Speaker-dependent speech recognition using speaker independent models
First Claim
Patent Images
1. A method of providing speech recognition of any word or phrase comprising the steps of:
- storing speaker-independent subword models and assigned identifiers for said models wherein Said identifiers are non-acoustic data pointers to said models;
receiving speech input;
recognizing said speech input as a series of occurrences of speaker-independent subword models and associating a label with a series of identifiers corresponding to the series of speaker-independent subword models recognized;
storing said label and said series of identifiers which represent the word or phrase, whereby said label and said series of identifiers are used to recognize the word or phrase; and
outputting said label recognized when input speech compares with speaker-independent subword model series defined by said series of identifiers.
0 Assignments
0 Petitions
Accused Products
Abstract
The memory and data management requirements for text independent speaker dependent recognition are drastically reduced by using a novel approach that eliminates the need for separate acoustic recognition models for each speaker. This is achieved by using speaker independent recognition models at the acoustic level. The speaker dependent data stored for each item to be recognized consists only of information needed to determine the speaker independent recognition model sequence for that item.
51 Citations
11 Claims
-
1. A method of providing speech recognition of any word or phrase comprising the steps of:
-
storing speaker-independent subword models and assigned identifiers for said models wherein Said identifiers are non-acoustic data pointers to said models; receiving speech input; recognizing said speech input as a series of occurrences of speaker-independent subword models and associating a label with a series of identifiers corresponding to the series of speaker-independent subword models recognized; storing said label and said series of identifiers which represent the word or phrase, whereby said label and said series of identifiers are used to recognize the word or phrase; and outputting said label recognized when input speech compares with speaker-independent subword model series defined by said series of identifiers. - View Dependent Claims (2, 3, 4, 5, 6)
-
-
7. An apparatus for providing speech recognition of any word or phrase comprising in combination:
-
a storage for storing speaker-independent subword models and assigning identifiers for said models wherein said identifiers are non-acoustic data pointers to said models; a receiver for receiving speech input; a first recognizer coupled to said storage and said receiver and responsive to said speech input and said speaker-independent subword models for recognizing said input speech as a series of occurrences of speaker-independent subword models and associating a label with a series of identifiers corresponding to said series of speaker-independent subword models recognized; and means coupled to said recognizer for storing said label and said series of identifiers which represent said word or phrase. - View Dependent Claims (8, 9, 10, 11)
-
Specification