each other of the plurality of speech segments.

15. A computer readable storage medium for use with a data communication network having a plurality of hosts, a plurality of edge nodes connected to the hosts, and at least one core node connected to the edge nodes, the storage medium comprising computer instructions for:

generating a first-language phoneme sequence from the at least one feature based upon a first-language phoneme model; and

determining a second-language speech segment from the first-language phoneme sequence based upon a second-language lexicon model.

16. The computer readable storage medium as defined in claim 15, wherein the computer instruction for generating comprises an instruction for generating a plurality of distinct language phoneme sequences based upon a plurality of first-language models;

and wherein the instruction for determining comprises an instruction for determining a second-language speech segment based upon a pairing of one of the plurality of first-language phoneme models and the second-language lexicon model.

17. The computer readable storage medium as defined in claim 16, wherein the pairing is based upon at least one confidence score.

18. The computer readable storage medium as defined in claim 17, wherein the instruction for determining comprises an instruction for determining a plurality of speech segments, including the second-language speech segment, each having a corresponding confidence score and being determined based upon a distinct pairing of one of the plurality of first-language acoustic models and the second-language lexicon model;

and wherein the confidence score associated with the second-language speech segment is greater than the confidence score corresponding to the confidence scores of each other of the plurality of speech segments.

19. The computer readable storage medium as defined in claim 15, wherein the instruction for generating comprises an instruction for generating a plurality of distinct language phoneme sequences based upon a plurality of distinct first-language acoustic models;

and wherein the instruction for determining comprises an instruction for determining a second-language speech segment based upon a pairing of one of the plurality of first-language acoustic models and one of a plurality of distinct second-language lexicon models.

20. The computer readable storage medium as defined in claim 19, wherein the pairing is based upon at least one confidence score based upon a distinct pairing of one of the plurality of first-language acoustic models and one of the plurality of second-language lexicon modules.

21. The computer readable storage medium as defined in claim 20, wherein the second-language lexicon module determines a plurality of speech segments, including the second-language speech segment, each having a corresponding confidence score and being determined based upon a distinct pairing of one of the plurality of first-language acoustic modules and one of the plurality of second-language lexicon modules;

and wherein the confidence score corresponding to the second-language speech segment is greater than the confidence score corresponding to each other of the plurality of speech segments.