ASSESSING SPEECH PROSODY

• US 20110270605A1
• Filed: 04/29/2011
• Published: 11/03/2011
• Est. Priority Date: 04/30/2010
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Qin, Yong, Shuang, Zhiwei, Shi, Qin, Zhang, Shi Lei
• Granted Patent: US 9,368,126 B2
• Time in Patent Office: Days
• US Class Current: 704/9
• CPC Class Codes: G10L 25/48 specially adapted for particular use

Abstract

A method, system and computer readable storage medium for assessing speech prosody. The method includes the steps of: receiving input speech data; acquiring a prosody constraint; assessing prosody of the input speech data according to the prosody constraint; and providing assessment result where at least of the steps is carried out using a computer device.

Claims

1. A method for assessing speech prosody, comprising:

receiving input speech data;

acquiring a prosody constraint;

assessing prosody of said input speech data according to said prosody constraint; and

providing an assessment result;

wherein at least one of the steps is carried out using a computer device.

2. The method according to claim 1, wherein said prosody constraint comprises a rhythm feature constraint.

3. The method according to claim 2, wherein:

said acquiring step further comprises the steps of;

acquiring a rhythm feature of said input speech data; and

acquiring a standard rhythm feature corresponding to said input speech data; and

said assessing step further comprises the step of comparing said rhythm feature of the input speech data with said corresponding standard rhythm feature.

4. The method according to claim 3, wherein:

said rhythm feature is represented as a phrase boundary location;

said standard rhythm feature is represented as a standard phrase boundary location; and

said phrase boundary and said standard phrase boundary containing a characteristic selected from the group consisting of silence and pitch reset.

5. The method according to claim 4, wherein said comparing step further comprises the step of determining whether said phrase boundary location matches with said standard phrase boundary location.

6. The method according to claim 4, wherein the step of acquiring said rhythm feature further comprises the steps of:

acquiring input text data corresponding to said input speech data;

aligning said input text data with said input speech data; and

determining said phrase boundary location.

7. The method according to claim 6, wherein the step of acquiring standard rhythm feature further comprises the steps of:

processing said input text data to acquire an input language structure which corresponds with said input text data;

matching said input language structure with said standard language structure to determine an occurrence probability of said phrase boundary location; and

extracting said standard phrase boundary location.

8. The method according to claim 7, wherein said step of extracting said standard phrase boundary location further comprises the steps of:

determining a standard occurrence probability of said standard phrase boundary location; and

extracting said standard phrase boundary location if said standard occurrence probability is above a certain threshold.

9. The method according to claim 7, where said matching step further comprises the step of traversing a decision tree of said standard language structure according to said input language structure of at least one word of the input text data to determine occurrence probability of said phrase boundary location of said at least one word.

10. The method according to claim 1 further comprising the step of acquiring a fluency feature of said input speech data wherein said prosody constraint comprises a fluency feature constraint.

11. The method according to claim 10, wherein said step of acquiring said fluency feature further comprises the steps of:

acquiring input text data corresponding to said input speech data;

aligning said input text data with said input speech data; and

measuring a fluency feature of said input speech data.

12. The method according to claim 11, wherein:

said fluency feature comprises a total number of phrase boundaries within a sentence in said input text data;

said phrase boundary containing a characteristic selected from the group consisting of silence and pitch reset;

said step of acquiring said prosody constraint further comprises the step of predicting a total number of phrase boundaries based on a length of said sentence; and

said step of assessing prosody further comprises the step of comparing said total number of phrase boundaries with said predicted total number of phrase boundaries.

13. The method according to claim 10, wherein:

said fluency feature comprises a silence duration within a first phrase boundary;

said step of acquiring prosody constraint further comprises the step of determining a standard silence duration corresponding to said input speech data; and

said step of assessing prosody further comprising the step of comparing said silence duration with said standard silence duration;

wherein said first phrase boundary is a phrase boundary of at least one word from said input text data.

14. The method according to claim 13, wherein the step of determining said standard silence duration further comprises the steps of:

processing said input text data to obtain a corresponding input language structure; and

matching said input language structure with a standard language structure of standard speech in a standard corpus to determine said standard silence duration.

15. The method according to claim 14, wherein said step of matching said input language structure comprises:

traversing a decision tree of said standard language structure according to said first phrase boundary to determine said standard silence duration of said first phrase boundary;

wherein said first phrase boundary is within said input language structure; and

wherein said standard silence duration is an average value of a silence duration within a second phrase boundary of said standard language structure.

16. The method according to claim 10, wherein:

said fluency feature comprises a repetition number wherein said repetition number represents a number of times a word is repeated within said input speech data;

said step of acquiring prosody constraint further comprises the step of acquiring a permissible value;

said step of assessing prosody further comprises the step of comparing said repetition number with said permissible value.

17. The method according to claim 10, wherein:

said fluency feature comprises a phone hesitation degree wherein said phone hesitation degree includes a metric selected from the group consisting of number of phone hesitation times and phone hesitation duration;

said step of acquiring prosody constraint further comprises the step of acquiring a permissible value; and

said step of assessing prosody further comprises the step of comparing said phone hesitation degree with said permissible value.

18. The method according to claim 1, further comprising the step of adding said input speech data into a corpus as standard speech data according to said assessment result.

19. The method according to claim 1, wherein said assessment result is a result selected from the group consisting of a score of prosody of the input speech data, a detailed analysis on prosody of said input speech data, and reference speech data.

20. A system for assessing speech prosody, comprising:

an input speech data receiver configured for receiving input speech data;

a prosody constraint acquiring means configured for acquiring a prosody constraint;

an assessing means configured for assessing prosody of said input speech data according to said prosody constraint; and

a result providing means configured for providing an assessment result.

21. The system according to claim 20, wherein the prosody constraint comprises a rhythm feature constraint.

22. The system according to claim 21, further comprising a rhythm feature acquiring means configured for acquiring a rhythm feature of said input speech data wherein said rhythm feature is represented as a phrase boundary location and said phrase boundary containing a characteristic selected from the group consisting of silence and pitch reset;

wherein said prosody constraint acquiring means is further configured for acquiring a standard rhythm feature corresponding to said input speech data; and

wherein said assessing means is further configured for comparing said rhythm feature of said input speech data with said corresponding standard rhythm feature.

23. The system according to claim 20, wherein the prosody constraint comprises a fluency feature constraint, the system further comprising:

a fluency feature acquiring means configured for acquiring a fluency feature of said input speech data,

wherein said prosody feature acquiring means is further configured for;

acquiring input text data corresponding to said input speech data;

aligning said input text data with said input speech data; and

measuring said fluency feature of said input speech data.

24. The system according to claim 23, wherein the fluency feature is a feature selected from the group consisting of total number of phrase boundaries, silence duration of said phrase boundary;

number of repetition times of a word; and

phone hesitation degree;

wherein said phrase boundary contains a characteristic selected from the group consisting of silence and pitch reset; and

wherein said phone hesitation degree includes a metric selected from the group consisting of a number of phone hesitation times and phone hesitation duration.

25. A computer readable storage medium tangibly embodying a computer readable program code having computer readable instructions which when implemented, cause a computer to carry out the steps of a method comprising:

receiving input speech data;

acquiring a prosody constraint;

assessing prosody of said input speech data according to said prosody constraint; and

providing an assessment result.