CN118314588A - Method and system for recognizing handwriting fonts based on AI technology - Google Patents

Abstract

The invention relates to the technical field of image recognition, and discloses a method for recognizing handwriting fonts based on an AI technology, which comprises the following steps: collecting an original handwriting image, carrying out gray level processing, simplifying the handwriting image after gray level processing by adopting a binarization technology, and highlighting the contrast between characters and a background to obtain the simplified handwriting image; cutting and scaling the simplified handwriting image to obtain a handwriting image with uniform size, generalizing the unified handwriting image through rotation and translation operation to obtain an expanded handwriting image, and forming a standard handwriting image set together with the simplified handwriting image; extracting high-dimensional features of images in a standard handwriting image set to obtain a characterization vector, and integrating the characterization vectors extracted from the handwriting images of the same handwriting type to obtain a characterization vector sequence; and constructing a continuous dependency analysis model to obtain an identification font semantic vector sequence, and decoding and mapping the output semantic vector sequence to obtain identification font content.

Description

A method and system for recognizing handwritten fonts based on AI technology

Technical Field

The present invention relates to the technical field of image recognition, and in particular to a method and system for recognizing handwritten fonts based on AI technology.

Background technique

With the in-depth development of digital transformation, people have put forward higher requirements for the efficiency and accuracy of information processing. As a bridge connecting traditional writing and digital information processing, handwriting recognition technology has important research value and application prospects. Handwriting recognition can not only be applied to automatic form filling, digitization of historical documents, personal identity verification and other fields, but also promote the development of barrier-free communication technology and help the visually impaired to better obtain information. However, in cursive writing or cursive writing, the boundaries between characters are not obvious, and traditional segmentation-based recognition methods are difficult to accurately handle this problem. In view of this, this patent proposes a method for recognizing handwriting based on AI technology, which improves the accuracy and real-time performance of handwriting recognition through intelligent segmentation, which can greatly enrich the human-computer interaction mode and enhance the user experience.

Summary of the invention

In view of this, the present invention provides a method and system for recognizing handwritten fonts based on AI technology, the purpose of which is to: 1) propose a method for recognizing handwritten fonts based on AI technology, which effectively improves the accuracy and robustness of handwritten font recognition through feature extraction, model design and correction strategy; 2) perform high-dimensional feature extraction on images in a standard handwritten image set, and integrate the handwritten image features of the same handwriting type into a representation vector sequence, effectively capture the key features of the handwritten image, and lay the foundation for subsequent continuous dependency analysis; 3) adopt a long short-term memory network (LSTM) fused with an attention mechanism as a continuous dependency analysis model, which can more effectively process sequence data, especially in capturing long-distance dependencies. The introduction of the attention mechanism enables the model to dynamically focus on the key information in the sequence when processing the input of each time step, thereby improving the accuracy and efficiency of the model; 4) in the decoding mapping stage, by combining the language model for recognition correction, not only the results of font recognition are considered, but also the language model is used to score and correct the recognition results, which can effectively reduce recognition errors, especially in complex or unclear handwritten images, and can significantly improve the recognition accuracy.

To achieve the above object, the present invention provides a method for recognizing handwritten fonts based on AI technology, comprising the following steps:

S1: Collect the original handwritten image and grayscale it, use binarization technology to simplify the grayscale processed handwritten image, highlight the contrast between the text and the background to obtain a simplified handwritten image;

S2: The simplified handwritten image is cropped and scaled to obtain a handwritten image of uniform size, and the unified handwritten image is generalized by rotation and translation operations to obtain an expanded handwritten image, which together with the simplified handwritten image constitutes a standard handwritten image set;

S3: extracting high-dimensional features from images in the standard handwriting image set to obtain representation vectors, and integrating the representation vectors extracted from handwriting images of the same handwriting type to obtain a representation vector sequence;

S4: Construct a continuous dependency analysis model, wherein the model takes a representation vector sequence as input and a font semantic vector sequence as output, wherein the LSTM integrated with the attention mechanism is the main implementation method of the continuous dependency analysis model;

S5: Decoding and mapping the output semantic vector sequence to obtain the recognized font content, wherein recognition and correction combined with the language model is the main implementation method of the decoding and mapping.

As a further improvement method of the present invention:

Optionally, in step S1, collecting the original handwritten image and performing grayscale processing includes:

S21: Scan the handwritten text to obtain an original handwritten image, where 300 dpi is the resolution of the scanned image;

S22: grayscale each pixel in the original handwritten image, and the calculation formula is:

M＝0.299×R+0.587×G+0.114×B

Where: R, G, B represent the intensity values of the red, green and blue channels of each pixel in the original color image, respectively, and M represents the calculated grayscale value.

Optionally, in step S3, high-dimensional feature extraction is performed on the images in the standard handwritten image set to obtain a representation vector, including:

An improved convolutional neural network is used for high-dimensional feature extraction, including:

S31: Extract features from the input image through the convolution layer. In the convolution layer, the convolution kernel is used to slide on the input image, and the element-wise product with the local area covered by it is summed to generate a feature map. The calculation formula is:

in:

F _ij represents the value of the feature map at position (i, j);

I _(i+m)(j+n) represents the pixel value of the input image at position (i+m,j+n);

K _mn represents the weight of the convolution kernel at the (m,n) position;

b represents the bias term;

m and n are index variables that traverse all positions of the convolution kernel;

S32: The pooling layer reduces the spatial dimension of the feature map, reduces the number of parameters and the amount of calculation, while keeping the features unchanged. The calculation formula is:

in:

P _ij represents the value of the feature map at position (i, j) after pooling;

F _(i+m)(j+n) represents the value of the feature map output by the convolutional layer at position (i+m,j+n);

M×N is the size of the pooling window;

S33: Concatenate the extracted feature map matrices into a global feature vector and obtain a representation vector through residual connection.

In the step S33, the representation vector is obtained by residual connection, including:

By introducing short-circuit connections to learn the residual mapping relationship between the global feature vector and the representation vector, the calculation formula is:

F(x)＝H(x)+x

in:

X represents the input of the residual connection;

H(x) represents the result of the stack of convolutional layers processing the input x;

F(x) represents the output of the residual module.

Optionally, constructing a continuous dependency analysis model in step S4 includes:

S51: extracting context information from the representation vector sequence to obtain a context vector c _t-1 , and using the extracted context vector and the representation vector x _t input at the current moment as inputs to the LSTM to obtain the output of the LSTM at the current moment;

S52: Using the LSTM output h _t and context vector c _t at the current moment, generate the semantic feature vector y _t at the current moment through a fully connected layer.

In the step S52, the LSTM output h _t and the context vector c _t at the current moment are used to generate the semantic feature vector y _t at the current moment through a fully connected layer, including:

S61: Calculate the attention weight according to the representation vector sequence. The calculation formula is:

e _t,s = a(h _t-1 ,h _s )

Where: e _t,s is the attention score; α _t,s represents the normalized attention weight, t,s is the time index; a() is the self-attention calculation function;

S62: According to the calculated attention weight, the calculation formula is:

Where: c _t represents the context vector after weighted averaging; h _s represents the LSTM output at time s.

Optionally, in step S5, decoding and mapping the output semantic vector sequence to obtain the identified font content includes:

S71: For each time step, the semantic vector is mapped through full connection to obtain the output of this step. The calculation formula is:

z _t = W·y _t + b

Where: z _t represents the output of the fully connected layer; W represents the weight matrix of the fully connected layer; b represents the bias term;

S72: The output of the fully connected layer is converted into a probability distribution through Softmax. For the fully connected layer output z _t at each time step t, the probability distribution calculation formula is:

in:

P(y _t = k|z _t ) represents the probability of predicting category k given input z _t at time step t;

z _t,k represents the component of vector z _t corresponding to category k;

K is the total number of categories, i.e. the number of different fonts that need to be recognized.

The recognition and correction combined with the language model in step S5 includes:

S81: Construct a language scoring model, wherein N-gram is the main implementation method of the scoring model, and the calculation formula is:

Where: (S = s ₁ , s ₂ , ..., s _N ) represents the font sequence at the previous N moments; S (s _i | s ₁ , s ₂ , ..., s _i-1 ) represents the conditional probability of the i-th font appearing given the previous i-1 fonts;

S82: The probability distribution of font recognition is integrated with the score of the language model, a comprehensive score of each possible sequence is calculated, and the sequence with the highest score is selected as the final recognition result.

In step S82, the probability distribution of font recognition is integrated with the score of the language model to calculate the comprehensive score of each possible sequence, including:

Calculate the scores of all possible sequences using the formula:

Score(S)＝αP _font (S)+βP _lang (S)

in:

P _font (S) represents the probability of obtaining sequence S by decoding mapping calculation;

P _lang (S) represents the probability calculated by the language scoring model for the sequence S;

α and β represent weight parameters used to balance the probability of font recognition and the probability of language model.

In order to solve the above problems, the present invention provides a system for recognizing handwritten fonts based on AI technology, the system comprising:

The data acquisition module is used to acquire the original handwritten image and perform grayscale processing to obtain the simplified handwritten image to form a standard handwritten image set, and to perform high-dimensional feature extraction on the images in the standard handwritten image set to obtain a representation vector, and to integrate the representation vectors extracted from the handwritten images of the same handwriting type to obtain a representation vector sequence;

A continuous dependency analysis module is used to construct a continuous dependency analysis model, and obtain a font semantic vector sequence based on a representation vector sequence as input;

The font recognition module is used to decode and map the output semantic vector sequence to obtain the recognized font content.

In order to solve the above problem, the present invention further provides an electronic device, the electronic device comprising:

A memory storing at least one instruction;

Communication interface, enabling electronic equipment to communicate; and

The processor executes the instructions stored in the memory to implement the above-mentioned method for recognizing handwritten fonts based on AI technology.

In order to solve the above problem, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the above-mentioned method for recognizing handwritten fonts based on AI technology.

Compared with the prior art, the present invention proposes a method for recognizing handwritten fonts based on AI technology, which has the following advantages:

First, this scheme proposes a method for handwriting recognition based on AI technology, which effectively improves the accuracy and robustness of handwriting recognition through feature extraction, model design and correction strategy;

At the same time, this scheme extracts high-dimensional features from images in the standard handwritten image set, and integrates the handwritten image features of the same handwriting type into a representation vector sequence, which effectively captures the key features of the handwritten image and lays the foundation for subsequent continuous dependency analysis. The long short-term memory network (LSTM) with integrated attention mechanism is used as the continuous dependency analysis model, which can process sequence data more effectively, especially in capturing long-distance dependencies. The introduction of the attention mechanism enables the model to dynamically focus on the key information in the sequence when processing the input of each time step, thereby improving the accuracy and efficiency of the model.

In addition, during the decoding and mapping stage, this scheme combines the language model for recognition correction. It not only takes into account the results of font recognition, but also uses the language model to score and correct the recognition results. It can effectively reduce recognition errors, especially in complex or unclear handwritten images, and can significantly improve the recognition accuracy.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a flow chart of a method for recognizing handwritten fonts based on AI technology provided by an embodiment of the present invention;

FIG2 is a functional module diagram of a system for recognizing handwritten fonts based on AI technology provided by an embodiment of the present invention;

FIG3 is a schematic diagram of the structure of an electronic device for implementing a method for recognizing handwritten fonts based on AI technology provided by an embodiment of the present invention.

The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings.

Detailed ways

It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

The embodiment of the present application provides a method for recognizing handwritten fonts based on AI technology. The execution subject of the method for recognizing handwritten fonts based on AI technology includes but is not limited to at least one of the electronic devices such as a server, a terminal, etc. that can be configured to execute the method provided by the embodiment of the present application. In other words, the method for recognizing handwritten fonts based on AI technology can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

Embodiment 1:

S1: The original handwritten image is collected and gray-scaled, and the gray-scaled handwritten image is simplified by using a binarization technique to highlight the contrast between the text and the background to obtain a simplified handwritten image.

In the step S1, the original handwritten image is collected and grayscaled, including:

S21: Scan the handwritten text to obtain an original handwritten image, where 300 dpi is the resolution of the scanned image;

S22: grayscale each pixel in the original handwritten image, and the calculation formula is:

M＝0.299×R+0.587×G+0.114×B

Where: R, G, B represent the intensity values of the red, green and blue channels of each pixel in the original color image, respectively, and M represents the calculated grayscale value.

S2: The simplified handwritten image is cropped and scaled to obtain a handwritten image of uniform size, and the unified handwritten image is generalized through rotation and translation operations to obtain an expanded handwritten image, which together with the simplified handwritten image constitutes a standard handwritten image set.

S3: extract high-dimensional features from images in the standard handwriting image set to obtain representation vectors, and integrate the representation vectors extracted from handwriting images of the same handwriting type to obtain a representation vector sequence.

In the step S3, high-dimensional feature extraction is performed on the images in the standard handwritten image set to obtain a representation vector, including:

An improved convolutional neural network is used for high-dimensional feature extraction, including:

S31: Extract features from the input image through the convolution layer. In the convolution layer, the convolution kernel is used to slide on the input image, and the element-wise product is performed with the local area covered by it, and then the sum is taken to generate a feature map. The calculation formula is:

in:

F _ij represents the value of the feature map at position (i, j);

I _(i+m)(j+n) represents the pixel value of the input image at position (i+m,j+n);

K _mn represents the weight of the convolution kernel at the (m,n) position;

b represents the bias term;

m and n are index variables that traverse all positions of the convolution kernel;

S32: The pooling layer reduces the spatial dimension of the feature map, reduces the number of parameters and the amount of calculation, while keeping the features unchanged. The calculation formula is:

in:

P _ij represents the value of the feature map at position (i, j) after pooling;

F _(i+m)(j+n) represents the value of the feature map output by the convolutional layer at position (i+m,j+n);

M×N is the size of the pooling window;

S33: Concatenate the extracted feature map matrices into a global feature vector and obtain a representation vector through residual connection.

In the step S33, the representation vector is obtained by residual connection, including:

By introducing short-circuit connections to learn the residual mapping relationship between the global feature vector and the representation vector, the calculation formula is:

F(x)＝H(x)+x

in:

X represents the input of the residual connection;

H(x) represents the result of the stack of convolutional layers processing the input x;

F(x) represents the output of the residual module.

S4: Construct a continuous dependency analysis model, which takes a representation vector sequence as input and an identification font semantic vector sequence as output, wherein the LSTM integrated with the attention mechanism is the main implementation method of the continuous dependency analysis model.

The continuous dependency analysis model is constructed in step S4, including:

S51: extracting context information from the representation vector sequence to obtain a context vector c _t-1 , and using the extracted context vector and the representation vector x _t input at the current moment as inputs to the LSTM to obtain the output of the LSTM at the current moment;

S52: Using the LSTM output h _t and context vector c _t at the current moment, generate the semantic feature vector y _t at the current moment through a fully connected layer.

In the step S52, the current LSTM output h _t and the context vector c _t are used to generate the current semantic feature vector y _t through a fully connected layer, including:

S61: Calculate the attention weight according to the representation vector sequence. The calculation formula is:

e _t,s = a(h _t-1 ,h _s )

Where: e _t,s is the attention score; α _t,s represents the normalized attention weight, t,s is the time index; a() is the self-attention calculation function;

S62: According to the calculated attention weight, the calculation formula is:

Where: c _t represents the context vector after weighted averaging; h _s represents the LSTM output at time s.

S5: Decoding and mapping the output semantic vector sequence to obtain the recognized font content, wherein recognition and correction combined with the language model is the main implementation method of the decoding and mapping.

In the step S5, decoding and mapping the output semantic vector sequence to obtain the recognized font content includes:

S71: For each time step, the semantic vector is mapped through full connection to obtain the output of this step. The calculation formula is:

z _t = W·y _t + b

Where: z _t represents the output of the fully connected layer; W represents the weight matrix of the fully connected layer; b represents the bias term;

S72: The output of the fully connected layer is converted into a probability distribution through Softmax. For the fully connected layer output z _t at each time step t, the probability distribution calculation formula is:

in:

P(y _t = k|z _t ) represents the probability of predicting category k given input z _t at time step t;

z _t,k represents the component of vector z _t corresponding to category k;

K is the total number of categories, i.e. the number of different fonts that need to be recognized.

The recognition and correction combined with the language model in step S5 includes:

S81: Construct a language scoring model, wherein N-gram is the main implementation method of the scoring model, and the calculation formula is:

Where: (S = s ₁ , s ₂ , ..., s _N ) represents the font sequence at the previous N moments; P (s _i | s ₁ , s ₂ , ..., s _i-1 ) represents the conditional probability of the i-th font appearing given the previous i-1 fonts;

S82: The probability distribution of font recognition is integrated with the score of the language model, a comprehensive score of each possible sequence is calculated, and the sequence with the highest score is selected as the final recognition result.

In step S82, the probability distribution of font recognition is integrated with the score of the language model to calculate the comprehensive score of each possible sequence, including:

Calculate the scores of all possible sequences using the formula:

Score(S)＝αP _font (S)+βP _lang (S)

in:

P _font (S) represents the probability of obtaining sequence S by decoding mapping calculation;

P _lang (S) represents the probability calculated by the language scoring model for the sequence S;

α and β represent weight parameters used to balance the probability of font recognition and the probability of language model.

Embodiment 2:

As shown in FIG. 2 , it is a functional module diagram of a system for recognizing handwritten fonts based on AI technology provided in an embodiment of the present invention, which can implement the method for recognizing handwritten fonts based on AI technology in Embodiment 1.

The system 100 for recognizing handwritten fonts based on AI technology of the present invention can be installed in an electronic device. According to the functions implemented, the system for recognizing handwritten fonts based on AI technology can include a data acquisition module 101, a continuous dependency analysis module 102 and a font recognition module 103. The module of the present invention can also be called a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.

The data acquisition module 101 is used to acquire the original handwritten image and perform grayscale processing to obtain simplified handwritten images to form a standard handwritten image set, and to perform high-dimensional feature extraction on the images in the standard handwritten image set to obtain representation vectors, and to integrate the representation vectors extracted from the handwritten images of the same handwriting type to obtain a representation vector sequence;

The continuous dependency analysis module 102 is used to construct a continuous dependency analysis model, and obtain a font semantic vector sequence according to the representation vector sequence as input;

The font recognition module 103 is used to decode and map the output semantic vector sequence to obtain the recognized font content.

In detail, the modules in the system 100 for recognizing handwritten fonts based on AI technology in the embodiment of the present invention adopt the same technical means as the method for recognizing handwritten fonts based on AI technology described in the above-mentioned Figure 1 when used, and can produce the same technical effects, which will not be repeated here.

Embodiment 3:

As shown in FIG. 3 , it is a schematic diagram of the structure of an electronic device for implementing a method for recognizing handwritten fonts based on AI technology provided by an embodiment of the present invention.

The electronic device 1 may include a processor 10 , a memory 11 , a communication interface 13 and a bus, and may also include a computer program stored in the memory 11 and executable on the processor 10 , such as a program 12 .

The memory 11 includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a mobile hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a magnetic memory, a disk, an optical disk, etc. The memory 11 may be an internal storage unit of the electronic device 1 in some embodiments, such as a mobile hard disk of the electronic device 1. The memory 11 may also be an external storage device of the electronic device 1 in other embodiments, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 1. Further, the memory 11 may also include both an internal storage unit of the electronic device 1 and an external storage device. The memory 11 may not only be used to store application software and various types of data installed in the electronic device 1, such as the code of the program 12, etc., but also be used to temporarily store data that has been output or is to be output.

The processor 10 may be composed of an integrated circuit in some embodiments, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips, etc. The processor 10 is the control core (Control Unit) of the electronic device, and uses various interfaces and lines to connect various components of the entire electronic device, and executes or executes programs or modules stored in the memory 11 (such as a program 12 for realizing handwriting font recognition based on AI technology), and calls data stored in the memory 11 to execute various functions of the electronic device 1 and process data.

The communication interface 13 may include a wired interface and/or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device 1 and other electronic devices, and to achieve connection and communication between internal components of the electronic device.

The bus may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to realize connection and communication between the memory 11 and at least one processor 10, etc.

FIG3 merely shows an electronic device with components. Those skilled in the art will appreciate that the structure shown in FIG3 does not limit the electronic device 1 and may include fewer or more components than shown in the figure, or a combination of certain components, or a different arrangement of components.

For example, although not shown, the electronic device 1 may also include a power source (such as a battery) for supplying power to each component. Preferably, the power source may be logically connected to the at least one processor 10 through a power management device, so that the power management device can realize functions such as charging management, discharging management, and power consumption management. The power source may also include one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The electronic device 1 may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

Optionally, the electronic device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), or a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode) touch device. The display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device 1 and to display a visual user interface.

It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. And the terms "including", "comprising" or any other variants thereof in this article are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "including a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims (10)

1. A method and a system for recognizing handwriting fonts based on AI technology are characterized in that the method comprises the following steps:

s1: collecting an original handwriting image, carrying out gray level processing, simplifying the handwriting image after gray level processing by adopting a binarization technology, and highlighting the contrast between characters and a background to obtain the simplified handwriting image;

S2: cutting and scaling the simplified handwriting image to obtain a handwriting image with uniform size, generalizing the unified handwriting image through rotation and translation operation to obtain an expanded handwriting image, and forming a standard handwriting image set together with the simplified handwriting image;

S3: extracting high-dimensional features of images in a standard handwriting image set to obtain a characterization vector, and integrating the characterization vectors extracted from the handwriting images of the same handwriting type to obtain a characterization vector sequence;

S4: constructing a continuous dependency analysis model, wherein the model takes a representation vector sequence as input, and recognizes a font semantic vector sequence as output, and LSTM fused with an attention mechanism is a main implementation method of the continuous dependency analysis model;

s5: and carrying out decoding mapping on the output semantic vector sequence to obtain the identification font content, wherein the identification deviation correction combined with the language model is a main implementation method of the decoding mapping.

2. The method and system for recognizing handwriting fonts based on AI technology as claimed in claim 1, wherein the step S1 of collecting an original handwriting image and performing graying processing includes:

s21: scanning the handwritten text to obtain an original handwritten image, wherein 300dpi is the resolution of the scanned image;

S22: gray processing is carried out on each pixel point in the original handwriting image, and a calculation formula is as follows:

M＝0.299×R+0.587×G+0.114×B

Wherein: r, G, B represent the intensity values of the red, green and blue channels, respectively, of each pixel in the original color image, M representing the calculated gray values.

3. The method and system for recognizing handwriting fonts based on AI technology as claimed in claim 2, wherein the step S3 of extracting high-dimensional features of images in the standard handwriting image set to obtain the characterization vector includes:

High-dimensional feature extraction using an improved convolutional neural network, comprising:

S31: features are extracted from an input image through a convolution layer, in the convolution layer, a convolution kernel is utilized to slide on the input image, element products are carried out on the convolution kernel and a local area covered by the convolution kernel, and then the convolution kernel and the local area are summed to generate a feature map, wherein a calculation formula is as follows:

Wherein:

f _ij represents the value of the feature map at position (i, j);

I _(i+m)(j+n) denotes the pixel value of the input image at position (i+m, j+n);

K _mn represents the weight of the convolution kernel at the (m, n) position;

b represents a bias term;

m and n traverse index variables for all positions of the convolution kernel;

S32: the space dimension of the feature map is reduced through the pooling layer, the parameter quantity and the calculated amount are reduced, the feature is kept unchanged, and the calculation formula is as follows:

Wherein:

P _ij represents the value of the pooled feature map at position (i, j);

F _(i+m)(j+n) represents the value of the feature map output by the convolution layer at position (i+m, j+n);

M×n is the size of the pooling window;

S33: and splicing the extracted feature map matrixes into total feature vectors, and obtaining characterization vectors through residual connection.

4. The method and system for recognizing handwriting fonts based on AI technology as claimed in claim 3, wherein the step S33 of obtaining the characterization vector through residual connection comprises:

By introducing a residual mapping relation between a short circuit connection learning global feature vector and a characterization vector, a calculation formula is as follows:

F(x)＝H(x)+x

Wherein:

x represents the input of the residual connection;

H (x) represents the result of the input x processing by the stack of convolutional layers;

F (x) represents the output of the residual block.

5. The method and system for recognizing handwriting fonts based on AI technology as claimed in claim 1, wherein constructing a continuous dependency analysis model in step S4 includes:

S51: extracting context information from the characterization vector sequence to obtain a context vector c _t-1, and taking the extracted context vector and a characterization vector x _t input at the current moment as inputs of LSTM to obtain the output of the LSTM at the current moment;

S52: the semantic feature vector y _t at the current time is generated by the full connection layer using the LSTM output h _t at the current time and the context vector c _t.

6. The method and system for recognizing handwriting according to claim 5, wherein generating the semantic feature vector y _t at the current time through the full connection layer by using the LSTM output h _t at the current time and the context vector c _t in the step S52 comprises:

s61: and calculating attention weight according to the characterization vector sequence, wherein the calculation formula is as follows:

e_t,s＝a(h_t-1,h_s)

Wherein: e _t,s is the attention score;

Alpha _t,s represents the normalized attention weight;

t, s is a time sequence index; a () is a self-attention computation function;

S62: according to the calculated attention weight, the calculation formula is as follows:

wherein: c _t denotes the context vector after weighted averaging;

h _s represents the LSTM output at s time.

7. The method and system for recognizing handwriting fonts based on AI technology as claimed in claim 1, wherein the step S5 of decoding and mapping the output semantic vector sequence to obtain the recognized font content includes:

s71: the semantic vector of each time step is mapped through full connection to obtain the output of the step, and the calculation formula is as follows:

z_t＝W·y_t+b

Wherein: z _t represents the full link layer output;

W represents a full connection layer weight matrix;

b represents a bias term;

s72: the output of the full-connection layer is converted into probability distribution through Softmax, and for the full-connection layer output z _t of each time step t, the probability distribution calculation formula is as follows:

Wherein:

P (y _t＝k|z_t) represents the probability of being predicted as class k given input z _t at time step t;

z _t,k represents the component of vector z _t that corresponds to category k;

k is the total number of categories, i.e. the number of different fonts that need to be identified.

8. The method and system for recognizing handwriting fonts based on AI technology as claimed in claim 1, wherein the step S5 combines recognition correction of language model, comprising:

s81: constructing a language scoring model, wherein N-gram is a main implementation method of the scoring model, and the calculation formula is as follows:

Wherein:

(s=s ₁,s₂,...,s_N) represents the font sequence at the first N times;

P (s _i|s₁,s₂,...,s_i-1) represents the conditional probability that the i-th font appears given the first i-1 fonts;

S82: and fusing the probability distribution of the font recognition with the score of the language model, calculating the comprehensive score of each possible sequence, and selecting the sequence with the highest score as a final recognition result.

9. The method and system for recognizing handwritten fonts based on AI technology as recited in claim 8, wherein the step S82 of fusing a probability distribution of font recognition with scores of a language model, calculating a composite score for each possible sequence, includes:

The scores of all possible sequences are calculated as:

Score(S)＝αP_font(S)+βP_lang(S)

Wherein:

P _font (S) represents the probability of the decoding map calculation to obtain the sequence S;

P _lang (S) represents the probability that the language scoring model calculates for sequence S;

alpha and beta represent weight parameters for balancing the font recognition probability and the language model probability.

10. A system for recognizing handwriting fonts based on AI technology, said system comprising:

The data acquisition module is used for acquiring an original handwriting image and carrying out graying treatment to obtain a simplified handwriting image to form a standard handwriting image set, carrying out high-dimensional feature extraction on images in the standard handwriting image set to obtain a characterization vector, and integrating the characterization vectors obtained by extracting handwriting images of the same handwriting type to obtain a characterization vector sequence;

the continuous dependency analysis module is used for constructing a continuous dependency analysis model and obtaining a font semantic vector sequence according to the characterization vector sequence as input;

A font recognition module, configured to decode and map the output semantic vector sequence to obtain a recognized font content, so as to implement the method for recognizing a handwritten font based on AI technology according to any one of claims 1-9.