CN113128296B - Electronic handwriting signature fuzzy label recognition system - Google Patents
Electronic handwriting signature fuzzy label recognition system Download PDFInfo
- Publication number
- CN113128296B CN113128296B CN201911422774.2A CN201911422774A CN113128296B CN 113128296 B CN113128296 B CN 113128296B CN 201911422774 A CN201911422774 A CN 201911422774A CN 113128296 B CN113128296 B CN 113128296B
- Authority
- CN
- China
- Prior art keywords
- signature
- data
- server
- handwriting
- classification
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 claims abstract description 21
- 238000012549 training Methods 0.000 claims abstract description 17
- 230000009467 reduction Effects 0.000 claims abstract description 10
- 238000010606 normalization Methods 0.000 claims abstract description 9
- 238000002372 labelling Methods 0.000 claims abstract description 7
- 238000012545 processing Methods 0.000 claims abstract description 6
- 238000005070 sampling Methods 0.000 claims description 15
- 238000013528 artificial neural network Methods 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 8
- 230000001133 acceleration Effects 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 5
- 239000000284 extract Substances 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 abstract description 7
- 238000011161 development Methods 0.000 abstract description 2
- 238000012423 maintenance Methods 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 10
- 230000008569 process Effects 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 3
- 230000001186 cumulative effect Effects 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 238000000513 principal component analysis Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 230000001815 facial effect Effects 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000012790 confirmation Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/30—Writer recognition; Reading and verifying signatures
- G06V40/33—Writer recognition; Reading and verifying signatures based only on signature image, e.g. static signature recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/30—Writer recognition; Reading and verifying signatures
- G06V40/37—Writer recognition; Reading and verifying signatures based only on signature signals such as velocity or pressure, e.g. dynamic signature recognition
- G06V40/394—Matching; Classification
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Collating Specific Patterns (AREA)
- Image Analysis (AREA)
- Character Discrimination (AREA)
Abstract
An electronic handwriting signature fuzzy-labeling recognition system comprises a signature terminal, a server and storage equipment, wherein the signature terminal submits records of electronic handwriting signature contents to a user, handwriting graphics are generated, and a normalization model is built. And obtaining a training model after dimension reduction processing, forcibly classifying by using a multi-classification method after the server is trained by the training model, performing fuzzy label classification, taking the result with the threshold value being greater than 0.8 calculated by the kernel function as fuzzy label classification data, and outputting the result. The electronic handwriting signature fuzzy-labeling recognition system provided by the invention has the advantages that the calculated amount required by accurate recognition is obviously reduced, the purchasing cost and the operation and maintenance difficulty are reduced along with the great reduction of the calculated amount, and a foundation is laid for the development and application of handwriting recognition technology.
Description
Background
In information technology, an authentication mode for identity is a continuously developed process. With the continuous upgrading and updating of computer processors, the calculated amount of the computer processor is also rapidly developed, and the 6-bit digital password which is commonly adopted at once is not safe as an authentication measure, but is combined with a complex password, so that people are hard to memorize. Therefore, the identity authentication mode is more prone to develop in the direction of adopting biological characteristics such as face recognition, fingerprint recognition and the like. However, facial features, fingerprints, etc. are not traditional authentication methods in human civilization, and users cannot avoid contradiction and objection when receiving these new authentication methods. At present, a commonly accepted identity authentication mode in society is handwriting signature. On important legal efficacy documents such as contracts, attorney and agreements, personal handwriting signature is an identity confirmation mode approved by society. And the identity verification modes such as facial recognition, fingerprint recognition and the like can not express whether the principal reads the legal document and agrees with the content of the document, and only can show that the document and the principal appear at the same place.
Therefore, in the information age, when electronic documents are becoming mainstream legal document representations, and signing of remote agreements is technically possible, a technology capable of discriminating whether it is the signatory itself is particularly important. If the non-applicant himself signs in a contract in the insurance industry, the contract does not have legal effectiveness. Therefore, the identification technology for identifying the identity of the signer is a key technology in the electronic contract adopting a handwriting signature mode.
Because of the development of neural networks, image recognition technology has made many technological breakthroughs, and computers can accurately recognize images through a large number of sample studies. In the field of piracy identification, the similarity between the piracy resource and the original resource is judged by scanning images through video frames, so that the unauthorized piracy resource is determined to be distributed on the internet.
However, in the case of a large sample size, a considerable amount of calculation is required for accurate identification. If all the detection samples are subjected to signature comparison in a massive signature feature database, the efficiency is low, and a great deal of waste is generated on computing resources.
Technical Field
The invention relates to a system for electronic handwriting signature recognition based on a neural network, in particular to a fuzzy labeling recognition for electronic handwriting signature.
Disclosure of Invention
The invention aims to add fuzzy labels to signature data submitted by users before accurate user signature handwriting recognition is carried out, and when accurate recognition is required, the samples submitted by users in a massive database do not need to be accurately recognized and compared one by one.
The invention is realized in the following way:
the utility model provides an electronic handwriting signature fuzzy label identification system, includes signature terminal, server, storage device, characterized by: the signature terminal records characteristic data of a user signature, wherein the characteristic data comprise coordinates, pressure values, speeds, tangential angles, curvatures, overall acceleration and probability density;
submitting data to a sign terminal for a plurality of times by a user, binding the data with the user by the storage device, transmitting the data to the server for normalization processing, establishing a normalization model, and generating handwriting graphics of the corresponding user;
recording written three-dimensional information { x } in source file t ,y t ,p t X, where x t ,y t Respectively the horizontal and vertical coordinates, p, of the signature track t Is the pressure value at the time of signing.
And submitting the signature source file by the user, storing the signature source file in a signature database, and binding the signature source file with the user.
The signature source files of the same user are subjected to unified preprocessing, noise is reduced, signals are flattened, invalid data are removed, and data size normalization is carried out on x and y respectively:the signature is scaled into a unified rectangular box. Wherein x' t For signing the coordinate point of each track in the source file, x' max For maximum value of the whole track coordinates, x' min For the minimum value of the maximum value of the whole track coordinate, M is the size of a rectangular frame, and in practice, the maximum width is appointed as the maximum width when a user writes according to the condition M of signature.
And the server reduces the dimension of the handwriting graph and reduces the sampling data.
The server maps the normalized sampling characteristic data and the handwriting graph after dimension reduction which are matched with each other in a deep neural network, and calculates to obtain a training model;
the server generates to-be-blurred label data after training the training model, forcedly classifies the to-be-blurred label data by using a multi-classification method, performs fuzzy labeling classification, and transmits the result to a storage device for storage after generating blurred label classification data;
and the new user submits a new signature sample from the signature terminal to the server, the server extracts fuzzy label classification data with a threshold value larger than 0.8 after kernel function calculation from the storage device, and outputs a classification result.
The dimension reduction method is weighted average and/or simplex center removal.
The simplex decentration method is that n sampling points are determined in the characteristic data, n-sided polygon space of sample data is constructed by the n sampling points, data of non-n vertexes are distributed on n-sided polygons as far as possible, and the sampling points are converted into a matrix to obtain the simplex vertexes.
When the number of users outputting the classification result is larger than 1, multi-classification matching is performed in the characteristic data of the output users.
The kernel function adopts a Sigmoid kernel function.
The expression of the Sigmoid kernel function isWhere x is the input value and the entry is the characteristic data of the new signature sample.
By submitting signature data to fuzzy labeling processing, the calculated amount required by accurate identification is obviously reduced, so that the calculated amount in the signature identification process is optimized, the energy consumption is reduced, and the optimization is provided for subsequent accurate identification. The reduction of the calculation requirement of the recognition system leads the hardware requirement of the system to be reduced, and when only adopting the accurate recognition step, the double-path server unit can provide qualified response speed, but through the preliminary screening step, the common single-path server unit can achieve similar effect along with the great reduction of the calculation amount, thereby saving energy, simultaneously saving calculation resources, simultaneously saving purchasing cost and reducing operation and maintenance difficulty, and laying an important foundation for the large-scale application of handwriting recognition technology.
Drawings
FIG. 1 is a flow chart of electronic handwriting signature recognition according to an embodiment of the present invention;
fig. two is a feature extraction flow chart of the PCA according to an embodiment of the present invention.
Detailed Description
Deep Neural Networks (DNNs) learn more useful features by building machine learning models with many hidden layers and massive training data, thereby ultimately improving the accuracy of classification or prediction.
As shown in the figure I, the electronic handwriting signature recognition method based on the ANN neural network is to analyze and calculate the acquired data of a signature data acquisition terminal, and the acquired data of the acquisition terminal records the trace information of handwriting { x } in time sequence t ,y t ,p t A collection of electronic signature data at a read rate of 200 points/second or more may be used in conjunction with the method of the present scheme.
The invention is realized in the following way:
the utility model provides an electronic handwriting signature fuzzy label identification system, includes signature terminal, server, storage device, characterized by: the signature terminal records characteristic data of a user signature, wherein the characteristic data comprise coordinates, pressure values, speeds, tangential angles, curvatures, overall acceleration and probability density;
submitting data to a sign terminal for a plurality of times by a user, binding the data with the user by the storage device, transmitting the data to the server for normalization processing, establishing a normalization model, and generating handwriting graphics of the corresponding user;
recording written three-dimensional information { x } in source file t ,y t ,p t X, where x t ,y t Respectively the horizontal and vertical coordinates, p, of the signature track t Is the pressure value at the time of signing.
The extraction method of the average acceleration comprises the following steps:
average acceleration extraction method
The ratio of the change in the velocity vector to the elapsed time Δt is referred to as the average acceleration over Δt time.
The step of selecting the principal component according to a principal component analysis algorithm (PCA) comprises the steps of:
calculating the feature set, and calculating a covariance matrix:
and calculating eigenvalues and eigenvectors of the covariance matrix, and arranging the eigenvalues and eigenvectors from large to small.
The feature values are m, and are arranged from large to small and are lambda 1 ≥λ 2 ≥...≥λ m The eigenvector is p 1 ,p 2 ,...p m 。
The contribution rate and the cumulative contribution rate of each principal component are calculated.
And calculating the principal component vector.
Y=XP
Element y h The method comprises the following steps:
and selecting principal components according to principal component vectors, wherein the cumulative contribution rate is selected to be more than 85%, and the information contained in the principal components accounts for more than 85% of the original information.
The number of principal components of different signers with different accumulated contribution rates (Table I)
And after the principal component characteristic values are selected, recording the types of the selected characteristic values, and calculating the characteristic values which are the same in characteristic value selection during identification.
Deep learning firstly performs layer-by-layer pre-training on each layer by using unsupervised learning to learn characteristics; training one layer at a time, and taking the training result as input of a higher layer; and then the top layer is changed to supervised learning to perform fine adjustment from top to bottom to learn the model.
1) The self-descending non-supervision learning is used, wherein the first layer is firstly trained by non-calibration data, and the parameters of the first layer are firstly learned when the first layer is trained (the first layer can be regarded as a hidden layer for obtaining a three-layer neural network with minimum difference between output and input), and the obtained model can learn the structure of the data due to the limitation and sparsity constraint of the model, so that the characteristics with more representation capability than the input are obtained; after learning to obtain the n-1 layer, the n-1 layer output is used as the n layer input, and the n layer is trained, so that the parameters of each layer are obtained respectively.
2) Top-down supervised learning, which is to further process parameters of the whole multi-layer model based on the parameters of each layer obtained in the first step, wherein the first step is a supervised training process; the first step is similar to the random initialization initial value process of the neural network, and because the first step of DL is not random initialization, but is obtained by learning the structure of input data, the initial value is closer to global optimum, so that a better effect can be obtained.
Firstly, large-scale signature data is used as a model training set, and a general three-layer network is trained, wherein the first two layers are nonlinear layers, and the last layer is a linear layer.
And a supervised neighbor element analysis method is adopted, and the distances of the same individuals are reduced, so that the distances of different individuals are increased to optimize a network model, and the recognition accuracy is improved.
The neighbor analysis is a distance metric learning method, and uses a squared euclidean distance to define the distance between single data and the rest data in a new conversion space:
x i ,x j ,x k mapping data corresponding to the I, j and k bit handwriting data respectively, wherein A is a mapping space and is expressed as a certain layer of the neural network.
C i Signature data corresponding to individual i.
The network model is iteratively optimized by solving the gradient of the loss function f for each layer of network a using a continuous gradient descent algorithm.
And submitting the signature source file by the user, storing the signature source file in a signature database, and binding the signature source file with the user.
Of the same userThe signature source file is subjected to unified preprocessing, noise is reduced, signals are flattened, invalid data are removed, and data size normalization is carried out on x and y respectively:the signature is scaled into a unified rectangular box. Wherein x' t For signing the coordinate point of each track in the source file, x' max For maximum value of the whole track coordinates, x' min For the minimum value of the maximum value of the whole track coordinate, M is the size of a rectangular frame, and in practice, the maximum width is appointed as the maximum width when a user writes according to the condition M of signature.
And the server reduces the dimension of the handwriting graph and reduces the sampling data.
The server maps the normalized sampling characteristic data and the handwriting graph after dimension reduction which are matched with each other in a deep neural network, and calculates to obtain a training model;
the server generates to-be-blurred label data after training the training model, forcedly classifies the to-be-blurred label data by using a multi-classification method, performs fuzzy labeling classification, and transmits the result to a storage device for storage after generating blurred label classification data;
and the new user submits a new signature sample from the signature terminal to the server, the server extracts fuzzy label classification data with a threshold value larger than 0.8 after kernel function calculation from the storage device, and outputs a classification result.
The dimension reduction method is weighted average and/or simplex center removal.
The simplex decentration method is that n sampling points are determined in the characteristic data, n-sided polygon space of sample data is constructed by the n sampling points, data of non-n vertexes are distributed on n-sided polygons as far as possible, and the sampling points are converted into a matrix to obtain the simplex vertexes.
When the number of users outputting the classification result is larger than 1, multi-classification matching is performed in the characteristic data of the output users.
The kernel function adopts a Sigmoid kernel function.
The electronic handwriting signature fuzzification identification system of claim 5, wherein: the expression of the Sigmoid kernel function isWhere x is the input value and the entry is the characteristic data of the new signature sample. And establishing a mapping relation from the multidimensional sample information to the one-dimensional judging result. Final output<0,1>The value of the interval serves as possible identity information.
When the embodiment is implemented, a large amount of signature sample data of a user is collected, all signature samples are preprocessed, characteristic values are calculated, main component analysis is performed to screen out main cost characteristic values, invalid data and unstable characteristic values are filtered after the processing, the invalid data and unstable characteristic values are brought into a deep neural network to train, a data model is generated, and then signature data to be identified is brought into the neural network to be identified.
The foregoing description of the embodiments has been provided for the purpose of illustrating the general principles of the invention, and is not meant to limit the scope of the invention, but to limit the invention to the particular embodiments, and any modifications, equivalents, improvements, etc. that fall within the spirit and principles of the invention are intended to be included within the scope of the invention.
Claims (6)
1. The utility model provides an electronic handwriting signature fuzzy label identification system, includes signature terminal, server, storage device, characterized by: the signature terminal records characteristic data of a user signature, wherein the characteristic data comprise coordinates, pressure values, speeds, tangential angles, curvatures, overall acceleration and probability density;
submitting data to a sign terminal for a plurality of times by a user, binding the data with the user by the storage device, transmitting the data to the server for normalization processing, establishing a normalization model, generating handwriting graphs of the corresponding user, and then extracting characteristic values; calculating the contribution rate and the accumulated contribution rate of each principal component according to the covariance matrix of the feature set, calculating principal component vectors, selecting principal component elements according to the principal component vectors, and selecting the same feature values for calculation during identification according to the type of the selected feature values to obtain the feature values;
the server reduces the dimension of the handwriting graph and reduces sampling data;
the server maps the normalized sampling characteristic values and the handwriting graph after dimension reduction which are matched with each other in a deep neural network, and calculates to obtain a training model;
the server generates to-be-blurred label data after training the training model, forcedly classifies the to-be-blurred label data by using a multi-classification method, performs fuzzy labeling classification, and transmits the result to a storage device for storage after generating blurred label classification data; and the new user submits a new signature sample to the server from the signature terminal, the server extracts fuzzy label classification data with a threshold value larger than 0.8 after kernel function calculation from the storage equipment, and the classification result is output.
2. The electronic handwriting signature fuzzification identification system of claim 1, wherein: the dimension reduction method is weighted average and/or simplex center removal.
3. The electronic handwriting signature fuzzification identification system of claim 2, wherein: the simplex decentration method is that n sampling points are determined in the characteristic data, n-sided polygon space of sample data is constructed by the n sampling points, data of non-n vertexes are distributed on n-sided polygons as far as possible, and the sampling points are converted into a matrix to obtain the simplex vertexes.
4. The electronic handwriting signature fuzzification identification system of claim 2, wherein: when the number of users outputting the classification result is larger than 1, multi-classification matching is performed in the characteristic data of the output users.
5. The electronic handwritten signature fuzzification identification system of claim 3 or 4, wherein: the kernel function adopts a Sigmoid kernel function.
6. The electronic handwriting signature fuzzification identification system of claim 5, wherein: the expression of the Sigmoid kernel function is characterized in that x is an input value, and the input item is characteristic data of a new signature sample.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911422774.2A CN113128296B (en) | 2019-12-31 | 2019-12-31 | Electronic handwriting signature fuzzy label recognition system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911422774.2A CN113128296B (en) | 2019-12-31 | 2019-12-31 | Electronic handwriting signature fuzzy label recognition system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113128296A CN113128296A (en) | 2021-07-16 |
CN113128296B true CN113128296B (en) | 2023-05-09 |
Family
ID=76770699
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911422774.2A Active CN113128296B (en) | 2019-12-31 | 2019-12-31 | Electronic handwriting signature fuzzy label recognition system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113128296B (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106407874A (en) * | 2016-03-25 | 2017-02-15 | 东南大学 | Handwriting recognition method based on handwriting coordinate sequence |
CN106803082A (en) * | 2017-01-23 | 2017-06-06 | 重庆邮电大学 | A kind of online handwriting recognition methods based on conditional generation confrontation network |
CN108960275A (en) * | 2018-05-08 | 2018-12-07 | 中国矿业大学 | A kind of image-recognizing method and system based on depth Boltzmann machine |
CN109284693A (en) * | 2018-08-31 | 2019-01-29 | 深圳壹账通智能科技有限公司 | Based on the banking operation prediction technique of writing key point, device, electronic equipment |
CN110188201A (en) * | 2019-05-27 | 2019-08-30 | 上海上湖信息技术有限公司 | A kind of information matching method and equipment |
Family Cites Families (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0654755B1 (en) * | 1993-11-23 | 2000-08-02 | International Business Machines Corporation | A system and method for automatic handwriting recognition with a writer-independent chirographic label alphabet |
AUPR824301A0 (en) * | 2001-10-15 | 2001-11-08 | Silverbrook Research Pty. Ltd. | Methods and systems (npw001) |
US20060294390A1 (en) * | 2005-06-23 | 2006-12-28 | International Business Machines Corporation | Method and apparatus for sequential authentication using one or more error rates characterizing each security challenge |
CN101872259B (en) * | 2010-06-03 | 2012-01-25 | 中国人民解放军第二炮兵工程学院 | Natural interaction pen capable of automatically inputting written contents and handwriting detection method |
CN102254196B (en) * | 2011-06-22 | 2014-05-14 | 江苏奥博洋信息技术有限公司 | Method for identifying handwritten Chinese character by virtue of computer |
CN103279770B (en) * | 2013-06-06 | 2016-03-23 | 哈尔滨工业大学 | Based on the person's handwriting recognition methods of stroke fragment and contour feature |
CN104517094A (en) * | 2013-09-30 | 2015-04-15 | 阿里巴巴集团控股有限公司 | Identity authentication method and identity authentication device |
CN108460406B (en) * | 2018-02-02 | 2022-04-29 | 北京交通大学 | Scene image attribute identification method based on minimum simplex fusion feature learning |
CN109086654B (en) * | 2018-06-04 | 2023-04-28 | 平安科技(深圳)有限公司 | Handwriting model training method, text recognition method, device, equipment and medium |
CN109377205A (en) * | 2018-12-06 | 2019-02-22 | 深圳市淘米科技有限公司 | A kind of cafeteria's intelligence settlement system based on depth convolutional network |
-
2019
- 2019-12-31 CN CN201911422774.2A patent/CN113128296B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106407874A (en) * | 2016-03-25 | 2017-02-15 | 东南大学 | Handwriting recognition method based on handwriting coordinate sequence |
CN106803082A (en) * | 2017-01-23 | 2017-06-06 | 重庆邮电大学 | A kind of online handwriting recognition methods based on conditional generation confrontation network |
CN108960275A (en) * | 2018-05-08 | 2018-12-07 | 中国矿业大学 | A kind of image-recognizing method and system based on depth Boltzmann machine |
CN109284693A (en) * | 2018-08-31 | 2019-01-29 | 深圳壹账通智能科技有限公司 | Based on the banking operation prediction technique of writing key point, device, electronic equipment |
CN110188201A (en) * | 2019-05-27 | 2019-08-30 | 上海上湖信息技术有限公司 | A kind of information matching method and equipment |
Also Published As
Publication number | Publication date |
---|---|
CN113128296A (en) | 2021-07-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Xin et al. | Research on image classification model based on deep convolution neural network | |
Chacko et al. | Handwritten character recognition using wavelet energy and extreme learning machine | |
CN103605972B (en) | Non-restricted environment face verification method based on block depth neural network | |
JP2009514110A (en) | Human detection by pause | |
CN112668482B (en) | Face recognition training method, device, computer equipment and storage medium | |
Alhabib et al. | Exploiting wavelet transform, principal component analysis, support vector machine, and k-nearest neighbors for partial face recognition | |
Praseetha et al. | Secure fingerprint authentication using deep learning and minutiae verification | |
Gupta et al. | Single attribute and multi attribute facial gender and age estimation | |
Dhieb et al. | A novel biometric system for signature verification based on score level fusion approach | |
CN116110089A (en) | Facial expression recognition method based on depth self-adaptive metric learning | |
Wu et al. | Comparative analysis and application of LBP face image recognition algorithms | |
Qin et al. | Finger-vein quality assessment based on deep features from grayscale and binary images | |
Hefny et al. | Online signature verification using deep learning and feature representation using Legendre polynomial coefficients | |
Jadhav et al. | HDL-PI: hybrid DeepLearning technique for person identification using multimodal finger print, iris and face biometric features | |
Saleem et al. | A new segmentation framework for arabic handwritten text using machine learning techniques | |
Houtinezhad et al. | Writer-independent signature verification based on feature extraction fusion | |
Wang | Examination on face recognition method based on type 2 blurry | |
Shinde et al. | Feedforward back propagation neural network (FFBPNN) based approach for the identification of handwritten math equations | |
Yuan et al. | Research on intelligent algorithm of identity authentication based on facial features | |
Das et al. | Thai automatic signature verification system employing textural features | |
Houtinezhad et al. | Off-line signature verification system using features linear mapping in the candidate points | |
Khosravi et al. | Recognition of Persian/Arabic handwritten words using a combination of convolutional neural networks and autoencoder (AECNN) | |
CN113128296B (en) | Electronic handwriting signature fuzzy label recognition system | |
Bhavani et al. | A multi-dimensional review on handwritten signature verification: strengths and gaps | |
Lei et al. | Student action recognition based on multiple features |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CP03 | Change of name, title or address | ||
CP03 | Change of name, title or address |
Address after: 401121 no.17-1, building 13, No.106, west section of Jinkai Avenue, Yubei District, Chongqing Patentee after: Chongqing Sign Digital Technology Co.,Ltd. Country or region after: China Address before: 401121 no.17-1, building 13, No.106, west section of Jinkai Avenue, Yubei District, Chongqing Patentee before: CHONGQING AOS ONLINE INFORMATION TECHNOLOGY CO.,LTD. Country or region before: China |