69 results about "Intelligent character recognition" patented technology

A character recognition system is described, in particular a system suitable for use in monitoring cargo container codes or vehicle number plates. An image of the code is first analyzed to extract potential characters. As part of this process, long horizontal and vertical line segments are filtered out. The extracted potential characters are then input to a two-level character recognition means. The first level comprises a neural network classifier that classifies a character into a smaller set of possible characters; and then the second level comprises another neural network classifier which identifies which character among the smaller set of possible characters the extracted character is.

A method and apparatus is disclosed for reading and decoding information extracted from a form. In the system of the present invention, packages are randomly placed on a conveyor belt, with their labels facing a two-camera subassembly. As the conveyor belt moves, the two-camera subassembly continuously takes images of the belt underneath the overhead camera. The design of the camera permits it to take a high resolution image of a non-singulated, unjustified package flow. A digital image of the packages within the field of view of the camera is then transferred to the processing system for analysis. The processing system identifies individual packages in the image, extracts them and then analyzes the information written on the package labels. The analysis process utilizes conventional Optical Character Recognition (OCR) and Intelligent Character Recognition (ICR) techniques to evaluate the information written on the package label. Once the information is decoded, the system either accesses a database record associated with the decoded machine-readable code, or creates a new record. When an unknown word image is encountered, the field-specific recognition process is aided by use of lexicon information, optimized based on installation-specific or user-specific criteria. The lexicon information is continuously revised based on processed form information. In a preferred embodiment, verified destination addresses associated with a user are alphabetized or rank-ordered based on frequency of occurrence. It is only after the system determines that the originating user is not stored in the database does it resort to the ZIP+4 or similar database to verify a destination address.

The invention discloses an off-line handwritten form Chinese character recognition method and belongs to the technical field of Chinese character recognition and extraction. The method comprises the steps of data collection, preprocessing, feature extraction and recognition extraction. A classification and recognition standard in the Chinese character recognition method comprises that a sample to be tested is classified according to a minimum distance standard, the feature recognition is performed based on a 4*4 elastic mesh, in rough classification recognition, the number of feature points such as end points, inflection points and cross points in each of meshes are conducted statistics on, and weight sum is respectively performed, in a second classification recognition, the number of transverse, vertical, left-falling and right-falling stroke vectors in each of the meshes is recorded, and handwritten form Chinese characters are recognized and extracted. According to the method, both off-line handwritten form Chinese characters and print form Chinese characters can be recognized, the feature extraction method of the Chinese character recognition method is simple, and the recognition speed is fast.

A method and apparatus is disclosed for reading and decoding information extracted from a form. In the system of the present invention, packages are randomly placed on a conveyor belt, with their labels facing a two-camera subassembly. As the conveyor belt moves, the two-camera subassembly continuously takes images of the belt underneath the overhead camera. The design of the camera permits it to take a high resolution image of a non-singulated, unjustified package flow. A digital image of the packages within the field of view of the camera is then transferred to the processing system for analysis. The processing system identifies individual packages in the image, extracts them and then analyzes the information written on the package labels. The analysis process utilizes conventional Optical Character Recognition (OCR) and Intelligent Character Recognition (ICR) techniques to evaluate the information written on the package label. Once the information is decoded, the system either accesses a database record associated with the decoded machine-readable code, or creates a new record. When an unknown word image is encountered, the field-specific recognition process is aided by use of lexicon information, optimized based on installation-specific or user-specific criteria. The lexicon information is continuously revised based on processed form information. In a preferred embodiment, verified destination addresses associated with a user are alphabetized or rank-ordered based on frequency of occurrence. It is only after the system determines that the originating user is not stored in the database does it resort to the ZIP+4 or similar database to verify a destination address.

A method for intelligent speech recognition for premature enunciators is disclosed. The method allows for speech recognition systems to adjust to commands of premature enunciators, who speak before the system is ready receive speech input. The method comprises activating the speech recognition system, informing the user that the speech recognition system is ready to receive speech input, and initiating a listening period. The method further comprises receiving input from a user before the system is ready to receive such input, identifying the user as a premature enunciator, and adjusting the system to allow for earlier detection of user speech input.

The invention relates to the field of image character recognition, and particularly relates to an image character sequence recognition system based on a recurrent neural network; the system comprises an image character input module, a convolutional neural network and a recurrent neural network classifier; the convolutional neural network extracts characteristics of a to-be-recognized character sequence input by the image character input module, and inputs to the recurrent neural network classifier; and the recurrent neural network classifier, according to sample characteristic data and output of the last moment, realizes continuous recognition of the character sequence. According to the system disclosed by the invention, the shortage that picture segmentation is carried out before OCR recognition is overcome, the earlier stage processing of the image character recognition is simplified, and a language model does not need to be constructed additionally to carry out optimization processing on a recognition result; while the recognition accuracy rate of character and word sequences is improved better, the processing efficiency of the character recognition is obviously improved; and the system has wide application prospect in the field of image character recognition.

The invention relates to a character recognition method based on SIFT feature and gray scale difference value histogram feature in the technical field of image character recognition. The character recognition method comprises the following steps: firstly, establishing a Chinese character template gallery, wherein pictures with different font classes, character sizes and directions under various complex backgrounds are stored into the Chinese character template gallery; secondly, performing feature extraction on the pictures in the template gallery; thirdly, calculating the feature vector of a Chinese character to be recognized; and fourthly, searching for Chinese characters which are mostly close to the Chinese character to be recognized from the template gallery. The recognition method does not need to pretreat the pictures and perform large-scale machine learning on massive character features, describes and matches characters in virtue of local features creatively and by skillful combination of gray scale statistic information in regions, has good character recognition effect as shown by the experimental result, and can well recognize different fond types of the same character.

The invention discloses an automatic detection method for a Chinese character area of a shop sign in a natural scene. The automatic detection method comprises the following steps of: A, acquiring vectors V1 and V2 which reflect main pixel colors of a sampling region image; B, determining a background outline of an original image by using the vectors V1 and V2; C, dividing an HSV (Hue, Saturation and Value) color space into eight color spaces including black, white, red, yellow, green, cyan, blue and fuchsine; C, carrying out color separation on an original image to obtain eight color distribution binary images; carrying out Chinese character connected area analysis on the eight color distribution binary images to obtain eight text line binary images; F, filtering text lines, which do not conform to a Chinese character writing rule, in each text line binary image to obtain refined Chinese character areas; G, mixing the refined Chinese character areas of different colors to obtain a Chinese character area of the original image. By using the automatic detection method for the Chinese character area of the shop sign in the natural scene, the detection accuracy of the Chinese character area can be efficiently improved, the automatic detection method has significant importance for application of Chinese character recognition in the natural scene.

The invention, which belongs to the technical field of image processing, discloses a depth-learning-based Chinese character recognition system realizing method. Preprocessing, segmentation, identification, reorganization are carried out on a picture with characters to form a text, thereby realizing conversion from the picture to the text. The method comprises: (1), picture pretreatment is carried out; to be specific, graying, binarization, and inclination correction are carried out on an inputted original picture to obtain a regular picture; (2), picture segmentation is carried out; to be specific, segmentation is carried out on an inputted picture with lots of characters, the picture is segmented into rows and then each row is segmented into single character; (3), picture identification is carried out; to be specific, the segmented single character is identified and an identification module is invoked for each single character picture to obtain an identification character result; and (4), text reorganization is carried out; to be specific, the obtained single character identification results are combined according to a sequence, correction is carried out, and then an identification result text segment is produced. With the image processing technique, graying and binarization are completed; and the generalization ability and the anti-interference capability are high.

An electronic forms management system is provided. The electronic forms management system includes a publishing unit, a data input unit and a server. The publishing unit is adapted to publish the form templates to the server. The data input unit is connected to the server, and is adapted to receive input data. The server is adapted to process the published form templates and the input data. In addition, the server includes an Intelligent Character Recognition (ICR) framework which is being adapted to receive one or more ICR engines for recognizing the input data.

One embodiment of the present invention provides a system that facilitates displaying automatically recognized text, wherein the automatically recognized text includes optical character recognition (OCR) text or intelligent character recognition (ICR) text. During operation, the system receives a source image containing text from a user. Next, the system performs a character-recognition operation on the source image to produce recognized text. The system then locates a data field within the source image. Next, the system identifies a subset of the recognized text that corresponds with the data field. After that, the system displays the source image to the user. At the same time, the system displays the subset of the recognized text in close proximity to the data field in the source image to facilitate comparing the recognized text with the text contained in the source image.

The invention discloses a document text extraction method and device, belonging to the field of data processing. The method includes: step 1: analyzing the document, obtaining the corresponding information of the font in the document, and obtaining a character mapping table according to the corresponding information; step 2: obtaining the font image corresponding to each character according to the font corresponding information; step 3: cutting the font image to obtain The inked area corresponding to the font image; Step 4: Perform character recognition on the inked area to obtain the recognition result of each character; Step 5: Update the character mapping table according to the recognition result, and extract text information from the document according to the updated character mapping table . The invention improves the flow of data processing, and also reduces the workload of data processing, so that the randomly coded packaged fonts will not become an obstacle to data processing. For a specific layout document, the correct text information can be obtained without identifying the page image, which minimizes manual intervention and preserves the format and logic information of the document.

The invention belongs to the image character recognition field and relates to an optical character sequence recognition method. According to the method of the invention, CNN (Convolutional Neural Network) and RNN (Recurrent Neural Network) technologies are adopted; feature extraction is performed on a whole picture containing a plurality of characters through a CNN; identical features are transmitted to an RNN so as to be subjected to repeatedly recursive use; and continuous prediction of the plurality of characters can be realized. With the method adopted, a defect that picture segmentation is required before OCR (optical character recognition) can be eliminated, the early-stage processing process of picture character recognition can be simplified, and the efficiency of character recognition can be significantly improved; and since the RNN recursively uses output data of the last round, and in model training, a language model of dependency relationships between characters and words can be obtained through learning, and therefore, a step in an OCR method, according to which a language model is required to be additionally built for post-processing after individual characters are recognized, can be avoided; and therefore, the recognition accuracy of character and word sequences can be better improved, and the processing efficiency of character recognition can be further improved.

Methods and apparatuses are described for automated execution of computer software using intelligent speech recognition techniques. A server captures a digitized voice segment from a remote device, the first digitized voice segment corresponding to speech submitted by a user of the remote device during a voice call. The server analyzes the digitized voice segment to convert the digitized voice segment into text and extract a set of keywords from the converted text. A client computing device displays the extracted keywords in a user interface of a display device. The server determines one or more computer software applications accessible to the client computing device and selects least one of the computer software applications that include functionality responsive to the keywords. The client computing device executes the functionality of the selected computer software applications that is responsive to the keywords.

The invention discloses an early child education system and a method based on Chinese character recognition. The system comprises a user login unit, a template unit, an image matching unit, an assessment unit and an output unit, wherein the user login unit is used for inputting personal information of the kid; the template unit is used for generating a corresponding Chinese character template according to the personal information of the kid in the user login unit; the image matching unit is used for being matched with the Chinese character template and finding a corresponding Chinese character related image; the assessment unit is used for building an assessment model through the Chinese character template in the template unit, and the assessment model is used for judging the Chinese character recognition correct rate by the kid; and the output unit is used for outputting an electronic learning report according to the kid Chinese character recognition degree in the assessment model. According to the early child education system provided by the invention, the kid character recognition effects are improved, the image corresponding to the Chinese character is changed into an image facilitating character recognition by the kid, and character recognition by the kid is more efficient; the information bearing amount is large enough through the image matching unit; and the character recognition condition by the kid is assessed via the assessment unit.

The invention discloses an end-to-end indefinite-length character recognition method and system, which is characterized in that: according to pictures and text character tags in a data set, a text detection model and a text recognition model are trained by using a depth neural network; Through the text detection model, the position of the text in the picture is located. Identify the specific content of the text through the text recognition model; The two models are combined to recognize the characters in the picture and locate the positions of the characters in the picture, so that the characters in the picture can be recognized and the positions of the characters in the picture can be known, which can be applied to the directions of the traditional character recognition and the certificate recognition, and the efficiency and accuracy of the character input can be greatly improved.

In an intelligent character recognition (ICR) method for recognizing hand-written text images using a long-short term memory (LSTM) recurrent neural network (RNN), text images are segmented into text line images, and the text lines images are pre-processed to normalize the line height and to equalize the word spacings in each text line. Both training images used to train the RNN network and test images containing text to be recognized by the trained RNN network are pre-processed to have identical heights and identical word spacings between words. This method improves character recognition accuracy.

The invention provides a medical-bill ICR (Intelligent Character Recognition) system and a medical-bill recognition method thereof. The medical-bill ICR system includes: a bill picture service collection module, which captures key points of a bill page; an ICR pre-processing module, which carries out filtering on interference graphics on the bill page; an ICR core engine module, which carries outrecognition of various characters of printed bodies and handwritten bodies; and a character library+industry database module, which stores text data of the printed bodies and the handwritten bodies, and is called by the ICR core engine module. Automated inputting, storage and management of various invoices are realized, a highly efficient, intelligent and convenient personalized solution is provided, work efficiency can be improved, and costs can be reduced.

The invention discloses an automatic recognition method of a digital instrument with a decimal point based on a convolution neural network, which comprises the following steps: the collected LED picture sample of the digital instrument is divided into independent LED character pictures; the LED character pictures are sent into a network model for training after being pretreated; the LED characterpictures are sent into a network model for training. The pictures to be recognized are input into the trained network model for recognition. The network model is composed of LED character convolutionneural network model and decimal point convolution neural network model. The pretreatment process of LED character image includes LED digital sample image pretreatment step and decimal point sample image pretreatment step. The invention scales the LED character picture containing decimal points, then segments the LED character picture and sends the LED character picture to the network model for training, that is, the regression positioning problem is converted into a classification problem. Because the decimal point and LED character recognition are two different networks, the results of modelrecognition will not interfere with each other, so it is more flexible in network debugging.

A character recognition device includes circuitry that is configured to remove duplicate successive points of a plurality of points in a handwritten stroke to form an enhanced handwritten stroke; space the plurality of points a uniform distance apart; detect primary strokes and secondary strokes of the enhanced handwritten stroke; merge the primary strokes; generate a primary merged stroke; extract raw point-based features from local features of the primary merged stroke; extract statistical features from computed statistics associated with the raw point-basal features to form primary merged stroke features; train and classify data from the primary merged stroke features and secondary stroke features to form stroke models; determine a plurality of primary merged stroke model candidates from the stroke models; compute a likelihood value for a combined set of primary stroke candidates and a set of secondary stroke candidates; and determine the handwritten stroke from the computing.

The present invention provides a five-dimensional lithology identification method based on conventional logging data. The five-dimensional lithology identification method based on conventional logging data includes: step 1, selecting multiple logging curves sensitive to lithology as identification parameters , extract its logging response values ​​for various lithologies; step 2, establish standard charts for different lithologies according to the data extracted in step 1; and step 3, extract the discriminant parameters of the samples to be predicted, according to the establishment of The standard plates of different lithologies are used for sample prediction. This method establishes a series of practical and effective lithology identification charts based on the five-dimensional analysis of conventional logging, which can assist the lithology identification of volcanic rocks in oil fields, and provide guidance and reference for lithology identification of other complex reservoirs.

The invention discloses an intelligent speech recognition to semantic comprehension result statistical system and method. The system includes a running state acquisition unit, a test voice selection unit, a response result acquisition unit, a test semantic acquisition unit and a test unit. An identification text which is closer to voice contents can be obtained by adopting the schemes; the accurate rates of semantic services can be calculated after contrast; and the system and method are convenient and fast, and can automatically see reports, so that efficiency can be greatly enhanced, and split speech and semantic test activities can be got together.

The invention belongs to the technical field of intelligent recognition, and discloses a face recognition security alarm method based on an embedded artificial intelligence chip, which can perform video acquisition, transmission, storage, face positioning, face recognition and sending of alarm information and video information based on the embedded artificial intelligence chip. Specifically, the method comprises: collecting video information through the video data collection module, transferring uploaded data through the data transfer module, judging whether someone invades or not through facerecognition conducted by the recognition and judgment module through the local development board, judging and starting an alarm after recognition, issuing alarm information to a user, and the user can obtain an alarm result in time and process the alarm result. According to the invention, real-time face recognition is directly carried out on a transmitted video without sending data to a third-party server for recognition, so that real-time monitoring alarm and evidence storage can be realized, and a user can directly use a mobile phone APP to play and display a real-time picture.

Embodiments of a computer system, a method, and a computer-program product (e.g., software) for use with the computer system are described. These embodiments may be used to identify and correct errorsin financial information that was extracted using character-recognition software, such as optical character recognition software and / or intelligent character recognition software. In particular, potential errors may be identified by comparing the financial information for a current financial transaction of a user with expected financial information from one or more previous financial transactionsof the user. Error metrics for these potential errors may be determined and used to correct at least some of the potential errors. For example, values of the Levenshtein edit distance may be determined based on the comparison, and one or more potential errors associated with one or more minimum values of the Levenshtein edit distance may be corrected.

The invention discloses a short message generating method and a cellphone. The short message generating method comprises the following steps: a, providing a menu option used for starting a camera in a short message editing interface, b, obtaining pictures with required courier number information through the camera, and c, conducting recognition processing to the obtained pictures and generating a short message which includes required courier number content. The recognition processing is conducted through an optical character recognition mode, an image character recognition mode or an intelligent character recognition mode in the step c. The generated short message in the step c further includes pre-defined information content. The short message generating method and the cellphone can rapidly generate the short message which includes the courier number content required by a user, the operation that the user inputs related courier number content manually is avoided, increases intelligence and practical applicability of the cellphone and improves user experience.

Disclosed herein is a method that improves the performance of handwriting recognition by calculating path scores so as to identify the path with the highest score as the basis for interpreting handwritten characters. Specifically, the method comprises the following steps: detecting connected regions in an input image comprising handwritten characters; determining a plurality of segmentation positions of the input image; obtaining a plurality of recognition results for each segment of each path in the input image, wherein each recognition result represents a character candidate for the segment and each path comprises one or more segments; obtaining a plurality of scores corresponding to the recognition results; calculating scores for each path in the input image based on segment lengths and the scores corresponding to the recognition results; and using the path with the highest score to interpret the handwritten characters in the input image.

The invention discloses a two-factor identity authentication method based on Chinese character format information. The two-factor identity authentication method comprises the steps that a user inputs simplified Chinese characters into a manual tablet; Chinese character recognition is conducted on the simplified Chinese characters through the manual tablet based on the Chinese character skeleton recognition algorithm, so that binary coding information is generated; encryption is conducted on the binary coding information through the SM3 encryption algorithm based on a time factor, so that a number-format challenge is generated; the user inputs the number-format challenge into a hardware token manually; encryption conversion is conducted on the number-format challenge through the hardware token by means of the SM3 algorithm based on the time factor, so that a dynamic password is obtained; the user inputs the dynamic password into an identity authentication server, and the identity authentication server authenticates the identity of the user according to the dynamic password. According to the two-factor identity authentication method based on the Chinese character format information, the challenge of two-factor identity authentication is generated based on the absoluteness and the uniqueness of the Chinese character format information, and therefore it is guaranteed that the challenge is not prone to being influenced by single parameters such as time.

An apparatus and a method for high-speed character recognition are disclosed. The character recognition method includes receiving a bit-stream decoded base on a symbol matching encoding scheme where the bit-stream including a symbol dictionary and a symbol information which is information of symbols included in an original image; decoding the symbol dictionary included in the bit-stream; performing a character recognition process of each of plural of symbols included in the decoded symbol dictionary; decoding the symbol information after completing the character recognition process; and generating a text file of the original image by using the result of the character recognition process and the decoded symbol information.

The invention discloses a method for intelligent speech recognition of a contact name in an address list. The method comprises the following steps that step 1, a dedicated database is created, whereinthe database involves creation of an address list contact pinyin library and creation of a voice search contact sorting method; step 2, a user searches for contacts by voice passwords, intelligent equipment converts the voice passwords into a text form to obtain corresponding text search results; step 3, the text search results are converted into pinyin, and the address list contact pinyin library in the database is compared to obtain corresponding contact search results. The method for the intelligent speech recognition of the contact name in the address list has the advantages that the comparison between the name pinyin of speech input and the corresponding database is facilitated, finding of corresponding contacts is facilitated, when multiple homophonic contacts are met, the homophonic contacts can be distinguished better, the contacts can be rapidly found, the time consumed by manual finding is reduced, the speech recognition rate of names is increased, and the error rate of speech recognition is reduced, thereby improving the user experience.