# US 11,113,519 B2 · Character recognition apparatus, character recognition program, and character recognition method > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,113,519 B2 - Document ID: US11113519B2 - Status: Granted - Kind: B2 - Application number: 16/502,085 - Filed: 2019-07-03 - Published: 2020-01-30 - Granted: 2021-09-07 - Earliest claimed: 2018-07-27 - Assignees: Fujitsu Limited - Applicants: Fujitsu Limited - Inventors: Yusuke Hamada; Misato Hishi - Web page: https://patentgrep.com/patent/US11113519B2 ## CPC - G06V 10/82: using neural networks - G06F 18/2178: {based on feedback of a supervisor} - G06V 20/62: Text, e.g. of license plates, overlay texts or captions on TV images - G06V 30/153: {using recognition of characters or words} - G06V 30/19173: {Classification techniques} - G06V 30/262: using context analysis, e.g. lexical, syntactic or semantic context - G06V 30/414: Extracting the geometrical structure, e.g. layout tree; Block segmentation, e.g. bounding boxes for graphics or text - G06V 30/10: Character recognition - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract A character recognition method for a moving image includes extracting a region corresponding to a character string included in each frame of a moving image to be recognized. The method reads a character string from the extracted region and corrects the character string read from each frame, based on appearance rule information that specifies an appearance rule of a character string corresponding to the order of the frames, such that the appearance order of the read character string conforms to the appearance rule. ## Claims ### Claim 1 A character recognition apparatus comprising: a memory, and a processor coupled to the memory and configured to: extract a region corresponding to a character string included in each of frames of a moving image to be recognized; read a character string from the extracted region of each of the frames; correct the character string read from each of the frames, based on appearance rule information that specifies an appearance rule of a character string corresponding to an order of the frames, such that an appearance order of the corrected character string conforms to the appearance rule; and correct the character string based on a comparison between the character string read from each frame and an estimation result estimated by inputting the character string read from each frame to a learning model obtained by learning the appearance rule information. ### Claim 2 (depends on 1) The character recognition apparatus according to claim 1, wherein the character string is a time code corresponding to a time in the moving image. ### Claim 3 (depends on 1) The character recognition apparatus according to claim 1, wherein the processor is further configured to extract a region in a frame to be extracted, based on a moving average of regions extracted from a predetermined number of frames nearest to the frame to be extracted. ### Claim 4 A non-transitory computer-readable storage medium having stored therein a character recognition program for causing a computer to execute a process comprising: extracting a region corresponding to a character string included in each of frames of a moving image to be recognized; reading a character string from the extracted region of each of the frames; correcting the character string read from each of the frames, based on appearance rule information that specifies an appearance rule of a character string corresponding to an order of the frames, such that an appearance order of the corrected character string conforms to the appearance rule; and correcting the character string based on a comparison between the character string read from each frame and an estimation result estimated by inputting the character string read from each frame to a learning model obtained by learning the appearance rule information. ### Claim 5 (depends on 4) The storage medium according to claim 4, wherein the character string is a time code corresponding to a time in the moving image. ### Claim 6 (depends on 4) The storage medium according to claim 4, wherein the extracting includes extracting a region in a frame to be extracted, based on a moving average of regions extracted from a predetermined number of frames nearest to the frame to be extracted. ### Claim 7 A character recognition method to be performed by a computer, the method comprising: extracting a region corresponding to a character string included in each of frames of a moving image to be recognized; reading a character string from the extracted region of each of the frames; correcting the character string read from each of the frames, based on appearance rule information that specifies an appearance rule of a character string corresponding to an order of the frames, such that an appearance order of the corrected character string conforms to the appearance rule; and correcting the character string based on a comparison between the character string read from each frame and an estimation result estimated by inputting the character string read from each frame to a learning model obtained by learning the appearance rule information. ### Claim 8 (depends on 7) The character recognition method according to claim 7, wherein the character string is a time code corresponding to a time in the moving image. ### Claim 9 (depends on 7) The character recognition method according to claim 7, wherein the extracting includes extracting a region in a frame to be extracted, based on a moving average of regions extracted from a predetermined number of frames nearest to the frame to be extracted.