A writing recognition device

Abstract

The present invention relates to a writing recognition device, belonging to the technical field of writing devices. The present invention includes a stylus pen and a recognition display terminal; the stylus pen includes a battery, a control circuit, a posture acquisition module, a battery management module, a writing device that realizes the basic functions of the pen, and a pen drop detection module; the pen drop detection module and the posture acquisition module are connected to the control circuit, and the battery management module is connected to the battery; the control circuit includes a radio frequency circuit and a main control chip; the control circuit is connected to an information receiving module in a host computer through the radio frequency circuit thereon; the recognition display terminal adopts a host computer, and the host computer includes an information receiving module, a preprocessing module, a neural network module, and a display output module; the information receiving module, the preprocessing module, the neural network module, and the display output module are connected in sequence. The present invention can recognize the strokes or characters written by the user, and finally achieve the effect of being able to recognize the Chinese characters and letters written by the user.

Description

Technical Field

[0001] The invention relates to a writing recognition device, belonging to the technical field of writing devices. Background Art

[0002] Handwriting input is a popular method of human-computer interaction. It significantly improves the user experience and convenience. It is currently used in various fields, such as human-computer interaction, password recognition, game entertainment, etc.

[0003] However, the existing handwriting recognition technology records the coordinates of pixel points and the vector value of the handwriting direction according to the handwriting direction (time order) while the user is writing, and divides them into different sequence sets according to the strokes (also called strokes), and processes these records to standardize each pixel point in each sequence set, and finally compares the handwritten font sequence set with the standard font sequence set in the standard font library, takes the Chinese character with the highest similarity, and returns it to the user as the recognition result; there are also methods that directly use image recognition to recognize the note image written by the user, and map the classification results of the image with the recognized Chinese characters, so as to achieve the purpose of recognition.

[0004] It can be seen from the above prior art that the current handwriting recognition technology uses the handwriting sequence set similarity of the entire handwritten Chinese character as the standard for recognizing Chinese characters, without considering the inherent characteristics of each stroke type and the differences caused by the different writing habits of each person. Moreover, the recognition of this method requires relying on a fixed device, such as a touch screen, while the present invention can be used as an input device for spatial handwriting, which is separated from traditional writing carriers such as paper and display screens. It only needs a pen, and the display device can be a mobile phone, computer, etc. that can realize deep learning operations and can display, and the present invention conforms to the current recognition method of people's handwriting habits. At the same time, the present invention also has the use of an ordinary pen, and can be used as an ordinary handwriting device when no recognition is performed.

[0005] The present invention mainly uses MEMS sensors to work. MEMS sensors are not affected by light when in use and have the characteristics of convenience, miniaturization, low cost, etc. These reasons make us focus on the MEMS-based handwriting recognition system. Summary of the invention

[0006] The technical problem to be solved by the present invention is: the present invention provides a writing recognition device to recognize the strokes or characters written by the user, and finally achieve the effect of being able to recognize the Chinese characters and letters written by the user.

[0007] The technical solution of the present invention is: a writing recognition device, including a stylus pen and a recognition display terminal;

[0008] The stylus pen includes a battery, a control circuit, a posture acquisition module, a battery management module, a writing device that realizes the basic functions of the pen, and a pen drop detection module;

[0009] The battery is used for power supply;

[0010] The control circuit, the posture acquisition module, the battery management module, and the pen drop detection module are integrated on a circuit board 14; the pen drop detection module and the posture acquisition module are connected to the control circuit, and the battery management module is connected to the battery; the circuit board 14 is arranged on a writing device that realizes the basic functions of the pen;

[0011] The control circuit includes a radio frequency circuit and a main control chip; the control circuit is connected to the information receiving module in the host computer through the radio frequency circuit thereon;

[0012] The identification and display terminal adopts a host computer, which includes an information receiving module, a preprocessing module, a neural network module, and a display output module; the information receiving module, the preprocessing module, the neural network module, and the display output module are connected in sequence.

[0013] As a further solution of the present invention, the main control chip adopts MCU, and the MCU adopts esp32 s3, SOC or single chip microcomputer.

[0014] As a further solution of the present invention, the preprocessing module is used to perform data enhancement and denoising on the posture data collected by the posture collection module;

[0015] The neural network module is used to recognize the posture data processed by the preprocessing module during the writing process, learn and recognize the strokes or characters written by the user through machine learning, and display them through the display output module;

[0016] As a further solution of the present invention, the writing device that realizes the basic functions of the pen includes a pen cap 11, a middle pen tube 12, a rear pen tube 15, a rear pen cover 16, and a pen core; the middle pen tube 12 is threadedly connected to the rear pen tube 15, the battery 13 and the circuit board 14 are arranged in the rear pen tube 15, and the rear pen cover 16 covers the rear pen tube 15.

[0017] As a further solution of the present invention, a metal ring is provided on the pen cap 11, and two contacts are provided on the pen tube 12, both of which are connected to the enable end of the main control chip. When the pen cap 11 is put on the pen tube 12, the metal ring connects the two contacts, thereby disabling the enable end of the main control chip and making the control circuit inoperative; when the pen cap 11 is removed, the two contacts are disconnected and the control circuit operates.

[0018] As a further solution of the present invention, a copper foil is attached to the end of the pen rear tube 15, which is connected to the capacitance measurement pin of the single-chip microcomputer in the control circuit. The working state of the control circuit is switched by detecting the capacitance of the copper foil to the ground. When the finger touches the end of the pen rear tube 15, the capacitance of the copper foil to the ground is changed, thereby transmitting a switch pressing signal to the single-chip microcomputer.

[0019] Through this pressing signal, the control circuit can control the working mode of the pen, for example: long press to enter the recognition mode, long press again to enter the normal working mode and use it as a normal pen.

[0020] As a further solution of the present invention, a planed surface is provided on the pen tube 12 for better positioning the direction of the posture acquisition module when holding the pen. The position of the posture acquisition module in the pen is fixed. Because the thumb is pressed on the planed surface, the position of the thumb and the posture acquisition module is relatively fixed, and the position of the entire hand and the posture acquisition module is relatively fixed, thereby reducing the complexity of data processing and helping to improve the comfort of holding the pen.

[0021] As a further solution of the present invention, the pen drop detection module includes a capacitance detection probe, which is connected to the metal pen tip. When the pen tip is suspended in the air or when the pen is dropped on paper, the capacitance value to the ground measured by the probe will be different. Through this characteristic, it is detected whether the pen is dropped. The wire connecting the capacitance detection contact on the circuit board and the metal pen tip needs to be shielded with a grounded metal wire to ensure that it is not affected by other factors, such as the proximity of a finger to affect the measured capacitance;

[0022] As a further solution of the present invention, the main control chip on the circuit board 14 uses a soc with a wifi bluetooth function, the main control soc collects the attitude sensor output attitude signal through the iic communication, and then sends it through the wireless network, the main control soc uses the embedded operating system freertos to perform task scheduling, and has two tasks, a reading task and a sending task, and the execution cycle of the task is set by the operating system, the attitude acquisition module used uses an attitude sensor to measure three-axis acceleration, three-axis angular velocity and three-axis magnetic field, the main control chip calculates three Euler angles roll, pitch, yaw, thereby calculating the information of relative displacement and attitude transformation, and by integrating the acceleration data, the attitude sensor collects the speed information and position information of the sensor in space, and the information sequence collected by these attitude sensors constitutes an information matrix uploaded to the host computer, and after the information is transmitted, the host computer pre-processes the information and sends it to the neural network for recognition. The Chinese characters recognized by the host computer are displayed according to the needs of the user, and can also be used as input information as the input of the association model to associate Chinese characters.

[0023] The beneficial effect of the present invention is that the present invention uses a neural network to identify the posture time series collected by the posture collection module during the writing process, and uses machine learning to identify the strokes or characters written by the user, and finally achieves the effect of being able to recognize the Chinese characters and letters written by the user. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a partial structural decomposition schematic diagram of the present invention;

[0025] Figure 2 It is a partial structural block diagram of the present invention;

[0026] Figure 3 It is a schematic diagram of the circuit structure of the power module of the present invention;

[0027] Figure 4 It is a schematic diagram of the circuit structure of the posture acquisition module of the present invention;

[0028] Figure 5 It is a schematic diagram of the structure of part of the control circuit of the present invention;

[0029] Figure 6 It is a schematic diagram of the peripheral circuit structure of the control belt path of the present invention.

[0030] Figure 1 The numbers in the figure are: 11-pen cap, 12-pen middle tube, 13-battery, 14-circuit board, 15-pen rear tube, 16-pen rear cover. DETAILED DESCRIPTION

[0031] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0032] Example 1: Figure 1-Figure 6 As shown, a writing recognition device includes a stylus pen and a recognition display terminal; the stylus pen includes a battery, a control circuit, a posture acquisition module, a battery management module, a writing device that realizes the basic functions of the pen, and a pen drop detection module; the battery is used for power supply;

[0033] The control circuit, the posture acquisition module, the battery management module, and the pen drop detection module are integrated on a circuit board 14; the pen drop detection module and the posture acquisition module are connected to the control circuit, and the battery management module is connected to the battery; the circuit board 14 is arranged on a writing device that realizes the basic functions of the pen;

[0034] The control circuit includes a radio frequency circuit and a main control chip; the control circuit is connected to the information receiving module in the host computer through the radio frequency circuit thereon;

[0035] The identification and display terminal adopts a host computer, which includes an information receiving module, a preprocessing module, a neural network module, and a display output module; the information receiving module, the preprocessing module, the neural network module, and the display output module are connected in sequence.

[0036] As a further solution of the present invention, the main control chip adopts MCU, and the MCU adopts esp32 s3 or a single-chip microcomputer.

[0037] As a further solution of the present invention, the preprocessing module is used to perform data enhancement and denoising on the posture data collected by the posture collection module;

[0038] The neural network module is used to recognize the posture data processed by the preprocessing module during the writing process, learn and recognize the strokes or characters written by the user through machine learning, and display them through the display output module;

[0039] As a further solution of the present invention, the writing device that realizes the basic functions of the pen includes a pen cap 11, a middle pen tube 12, a rear pen tube 15, a rear pen cover 16, and a pen core; the middle pen tube 12 is threadedly connected to the rear pen tube 15, the battery 13 and the circuit board 14 are arranged in the rear pen tube 15, and the rear pen cover 16 covers the rear pen tube 15.

[0040] As a further solution of the present invention, a metal ring is provided on the pen cap 11, and two contacts are provided on the pen tube 12, both of which are connected to the enable end of the main control chip. When the pen cap 11 is put on the pen tube 12, the metal ring connects the two contacts, thereby disabling the enable end of the main control chip and making the control circuit inoperative; when the pen cap 11 is removed, the two contacts are disconnected and the control circuit operates.

[0041] As a further solution of the present invention, a copper foil is attached to the end of the pen rear tube 15, which is connected to the capacitance measurement pin of the single-chip microcomputer in the control circuit. The working state of the control circuit is switched by detecting the capacitance of the copper foil to the ground. When the finger touches the end of the pen rear tube 15, the capacitance of the copper foil to the ground is changed, thereby transmitting a switch pressing signal to the single-chip microcomputer.

[0042] Through this pressing signal, the control circuit can control the working mode of the pen, for example: long press to enter the recognition mode, long press again to enter the normal working mode and use it as a normal pen.

[0043] As a further solution of the present invention, a planed surface is provided on the pen tube 12 for better positioning the direction of the posture acquisition module when holding the pen. The position of the posture acquisition module in the pen is fixed. Because the thumb is pressed on the planed surface, the position of the thumb and the posture acquisition module is relatively fixed, and the position of the entire hand and the posture acquisition module is relatively fixed, thereby reducing the complexity of data processing and helping to improve the comfort of holding the pen.

[0044] As a further solution of the present invention, the pen drop detection module includes a capacitance detection probe, which is connected to the metal pen tip. When the pen tip is suspended in the air or dropped on paper, the capacitance value to the ground measured by the probe will be different. This characteristic is used to detect whether the pen has dropped.

[0045] As a further solution of the present invention, the main control chip on the circuit board 14 uses a soc with wifi bluetooth function, and the posture acquisition module uses a posture sensor to measure three-axis acceleration, three-axis angular velocity and three-axis magnetic field. The main control chip calculates three Euler angles roll, pitch, and yaw, thereby calculating the information of relative displacement and posture transformation. By integrating the acceleration data, the posture sensor collects the speed information and position information of the sensor in space. The information sequence collected by these posture sensors constitutes an information matrix uploaded to the host computer. After the information is transmitted, the host computer pre-processes the information and sends it to the neural network for recognition. The Chinese characters recognized by the host computer are displayed according to the needs of the user, and can also be used as input information as the input of the association model to associate Chinese characters.

[0046] The pen tube of the present invention is provided with a small planed surface with a width of 3.74mm and a depth of 0.3mm at the pen holding position, which helps the user to stabilize the pen holding state when holding the pen, and at the same time facilitates the posture sensor to locate the writing posture; the middle and rear ends of the pen rear tube 15 have two card slots, one for fixing the battery 13, and the other for fixing the circuit board 14; the position where the lower end of the pen tube 12 contacts the pen cover 11 is also provided with two semicircular small grooves for placing metal foil as contacts. When the pen cover 11 is covered at the pen tip, the main control chip is in a disabled state and the control circuit does not work; after removing the pen cover 11, the control circuit works, and then the user selects the working mode of the pen through the touch sensing switch (the position of the copper foil) at the end of the pen rear tube 15. In normal mode, the present invention can be used as an ordinary pen, and the product functions described in the present invention are mainly in the identification working mode.

[0047] Figure 1The composition of the various parts of the pen, the pen cap 11 of the present invention, the outer diameter is 20.16mm, the inner diameter is 17.96mm, the overall length is 100.82mm, one side has a fixing structure that pen caps basically have, mainly used to be fixed to books, it cooperates with the pen tube in a conventional manner, there is a hemisphere with a radius of 0.74mm 1.5mm away from the end of the pen cap inside, four of which are linearly distributed in a circle, they can be tightly matched with the pen tube, the pen tube 12, length 88.72mm, outer diameter 14.61mm, inner diameter 12.41mm, the pen tube is provided with a small planed surface with a depth of 0.37mm on the pen grip, which is used to locate the approximate direction of the posture sensor when holding the pen, and also helps to improve the comfort of holding the pen, and is convenient for fixing the way of holding the pen when writing. The connection between the pen tube and the rear tube of the pen is a threaded connection. There are two contacts at the contact point between the pen tube and the pen cap. The metal ring on the pen cap will short the contact points and disable the main control chip. After the pen cap is removed, the two contacts are disconnected and the system is turned on. Battery 13 (size: 6mm thick, 12mm wide, 30mm long), pen back tube 15, the front half of which has a 5mm thick partition for isolating the pen core and other internal parts, the middle of the partition has a 4mm diameter and 2.5mm deep groove for fixing the end of the pen core, and there are battery slots and circuit board slots inside to facilitate fixing the battery and circuit board to the middle and rear end of the pen tube, pen back cover 16, it cooperates with the pen back tube as a plug cover, and there is a circle of protruding parts at the rear half of the pen back cover, which can be used to more conveniently remove the back cover from the back tube, and the rear end of the pen back tube can be opened to facilitate charging the battery. Figure 3 This is a schematic diagram of the circuit structure of the power module of the present invention. The battery is charged by using the battery management chip TP4056, and the battery voltage is adjusted to 3.3 volts using AMS1117-3.3 to power the circuit. Figure 4 The attitude acquisition module uses a direction sensor, which is the direction sensor MPU6050, to collect information such as direction and acceleration and send it to the MCU through the iic protocol. Figure 5 The main control chip ESP-S3 receives data from MPU6050 and sends it to the host computer via WIFI for processing. Figure 6 This is the schematic diagram of the peripheral function circuit of ESP32-S3.

[0048] A slot is provided inside the rear tube 15 of the stylus pen for fixing detachable components such as the circuit board 14 and the battery 13 , as well as fixing the pen core.

[0049] The pen back cover 16 can be removed, and after removal, the Type-C interface on the circuit board 14 can be connected to a charger to charge the battery.

[0050] When working in normal mode, its function is no different from that of an ordinary pen, and it can write and replace refills normally. The refills used can be purchased on the market. At this time, the circuit inside the pen works in ultra-low power consumption mode, and the button at the end of the pen back tube 15 can be used to switch modes.

[0051] By long pressing the touch-sensitive switch at the end of the pen, it starts trying to connect to the host computer. At this time, the corresponding indicator light gives a signal. After the connection is successful, the indicator light gives another signal, indicating that the connection is successful, and then the collection and transmission of posture information begins.

[0052] By long pressing the touch-sensitive switch at the end of the pen, it starts trying to connect to the host computer. At this time, the indicator light on the control circuit gives a signal. (The thickness of the pen tube designed at the indicator light position of the early product is thinner, so the indicator signal can be seen more clearly.) After the connection is successful, the indicator light gives another signal, indicating that the connection is successful, and then detects the pen drop by sensing the change in capacitance of the pen tip.

[0053] The present invention uses short refills available on the market, such as the mg0097 gel pen refill produced by Morning Glory Stationery Company. The pen tip and the capacitance measurement pin of the control microcontroller are connected to a shielded wire, and then the wire with the shielding layer is grounded to the shielding layer to shield the influence of irrelevant electric fields on capacitance measurement.

[0054] The electromagnetic information part of the pen consists of a control circuit, a posture acquisition module, a battery management module, and a pen drop detection module. The control circuit can be a low-power WiFi Bluetooth SOC produced by Espressif - ESP32 S3, which has Bluetooth 5.0, WiFi and other functions. It also has a wealth of peripheral interfaces to facilitate further expansion of functions.

[0055] Before fixing the main control chip to the circuit board, the first program is burned using the serial port. For subsequent program burning, use OTA to download the compiled firmware to ESP32 S3 to facilitate future function updates and upgrades.

[0056] The attitude acquisition module can use the six-axis acceleration and angular velocity sensor produced by Invensense, which includes three-axis acceleration and three-axis angular velocity. It is connected to the control circuit through the IIC interface. The control circuit performs low-pass filtering on the read data to filter out the high-frequency noise in the attitude signal, and then uses the Kalman filter algorithm to fuse the acceleration and angular velocity data to obtain the module attitude expressed by Euler angles, including pitch angle pitch, roll angle roll, and yaw angle yaw. Then the three-axis acceleration is normalized. This part of the collected data and the calculated data together constitute a frame of data to be sent.

[0057] When sending data, the data header is sent first, and then the data frame in byte form is sent to the host computer. A data packet contains the collected and calculated data as described above, and also contains a checksum calculated through checksum calculation. When the host computer receives data, it first receives the data header. After receiving the data header, it then receives data of a certain length. Finally, it calculates the checksum of this section of data. If the checksum of this data is the same as the checksum data in the data packet, the data is retained; otherwise, the data is discarded.

[0058] After detecting the pen tip touching down, start transmitting the pen's attitude data, and stop sending the data after lifting the pen. The host computer preprocesses the attitude data sequence during the period when the pen tip touches down, such as normalizing, scaling, and interpolating, and then sends it into the neural network for learning and classification.

[0059] The neural network is built using Google's machine learning framework TensorFlow and includes fully connected layers, pooling layers, convolutional layers, and RELU layers. The neural network has the following characteristics:

[0060] a. Local perception: When performing convolution, the data area covered by the convolution kernel each time is only a small part, which is local features, so it is called local perception. CNN is a process from local to global (the realization from local to global is in the fully connected layer), while the traditional neural network is only a global process. b. Weight sharing: The number of parameters in the traditional neural network is extremely large. In CNN, except for the fully connected layer, the parameters of the convolutional layer completely depend on the settings of the filter size. Compared with the traditional neural network, the new neural network has fewer parameters and less computational complexity, and the entire data matrix shares a set of filter parameters. c. Multiple convolutional kernels: One convolutional kernel represents one feature. To obtain more different feature sets, there are multiple convolutional kernels in the convolutional layer to generate different features.

[0061] The completion of the model includes the production of the dataset. When making the dataset, use the pen described above to make the dataset one by one. First, create folders according to the number of labels and label names to be classified. For example, if the model needs to recognize the stroke of a left-falling stroke, a folder needs to be created and named "pie" in Chinese pinyin. Then, write a large number of pen attitude sequences received by the host computer when writing this stroke into the folder and save them in the csv file format. One csv file contains one piece of data, so that the folder containing a large number of csv files constitutes the dataset for writing classification.

[0062] The dataset includes a noise dataset, that is, in addition to normal data categories, blank data without corresponding strokes or characters is added to distinguish between valid writing and invalid writing.

[0063] After receiving the data, the host computer pre-processes the data and further extracts the data features. After enhancing the data, the host computer sends it to the trained convolutional neural network model to obtain the output classification data, thereby recognizing the characters written by the user;

[0064] The program of the host computer further analyzes the received program and extracts other data generated during writing, such as maximum acceleration, maximum angular velocity, average acceleration, average angular velocity, and the difference between the posture data at each moment and the average value. The data is then combined into a matrix and normalized and centralized. After data preprocessing, the data will be sent to the built deep learning network for learning or prediction.

[0065] The convolutional neural network used in the neural network model is built using the open source machine learning framework tensorflow. The hidden layers of the model include convolutional layer, pooling layer, fully connected layer, and relu layer. The output classification data corresponds to the corresponding numbers and related strokes and characters.

[0066] When the host computer recognizes Chinese characters or English characters, it will group the continuously recognized characters into a stroke sequence, and then compare the recognized stroke sequence with the stroke sequence in the Chinese character library, and infer the Chinese character with the highest probability based on the similarity of the stroke sequence, and take the Chinese character with the highest recognition probability as the recognition result;

[0067] When the neural network of the host computer uses the data set to train the model, it first establishes a correspondence between the file path and the characters and strokes to which the file belongs, and reads them into the memory in a format corresponding to the matrix and label. Set the random number seed, divide the data set, validation set, and test set, and then train the neural network. Adjust the model parameters according to the model accuracy, and finally obtain the ideal character, stroke, and classification model.

[0068] The specific embodiments of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments. Various changes can be made within the knowledge scope of ordinary technicians in this field without departing from the purpose of the present invention.

Claims

1. A handwriting recognition device, characterized in that: Including stylus pen, identification display terminal; The stylus pen includes a battery, a control circuit, a posture acquisition module, a battery management module, a writing device that realizes the basic functions of the pen, and a pen drop detection module; The battery is used for power supply; The control circuit, the posture acquisition module, the battery management module, and the pen drop detection module are integrated on a circuit board (14); the pen drop detection module and the posture acquisition module are connected to the control circuit, and the battery management module is connected to the battery; the circuit board (14) is arranged on a writing device that realizes the basic functions of a pen; The control circuit includes a radio frequency circuit and a main control chip; the control circuit is connected to the information receiving module in the host computer through the radio frequency circuit thereon; The identification and display terminal adopts a host computer, which includes an information receiving module, a preprocessing module, a neural network module, and a display output module; the information receiving module, the preprocessing module, the neural network module, and the display output module are connected in sequence; The preprocessing module is used to perform data enhancement and denoising on the posture data collected by the posture collection module; The neural network module is used to recognize the posture data processed by the preprocessing module during the writing process, learn and recognize the strokes or characters written by the user through machine learning, and display them through the display output module; The writing device realizing the basic functions of a pen comprises a pen cap (11), a middle pen tube (12), a rear pen tube (15), a rear pen cover (16), and a pen core; the middle pen tube (12) is threadedly connected to the rear pen tube (15), a battery (13) and a circuit board (14) are arranged in the rear pen tube (15), and the rear pen cover (16) covers the rear pen tube (15); The pen cap (11) is sleeved with a metal ring, and the pen tube (12) is provided with two contacts, both of which are connected to the enable end of the main control chip. When the pen cap (11) is sleeved on the pen tube (12), the metal ring connects the two contacts, thereby disabling the enable end of the main control chip and making the control circuit inoperative; when the pen cap (11) is removed, the two contacts are disconnected and the control circuit operates; A copper foil is attached to the end of the pen rear tube (15), which is connected to the capacitance measurement pin of the single-chip microcomputer in the control circuit. The working state of the control circuit is switched by detecting the capacitance of the copper foil to the ground. When the finger touches the end of the pen rear tube (15), the capacitance of the copper foil to the ground is changed, thereby transmitting a switch pressing signal to the single-chip microcomputer; Through this pressed signal, the control circuit can control the working mode of the pen; The pen drop detection module includes a capacitance detection probe, which is connected to the metal pen tip. When the pen tip is suspended in the air or when the pen is dropped on paper, the capacitance value to the ground measured by the probe will be different. Through this characteristic, it is detected whether the pen is dropped; The main control chip on the circuit board (14) uses a soc with wifi and bluetooth functions. The attitude acquisition module uses an attitude sensor to measure three-axis acceleration, three-axis angular velocity and three-axis magnetic field. The main control chip calculates three Euler angles roll, pitch and yaw, thereby calculating relative displacement and attitude change information. By integrating the acceleration data, the attitude sensor collects the speed information and position information of the sensor in space. The information sequences collected by these attitude sensors constitute an information matrix uploaded to the host computer. After the information is transmitted, the host computer pre-processes the information and sends it to the neural network for recognition.

2. The handwriting recognition device according to claim 1, characterized in that: The main control chip adopts MCU, and the MCU adopts esp32 s3, SOC or single chip microcomputer.

3. The handwriting recognition device according to claim 1, characterized in that: The pen tube (12) is provided with a planed surface for better positioning the direction of the posture acquisition module when holding the pen. The position of the posture acquisition module in the pen is fixed. Because the thumb is pressed on the planed surface, the position of the thumb and the posture acquisition module is relatively fixed. Then, the position of the entire hand and the posture acquisition module is relatively fixed, thereby reducing the complexity of data processing and helping to improve the comfort of holding the pen.

Images