US 12,361,105 B2Granted

Signature verification

AssigneeMicrosoft Technology Licensing, LLC

InventorsTianyi Chen, Sheng Yi

Life of this patent

Estimating the term…

Loading patent…

Abstract

Methods, systems, and computer program products are provided for signature verification. Signature verification may be provided for target signatures using genuine signatures. A signature verification model pipeline may extract features from a target signature and a genuine signature, encode and submit both to a neural network to generate a similarity score, which may be repeated for each genuine signature. A target signature may be classified as genuine, for example, when one or more similarity scores exceed a genuine threshold. A signature verification model may be updated or calibrated at any time with new genuine signatures. A signature verification model may be implemented with multiple trainable neural networks (e.g., for feature extraction, transformation, encoding, and/or classification).

Claims20

Dependent claims

A signature verification system, comprising:

a processor;
and
a memory device that stores program code for execution by the processor, the program code comprising:
a model comprising a chain of long short term memory (LSTM) cells, the chain comprising a first LSTM cell and a plurality of subsequent LSTM cells, wherein the model:
transforms target signature features of a target signature into an encoded target feature matrix comprising:
a first column corresponding to a first input of the target signature at a first timestamp, and
a second column corresponding to a second input of the target signature at a second timestamp;
utilizes the chain of LSTM cells to
transform the encoded target feature matrix into an encoded target feature vector by:
utilizing the first LSTM cell to transform the first column into a first target feature vector, and
utilizing a last LSTM cell in the plurality of subsequent LSTM cells to transform the second column into the encoded target feature vector based on an output of a previous LSTM cell in the chain of LSTM cells;
and
verifies the target signature based on the encoded target feature vector and an encoded genuine feature vector generated based on the chain of LSTM cells and a first genuine signature.

10 more elements

A method comprising:

transforming target signature features of a target signature into an encoded target feature matrix comprising:
a first column corresponding to a first input of the target signature at a first timestamp, and
a second column corresponding to a second input of the target signature at a second timestamp;
utilizing a chain of long short term memory (LSTM) cells to transform the encoded target feature matrix into an encoded target feature vector by:
utilizing a first LSTM cell in the chain to transform the first column into a first target feature vector, and
utilizing a last LSTM cell in a plurality of LSTM cells subsequent to the first LSTM cell to transform the second column into the encoded target feature vector based on an output of a previous LSTM cell in the chain;
and
verifying the target signature is genuine based on the encoded target feature vector and an encoded genuine feature vector generated based at least on the chain of LSTM cells and a first genuine signature.

5 more elements

A computer-readable storage medium having program instructions recorded thereon that, when executed by a processing circuit, perform a method comprising:

transforming target signature features into an encoded target feature matrix comprising:
a first column corresponding to a first input of the target signature at a first timestamp, and
a second column corresponding to a second input of the target signature at a second timestamp;
utilizing a chain of long short term memory (LSTM) cells to transform the encoded target feature matrix into an encoded target feature vector by:
utilizing a first LSTM cell in the chain to transform the first column into a first target feature vector, and
utilizing a last LSTM cell in a plurality of LSTM cells subsequent to the first LSTM cell to transform the second column into the encoded target feature vector based on an output of a previous LSTM cell in the chain;
and
verifying the target signature is genuine based on the encoded target feature vector and an encoded genuine feature vector generated based at least on the chain of LSTM cells and a first genuine signature.

5 more elements

Dependencies are inferred from the claim text.

Facts

Application: 17/820,036Filed 16 Aug 2022 · continuation
Priority: 28 Oct 2019
Published: US 2022/0392265 A18 Dec 2022
Granted: 15 Jul 2025
Term ends: Not estimatedThe source records don’t support an estimate.

Classification21 codes

G06F 21/32
User authentication: using biometric data, e.g. fingerprints, iris scans or voiceprints
G06F 18/214
Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation: generating training…
G06F 18/253
Fusion techniques: of extracted features
G06N 3/0442
Recurrent networks, e.g. Hopfield networks: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]

Family

Application16/665,921Parent of this continuation · filed 28 Oct 2019
US 2022/0392265 A1Pre-grant publication · published 8 Dec 2022
US 11,450,150 B2Parent of this continuation
US 2021/0124905 A1Published application
US 12,361,105 B2This patent

From the US filing history. Foreign family members aren’t listed.