Learning device and pattern recognition device
AssigneeNEC Corporation
InventorsTatsuya KOMATSU, Reishi KONDO, Sakiko MISHIMA
Life of this patent
Estimating the term…
Abstract
The acoustic feature extraction means 82 extracts an acoustic feature, using predetermined parameters, from an acoustic pattern obtained as a result of processing on an acoustic signal. The language vector calculation means 83 calculates a language vector from a given label that represents an attribute of a source of the acoustic signal and that is associated with the acoustic pattern. The similarity calculation means 84 calculates a similarity between the acoustic feature and the language vector. The parameter update means 85 learns parameters so that the similarity becomes larger, and updates the predetermined parameters to the parameters obtained by learning.
Claims11
A learning device comprising:
- an acoustic feature extraction unit that extracts an acoustic feature, using predetermined parameters, from an acoustic pattern obtained as a result of processing on an acoustic signal;
- a language vector calculation unit that calculates a language vector from a given label that represents an attribute of a source of the acoustic signal and that is associated with the acoustic pattern;
- a similarity calculation unit that calculates a similarity between the acoustic feature and the language vector;
- and
- a parameter update unit that learns parameters so that the similarity becomes larger, and updates the predetermined parameters to the parameters obtained by learning.
A pattern recognition device comprising:
- an acoustic feature extraction unit that extracts an acoustic feature from an acoustic pattern using parameters for extracting the acoustic feature from the acoustic pattern, the parameters learned based on a pair of an acoustic pattern obtained as a result of processing on an acoustic signal and a label representing an attribute of a source of the acoustic signal;
- and
- an identification unit that identifies a label representing the attributes of the source of the acoustic signal on which the acoustic pattern is based, based on the acoustic feature.
A learning method, implemented by a computer, comprising:
- an acoustic feature extraction processing of extracting an acoustic feature, using predetermined parameters, from an acoustic pattern obtained as a result of processing on an acoustic signal;
- a language vector calculation processing of calculating a language vector from a given label that represents an attribute of a source of the acoustic signal and that is associated with the acoustic pattern;
- a similarity calculation processing of calculating a similarity between the acoustic feature and the language vector;
- and
- a parameter update processing of learning parameters so that the similarity becomes larger, and updating the predetermined parameters to the parameters obtained by learning.
11-20. (canceled)
Dependencies are inferred from the claim text.