Joint sound model generation techniques
AssigneeAdobe Systems Incorporated
InventorsDennis L. Sun, Gautham J. Mysore
Life of this patent
Estimating the term…
Abstract
Joint sound model generation techniques are described. In one or more implementations, a plurality of models of sound data received from a plurality of different sound scenes are jointly generated. The joint generating includes learning information as part of generating a first said model of sound data from a first one of the sound scenes and sharing the learned information for use in generating a second one of the models of sound data from a second one of the sound scenes.
Claims20
A method implemented by one or more computing devices, the method comprising:
- obtaining sound data from a plurality of different recordings that correspond to a plurality of different sound scenes;
- jointly generating a plurality of models of sound data, each of the models serving as a dictionary that represents characteristics of a sound source included in at least one of the plurality of different sound scenes, and the joint generating including:
- learning information as part of generating a first said model of sound data from a first said recording of a first said sound scene;
- and
- sharing the learned information to generate a second said model of sound data from a second said recording of a second said sound scene;
- and
- decomposing sound data using at least one of the jointly generated plurality of models and without using training data.
A system comprising:
- one or more modules implemented at least partially in hardware and configured to jointly generate a plurality of individual models for respective ones of a plurality of sound data by sharing information during the generating to utilize a penalty that encourages spectral components of the respective said models to have similar values, an individual model serving as a dictionary that represents characteristics of a sound source included in at least one of a plurality of sound scenes that correspond to the plurality of sound data;
- and
- at least one module implemented at least partially in hardware and configured to decompose sound data using at least one of the jointly generated plurality of models and without using training data.
One or more computer readable storage media comprising instructions stored thereon that, responsive to execution by a computing device, causes the computing device to perform operations comprising:
- jointly generating a plurality of individual models for respective ones of a plurality of sound data by sharing information learned during the generating from a plurality of recordings that correspond the plurality of sound data, the information shared to utilize a penalty that encourages spectral components of the respective said models to have similar values;
- and
- decomposing sound data using at least one of the jointly generated plurality of models and without using training data.
Dependencies are inferred from the claim text.