T-netix, Inc.

Other

Lucent Technologies Inc.

Rutgers, the State University of New Jersey

Xiaoyu Zhang

Richard James Mammone

Manish Sharma

Christine Irene Podilchuk

Kevin Farrell

Han-Seng Liou

Khaled Assaleh

Devang Naik

The voice print system of the present invention is a subword-based, text-dependent automatic speaker verification system that embodies the capability of user-selectable passwords with no constraints on the choice of vocabulary words or the language. An estimate of the enrollement channel and of the test channel is developed for inverse filtering of the enrollment or the test speech, respectively. Automatic blind speech segmentation allows speech to be segmented into subword units without any linguistic knowledge of the password. Subword modeling is performed using a multiple classifiers. The system also takes advantage of such concepts as multiple classifier fusion and data resampling to successfully boost the performance. Key word/key phrase spotting is used to optimally locate the password phrase. Numerous adaptation techniques increase the flexibility of the base system, and include: channel adaptation, fusion adaptation, model adaptation and threshold adaptation.

Subword-based speaker verification using multiple-classifier fusion, with channel, fusion, model and threshold adaptation

The voice print system of the present invention is a subword-based, text-dependent automatic speaker verification system that embodies the capability of user-selectable passwords with no constraints on the choice of vocabulary words or the language. Automatic blind speech segmentation allows speech to be segmented into subword units without any linguistic knowledge of the password. Subword modeling is performed using a multiple classifiers. The system also takes advantage of such concepts as multiple classifier fusion and data resampling to successfully boost the performance. Key word/key phrase spotting is used to optimally locate the password phrase. Numerous adaptation techniques increase the flexibility of the base system, and include: channel adaptation, fusion adaptation, model adaptation and threshold adaptation.

Subword-based speaker verification with multiple-classifier score fusion weight and threshold adaptation

The present invention relates to a robust speech processing method and system which models channel and noise variations with affine transforms to reduce mismatched conditions between training and testing. The affine transform relating the training vectors C.sub.k with the vectors for testing condition c.sub.k', is represented by the form:

Robust speech processing with affine transform replicated data

The present invention relates to a pattern recognition system which uses data fusion to combine data from a plurality of extracted features and a plurality of classifiers. Speaker patterns can be accurately verified with the combination of discriminant based and distortion based classifiers. A novel approach using a training set of a 'leave one out' data can be used for training the system with a reduced data set. Extracted features can be improved with a pole filtered method for reducing channel effects and an affine transformation for improving the correlation between training and testing data.

Speaker verification system using decision fusion logic

A method and apparatus for recognition of objects such as faces in images using signal compression techniques (i.e., coding techniques) in which a portion of the image which includes the object to be recognized (e.g., the face) is coded, and the resultant coded data is matched against previously coded and stored training data which makes up a known object database. A given object in an input image signal is matched to one of a plurality of known objects stored in a database, wherein the stored representation of each of the known objects comprises a codebook generated based on training image signals comprising the known object. A first illustrative embodiment comprises the steps of decomposing the given object into blocks; performing a plurality of encodings of the given object, each encoding comprising coding the object with use of one of the codebooks; determining a coding error for each encoding; and matching the given object to one of the known objects based on the coding errors. A second illustrative embodiment comprises the steps of decomposing the given object into blocks; generating a codebook corresponding to the given object based on the blocks; comparing the codebook corresponding to the given object with the codebooks corresponding to each of the known objects; and matching the given object to one of the known objects based on the comparison of the codebooks.

Growing community of inventors

Piscataway, NJ, United States of America

Xiaoyu Zhang