Method for recognizing speech/speaker using emotional change to govern unsupervised adaptation
AssigneeSony Deutschland GmbH
InventorThomas Kemp
Life of this patent
Estimating the term…
Abstract
To improve the performance and the recognition rate of a method for recognizing speech in a dialogue system, or the like, it is suggested to derive emotion information data (EID) from speech input (SI) being descriptive for an emotional state of a speaker or a change thereof based upon which a process of recognition is chosen and/or designed.
Claims14
Method for recognizing speech/speaker in a dialogue system, or the like,
- wherein speech input (SI) of at least one speaker is received,
- wherein a process of recognition based on said speech input (SI) is performed,
- emotion information data (EID) being descriptive for an emotional state of a speaker or a change thereof are derived from said speech input (SI) or from parts thereof, and
- wherein said process of recognition is chosen and/or designed and/or modified based on said emotion information data (EID).
Method according to any one of the preceding claims,
- wherein the step of choosing and/or designing the process of recognition includes a step of selecting acoustic models (AM1, AM2), or the like and/or sets thereof to be involved in the recognition process.
Method according to any one of the preceding claims,
- wherein said emotion information data (EID) are generated so as to reflect emotional states of a given speaker before and after a recognition result is presented and/or a system action (SA) in response to the recognition result is performed.
Method according to any one of the preceding claims,
- wherein a first set of acoustic models is selected for the process of recognition until said emotion information data (EID) show a predetermined emotional shift of a speaker, upon which a second set of acoustic models is selected.
Method according to any one of the preceding claims,
- wherein said process of recognition at least includes a sub-process of recognizing speech.
Method according to any one of the preceding claims,
- wherein said process of recognition at least includes a sub-process of recognizing a speaker.
Dependencies are inferred from the claim text.