Method for creating a speech model
View Patent ↗A transformation can be derived which would represent that processing required to convert a male speech model to a female speech model. That transformation is subjected to a predetermined modification, and the modified transformation is applied to a female speech model to produce a synthetic children's speech model. The male and female models can be expressed in terms of a vector representing key values defining each speech model and the derived transformation can be in the form of a matrix that would transform the vector of the male model to the vector of the female model. The modification to the derived matrix comprises applying an exponential p which has a value greater than zero and less than 1.
1. A computerized method for generating a speech model for a third speech group comprising the steps of:
deriving, with a computer, a transformation that would produce an existing second speech group model when operating on an existing first speech group model; and
applying, with a computer, the transformation to the second speech group model to produce a model for the third speech group,
wherein the first speech group constitutes adult males, the second speech group constitutes adult females, and the third speech group constitutes children, and wherein the model for the third speech group is used to recognize, with a computer, speech from members of said third group.
2. The method of claim 1 wherein the transformation is subjected to modification prior to the applying step.
3. The method of claim 2 wherein the modification comprises applying an exponential operation to the transformation.
4. The method of claim 3 wherein the exponential operation is done with an exponential value between zero and one.
5. The method of claim 4 wherein the exponential operation is done with an exponential value between approximately 0.25 and approximately 0.7.
6. The method of claim 4 wherein the exponential operation is done with an exponential value between approximately 0.4 and approximately 0.5.
7. The method of claim 4 wherein the exponential operation is done with an exponential value of approximately 0.5.
8. The method of claim 1 wherein the children's speech group model includes covariant values associated with its states and the covariant values are scaled to account for variability in children's speech.
9. The method of claim 8 wherein the covariance values are in the form of a diagonal matrix and the first six covariant features are scaled as follows:
1.40
1.33
1.27
1.21
1.15
1.09
and energy, delta-energy and delta-delta-energy values are scaled as follows:
1.45
1.35
1.15.
10. The method of claim 1 wherein the male and female speech models are represented as vectors of values representing states in a statistically modeled system, the transformation being a matrix that would transform the male vectors into the female vectors.
11. The method of claim 10 wherein the matrix is subjected to an exponential operation.
12. The method of claim 11 wherein the exponential operation is done with an exponential value between zero and one.
13. The method of claim 11 wherein the exponential operation is done with an exponential value between approximately 0.25 and approximately 0.7.
14. The method of claim 11 wherein the exponential operation is done with an exponential value between approximately 0.4 and approximately 0.5.
15. The method of claim 11 wherein the exponential operation is done with an exponential value of approximately 0.5.
16. The method of claim 10 wherein the children's speech group model includes covariance values associated with its states and the covariant values are scaled to account for variability in children's speech.
17. The method of claim 16 wherein the covariant values are in the form of a diagonal matrix and the first six covariant features are scaled as follows:
1.40
1.33
1.27
1.21
1.15
1.09
and energy, delta-energy and delta-delta-energy values are scaled as follows:
1.45
1.35
1.15.
18. The method of claim 1 wherein the transformation is subjected to modulation prior to the applying step.
19. The method of claim 18 wherein the modulation comprises applying an exponential operation to the transformation.
20. The method of claim 19 wherein the exponential operation is done with an exponential value between zero and one.
21. The method of claim 19 wherein the exponential operation is done with an exponential value between approximately 0.25 and approximately 0.7.
22. The method of claim 19 wherein the exponential operation is done with an exponential value between approximately 0.4 and approximately 0.5.
23. The method of claim 19 wherein the exponential operation is done with an exponential value of approximately 0.5.
24. The method of claim 1 wherein the third speech group model includes covariant values associated with its states and the covariant values are scaled to account for variability in speech being modeled.
25. A method comprising deriving, with a computer, a transformation function from a set of models of speech from first and second population types which differ from one another, and using the derived transformation function, with a computer, to transform a speech model from one of said first and second population types to a third population type, said third population type being different from both of said first and said second population types, and recognizing, with a computer, speech from individuals in said third population type using said derived speech model.
26. The method of claim 25 wherein said first and second population types are male and female adults, respectively, and said third population type is children.
27. The method of claim 25 wherein after said transformation to said third population type, said speech is used to train someone in a new language.