International Business Machines Corporation

Michigan State University

Ford Global Technolgoies, LLC

Anil K Jain

Rudolf Maarten Bolle

Sharathchandra Umapathirao Pankanti

Lin Hong

Matthew Michael Pohlman

Jacob O Miller

Jason R Gilder

Kwaku O Prakah-Asante

Krishnaswamy Venkatesh Prasad

Arun Ross

Debayan Deb

Kai Cao

Steven Grosz

Charlie Lougheed

Joshua J Engelsma

Doug Meil

Brad Jarrell

A method and system for fingerprint recognition includes generating a first set of feature embeddings for a first image at a feature extractor, generating a second set of feature embeddings for a second image at the feature extractor, generating a third set of embeddings at a first classification head from the first set of feature embeddings, generating a fourth set of embeddings at a second classification head from the first set of feature embeddings, generating a fifth set of embeddings at the first classification head from the second, set of feature embeddings, generating a sixth set of embeddings at the second classification head from the second set of feature embeddings, generating a first similarity score based on the third and fourth sets of embeddings and the fifth and sixth set of embeddings and generating a similarity indicator based on the first similarity score.

Attention driven and latent fingerprint recognition network

A passive authentication method includes, in response to receiving a requested action from a first user, obtaining a set of sensor data and categorizing first sensor data of the set of sensor data into a first modality of a set of modalities. The method includes, for the first modality of the set of modalities, determining a distance value by applying a first modality model to the first sensor data and comparing the distance value to a first verified value of the first user for the first modality. The method includes, based on the comparison, determining a first authentication decision of the distance value. The method includes, in response to the first authentication decision indicating the first sensor data corresponds to the first user, performing the requested action.

Integrated systems and methods for passive authentication

A computer-implemented method for generating a representation for a fingerprint includes receiving, by a computer processor, an image of a given fingerprint. The method extracts particular attributes for the given fingerprint from the image using a first neural network. The first neural network is trained to identify particular attributes in fingerprints and constructs a first feature vector from the extracted particular attributes, where the first feature vector has a first fixed length. The method includes extracting textural features of the given fingerprint from the image using a second neural network, where the second neural network is trained to identify textural features that are not limited to particular attributes and constructing a second feature vector from the extracted textural features, where the second feature vector has a second fixed length. The method includes concatenating the first feature vector with the second feature vector to form a representation for the given fingerprint.

Fixed length fingerprint representation

A system links data objects associated with a common entity and includes at least one processor. The system compares data objects within one or more source systems to identify candidate data objects associated with a corresponding common entity based on information within those data objects. The candidate data objects are analyzed based on supplemental temporal information within the one or more source systems pertaining to the candidate data objects to determine resulting data objects associated with the corresponding common entity. The resulting data objects are linked to form a set of data objects for the common entity. Embodiments of the present invention further include a method and computer program product for linking data objects associated with a common entity.

Association of entity records based on supplemental temporal information

In a method for measuring healthcare quality, a plurality of computers operating in parallel receive from a patient dataset representative of patients within a healthcare organization, the patient dataset being distributed across the plurality of computer as subsets of patient datasets. The plurality of computers receive at least one healthcare quality measure definition comprising a numerator and a denominator. The plurality of computers identify, in the subsets of patient datasets, patient data that corresponds to the measure definition. The plurality of computers attribute the identified patient data to at least one provider based on an attribution rule. The plurality of computers calculate at least one healthcare quality measure for the at least one provider, according to the healthcare quality measure definition, based on the attributed patient encounters. The plurality of computers store the at least one calculated healthcare quality measure in a measure dataset.

System and method for measuring healthcare quality

The invention is a system and method for deriving a single dimensional representation for a set of points e.g.,minutiae, in image of a two dimensional pattern of lines, e.g. a fingerprint, by creating a one dimensional (string) representation of one or more points (e.g., minutiae) and the respective attributes of each point therein. A landmark point is selected from the two dimensional image, preferably from the set of the points to be represented in single dimension. The relationships of each of the points with reference to the landmark determines a linear order for the points and the attributes associated with each point.

System and method for deriving a string-based representation of a fingerprint image

The invention is a system and method for determining alignment parameters (rotation angle and translation vector) between two images of scene based on the attributes of segments associated with a pair of corresponding points, each point extracted from each of the images.

Determining an alignment estimation between two (fingerprint) images

A reliable estimate of image pixel (fingerprint ridge) orientations, and block orientations for a given block is determined by finding an optimal block size in an area of an image, e.g., a fingerprint image. An initial block size is chosen. Then a block direction is determined, e.g., by analyzing intensity gradients in two directions, preferably two orthogonal directions, at a plurality (preferably all) of the pixels locations in the block. Block directions of blocks in a given area of the images (fingerprint) are compared. If the compared directions are within a tolerance of one another, the determined direction of the block is considered correct. However, if the compared directions are not within a tolerance of one another, the block size is changed and the process is repeated.

System and method for determining block direction in fingerprint images

Disclosed in a system and method for segmenting foreground pixel from background pixels in a digital image, specifically a fingerprint image. First the image is divided into blocks of optimal size. Then, for a given block, an overall gradient magnitude and a projected gradient magnitude is determined. The overall gradient magnitude is a measure of the amount of intensity variation in the current block. If a large number (or all) of the pixels have a direction that is the same as the overall gradient direction, the projected gradient magnitude will be nearly equal (will be equal) to the overall gradient magnitude. This indicates that the current block is a block in the foreground of the image. A relationship, e.g., a ratio, of the over gradient magnitude to the overall projected gradient magnitude is taken certainty level that the current block is in the foreground or background of the image. Connected blocks making up the largest region of foreground blocks are considered the foreground region, the rest of the blocks/regions are discarded (considered as background).

System and method for identifying foreground and background portions of digitized images

The invention is a system and method for matching a two dimensional pattern of lines, e.g. a fingerprint, by creating a one dimensional representation of one or more points (e.g., minutiae) on the lines in the patterns to be matched. The one dimensional representations are created by finding a corresponding reference point (minutiae) in each of the patterns and creating an index of other points in each of the respective patterns with respect to the reference point (minutiae). The indexes are adaptively compared taking into account the elastic distortions of the lines.

System and method for matching (fingerprint) images an aligned string-based representation

The invention is a system and method for determining a correspondence between a point selected from a set of points extracted from one image and an another point selected from the second set of points extracted from another image based on similarity of relationships of attributes associated with segments attached to the points.

Growing community of inventors

Solon, OH, United States of America

Anil K Jain