US 12,361,737 B2
Method of recognizing magnetic ink characters in ATM
Hyun Su Kim, Seoul (KR)
Assigned to Hyosung TNS Inc., Seoul (KR)
Filed by Hyosung TNS Inc., Seoul (KR)
Filed on Sep. 1, 2022, as Appl. No. 17/901,709.
Claims priority of application No. 10-2022-0057443 (KR), filed on May 10, 2022.
Prior Publication US 2023/0368554 A1, Nov. 16, 2023
Int. Cl. G06V 30/224 (2022.01); G06Q 20/10 (2012.01)
CPC G06V 30/2253 (2022.01) [G06Q 20/1085 (2013.01)] 6 Claims
OG exemplary drawing
 
1. A method of recognizing magnetic ink characters, and the method comprising:
obtaining character-based waveforms of magnetic ink characters on a check by a magnetoresistive (MR) sensor of an automated teller machine that handles bills and checks;
obtaining a waveform group corresponding to each of the magnetic ink characters from a plurality of waveforms by executing a waveform authentication algorithm on a microprocessor of the automated teller machine to determine a waveform group to which each of the character-based waveforms corresponds according to a preset authentication reference for the waveform groups, based on each of the character-based waveforms acquired by recognizing the magnetic ink characters with the MR sensor, wherein the preset authentication reference for the waveform groups is derived through reiterative deep learning of character-based waveform data of the MR sensor that are acquired on a per-character basis from legitimate checks on each of which magnetic ink characters are properly printed, the reiterative deep learning using Convolution Neural Network (CNN);
obtaining an image of the check by a contact image sensor of the automated teller machine;
performing optical character recognition (OCR) on the obtained image by the microprocessor of the automated teller machine to obtain results of the OCR; and
comparing, for verification, the results of the OCR and waveform groups of the magnetic ink characters.