CN114519374A - Handwritten letter signal recognition system and method based on flexible sensor - Google Patents
Handwritten letter signal recognition system and method based on flexible sensor Download PDFInfo
- Publication number
- CN114519374A CN114519374A CN202210137709.0A CN202210137709A CN114519374A CN 114519374 A CN114519374 A CN 114519374A CN 202210137709 A CN202210137709 A CN 202210137709A CN 114519374 A CN114519374 A CN 114519374A
- Authority
- CN
- China
- Prior art keywords
- convolution kernel
- signal
- layer
- strain
- handwritten
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
- G01B7/00—Measuring arrangements characterised by the use of electric or magnetic techniques
- G01B7/16—Measuring arrangements characterised by the use of electric or magnetic techniques for measuring the deformation in a solid, e.g. by resistance strain gauge
- G01B7/18—Measuring arrangements characterised by the use of electric or magnetic techniques for measuring the deformation in a solid, e.g. by resistance strain gauge using change in resistance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Character Discrimination (AREA)
Abstract
The invention discloses a handwritten letter signal recognition system and method based on a flexible sensor. The flexible sensor is arranged at the finger joint, senses the voltage change generated by the strain under the bending of the finger joint in the process of writing letters by hand and sends the voltage change to the semiconductor characteristic analyzer; the semiconductor characteristic analyzer obtains a real-time resistance signal of the flexible sensor and transmits the real-time resistance signal to the signal processor; and the signal processor is internally provided with a multi-stream convolution neural network, receives the signal of the semiconductor characteristic analyzer, and further obtains a classification result of the handwritten letters corresponding to the output handwritten strain signal through analysis and processing of the multi-stream convolution neural network. The conductive material of the flexible sensor is a liquid metal material, so that the flexible sensor has very high sensitivity and conductivity, and the feasibility and accuracy of letter recognition are ensured. The identification method can effectively judge the letter category according to the transmitted resistance change, and is more intuitive and accurate.
Description
Technical Field
The invention relates to a signal recognition system, in particular to a handwritten letter signal recognition system and method based on a flexible sensor.
Background
In recent years, with the synergistic progress in the fields of computer science, optoelectronics, material science, mechanics and the like, the artificial intelligence technology has not been developed rapidly before, and intelligent wearable equipment for converting external stimuli such as force, strain, temperature, humidity and the like into electric signals is developed vigorously. The traditional silicon-based semiconductor and metal-based rigid strain sensor has the advantages of high resolution, small error, capability of measuring dynamic and static forces and the like, but the inherent rigidity and brittleness of the sensor cause poor flexibility and inextensibility, and a very small strain range (less than 5 percent) is difficult to meet the requirements of high stretching and high sensitivity at the same time, so that the measurement stability, the measurement precision and the measurement range of the sensor are seriously influenced, the normal physiological activity of a human body can bear 30 percent of strain on the skin and about 100 percent of strain on joints, and the traditional rigid material can fail under the strain of 1 to 3 percent, so that the development and the application of the sensor are limited. As one of the indispensable important components in smart wearable devices, flexible strain sensors naturally become a hotspot of scientific research today. Although the sensitivity of the conventional strain sensor can be greatly improved by the materials such as silver nanowires, carbon nanotubes, graphene and the like, the mechanical strength of the sensor is reduced due to the inherent brittleness of the materials, so that the practical use of the sensor is limited, and the durability and the reliability of the sensor are influenced.
Disclosure of Invention
In order to solve the problems in the background art, the invention provides a handwritten letter signal recognition system and method based on a flexible sensor.
The technical scheme adopted by the invention is as follows:
a handwritten letter signal recognition system based on a flexible sensor comprises:
the handwritten letter signal recognition system comprises a handwritten strain signal acquisition device and a signal processor, wherein the handwritten strain signal acquisition device is electrically connected with the signal processor; the handwriting strain signal acquisition device comprises a flexible sensor and a 4200 semiconductor characteristic analyzer, wherein the flexible sensor is electrically connected with the 4200 semiconductor characteristic analyzer, and the 4200 semiconductor characteristic analyzer is electrically connected with a signal processor.
The flexible sensor is arranged at the finger joint and used for sensing voltage change generated by strain under bending at the finger joint in the process of writing letters and sending the voltage change to the 4200 semiconductor characteristic analyzer; 4200 the semiconductor characteristic analyzer is used for receiving the voltage change from the flexible sensor and obtaining a real-time resistance signal of the flexible sensor and transmitting the signal to the signal processor.
The signal processor is internally provided with a multi-stream convolution neural network and used for receiving 4200 semiconductor characteristic analyzer signals, and then classification results of handwritten letters corresponding to the output handwritten strain signals are obtained through analysis processing of the multi-stream convolution neural network.
Secondly, a handwritten letter signal recognition method based on a flexible sensor comprises the following steps:
the handwritten letter signal recognition method comprises the following steps:
1) and (4) preparing a flexible sensor.
2) The flexible sensor is placed at the finger joint and electrically connected 4200 to the semiconductor feature analyzer.
3) The method comprises the steps of handwriting a plurality of letters by a pen, repeatedly handwriting the letters for a plurality of times, receiving bending strain change of a flexible sensor during handwriting through a 4200 semiconductor characteristic analyzer, calculating strain resistance signals during handwriting of the letters to serve as handwriting strain signals, further obtaining a plurality of handwriting strain signals corresponding to the handwritten letters, and forming a signal data set by the handwriting strain signals corresponding to the handwritten letters.
4) Carrying out standardization processing on the signal data set, and selecting a plurality of handwriting strain signals as a training set; the remaining part serves as a test set.
5) Building a multi-stream convolution neural network, and arranging the multi-stream convolution neural network in a signal processor; inputting a training set into a multi-stream convolution neural network in a signal processor for training;
and inputting the test set into the trained multi-stream convolution neural network for verification, and obtaining a verification classification result of the handwritten letters corresponding to each handwritten pressure signal in the test set.
6) And repeating the step 3), obtaining a plurality of handwriting strain signals again to form a target signal data set, inputting the target signal data set into the trained multi-stream convolution neural network for processing, and obtaining a classification result of the handwritten letters corresponding to each handwriting pressure strain in the target signal data set.
In the step 1), the preparation of the flexible sensor comprises the following steps:
1.1) adding acrylamide and acrylic acid into deionized water for dissolving, and then sequentially adding zirconium oxychloride octahydrate and a photoinitiator 2-hydroxy-4' - (2-hydroxyethoxy) -2-methyl propiophenone to obtain a mixed solution.
1.2) placing the mixed solution on a magnetic stirrer and stirring for 30min until solid small particles in the mixed solution are fully dissolved to form colorless transparent liquid, and then introducing argon for 30min to exhaust dissolved oxygen in the colorless transparent solution to obtain a transparent solution.
1.3) injecting the transparent solution into a hydrogel mould with a thickness of 0.5mm, and placing under an ultraviolet lamp for illumination polymerization for 5min to obtain a semi-polymerized poly (acrylamide-co-acrylic acid)/zirconium ion hydrogel without swelling equilibrium; half of the semi-polymerized poly (acrylamide-co-acrylic acid)/zirconium ion hydrogel is placed in a large amount of deionized water to be soaked for 7 days until the swelling balance is achieved, the quality of the hydrogel is kept unchanged, and the poly (acrylamide-co-acrylic acid)/zirconium ion hydrogel substrate is obtained.
1.4) placing the poly (acrylamide-co-acrylic acid)/zirconium ion hydrogel substrate in a glass dish, printing a liquid metal conductive circuit with uniform line width on the poly (acrylamide-co-acrylic acid)/zirconium ion hydrogel substrate by using a 3D printer, and respectively bonding and electrically connecting two ends of the liquid metal conductive circuit with one end of two copper electrodes; and simultaneously transferring the poly (acrylamide-co-acrylic acid)/zirconium ion hydrogel substrate, the liquid metal conductive circuit and the two copper electrodes in the glass dish into another hydrogel mold prepared in advance and having the thickness of 1mm, and uniformly coating the other half of the semi-polymerized poly (acrylamide-co-acrylic acid)/zirconium ion hydrogel on the copper electrodes and the liquid metal conductive circuit until the semi-polymerized poly (acrylamide-co-acrylic acid)/zirconium ion hydrogel completely covers the liquid metal conductive circuit and the two copper electrodes to obtain the semi-polymerized conductive hydrogel strain sensor.
1.5) placing the semi-polymerized conductive hydrogel strain sensor under an ultraviolet lamp for illumination polymerization to obtain a liquid metal/hydrogel strain sensor, namely a flexible sensor.
In the step 1.1), 0.11 to 3.73g of acrylamide, 0.54 to 6.81g of acrylic acid, 0.48 to 4.83g of zirconium oxychloride octahydrate and 0.03 to 0.24g of photoinitiator are added by mass.
Mainly comprises 10-50mol percent of acrylamide, 10-50mol percent of acrylic acid, 0.1-1.0mol/L of zirconium oxychloride octahydrate and 1.0mol percent of photoinitiator in terms of mole fraction (or concentration).
In the step 1.4), the other ends of the two copper electrodes protrude out of the flexible sensor.
In the step 2), the other ends of the two electrode bulges are respectively and electrically connected with a red and black wire of a 4200 semiconductor characteristic analyzer.
In the step 4), for each handwriting strain signal in the signal data set, the following operations are performed:
and (3) carrying out standardization processing on the handwriting strain signal, scaling the original data of the handwriting strain signal between 0 and 1, and keeping the distribution of the original data of the handwriting strain signal.
In the step 5), the multi-stream convolutional neural network is mainly composed of a multi-stream decomposition stage and a fusion stage which are sequentially performed.
The multi-stream decomposition stage comprises four same decomposition parts which are sequentially arranged, wherein each decomposition part comprises a 32 convolution kernel convolution layer, a 64 convolution kernel convolution layer, a 128 convolution kernel convolution layer and a 128 convolution kernel local connection layer; the fusion stage includes 512, 256, 128, and 64 convolutional-kernel fully-connected layers.
The input of the multi-stream convolutional neural network is averagely divided into four input parts with equal length, the four input parts respectively correspond to four same decomposition parts which are sequentially arranged in a multi-stream decomposition stage, and the four input parts are respectively and correspondingly input into the four decomposition parts; for each input portion and corresponding decomposition portion, the following operations are performed:
the input part is sequentially input into the 32 convolution kernel convolutional layer, the 64 convolution kernel convolutional layer, the 128 convolution kernel convolutional layer and the 128 convolution kernel local connecting layer of the decomposition part for processing, and a linear rectification function and a BN algorithm which are sequentially performed are connected between the 32 convolution kernel convolutional layer and the 64 convolution kernel convolutional layer, between the 64 convolution kernel convolutional layer and the 128 convolution kernel convolutional layer, between the 128 convolution kernel convolutional layer and the 128 convolution kernel local connecting layer and behind the 128 convolution kernel local connecting layer.
The outputs of the 32 convolution kernel convolutional layer, the 64 convolution kernel convolutional layer, the 128 convolution kernel convolutional layer and the 128 convolution kernel local connecting layer are input into a linear rectification function for processing, and the output of the linear rectification function is processed by a BN algorithm to be used as the input of the 64 convolution kernel convolutional layer, the 128 convolution kernel convolutional layer, the 128 convolution kernel local connecting layer and the final output of the decomposition part.
And a linear rectification function is connected between the 512 convolution kernel full-connection layer and the 256 convolution kernel full-connection layer, between the 256 convolution kernel full-connection layer and the 128 convolution kernel full-connection layer, and between the 128 convolution kernel full-connection layer and the 64 convolution kernel full-connection layer, and a SoftMax classifier is connected behind the 64 convolution kernel full-connection layer.
And finally outputting the decomposition parts, performing characteristic fusion in a fusion stage, sequentially processing a fusion result by a 512 convolution kernel full-link layer, a 256 convolution kernel full-link layer, a 128 convolution kernel full-link layer and a 64 convolution kernel full-link layer, inputting the outputs of the 512 convolution kernel full-link layer, the 256 convolution kernel full-link layer and the 128 convolution kernel full-link layer into a linear rectification function for processing, respectively serving as the inputs of the 256 convolution kernel full-link layer, the 128 convolution kernel full-link layer and the 64 convolution kernel full-link layer, and finally processing the output of the 64 convolution kernel full-link layer by a SoftMax classifier to output a classification result.
The invention has the beneficial effects that:
1. among a plurality of flexible substrate materials, the hydrogel material has the advantages of good water absorption, flexibility, elasticity, biocompatibility and the like; liquid metal, which is a conductive material that is liquid at room temperature, is one of the most flexible and deformable conductors. Unlike traditional metal materials, liquid metal has the advantages of mechanical flexibility, high reliability, fast response, low hysteresis, high conductivity, low toxicity, etc.
2. The handwritten letter signal identification method is to judge the written letters from the resistance change, so that the letters can be identified more intuitively and more accurately.
3. The flexible sensor is the original sensor, and due to the special characteristics of the material (liquid metal) of the flexible sensor, the flexible sensor has very high sensitivity and conductivity, so that the feasibility and the accuracy of letter recognition are ensured. Based on high sensitivity and high accuracy of identification, written letters can be effectively obtained according to the transmitted resistance change, so that the method can be widely applied to the field of intellectualization.
Drawings
FIG. 1 is a graph of resistance change for different letters;
FIG. 2 is a graph of resistance changes for different letters being handwritten consecutively;
FIG. 3 is a schematic diagram of a multi-stream convolutional neural network structure;
fig. 4 shows the recognition accuracy of different letters.
Detailed Description
The invention is described in further detail below with reference to the figures and the embodiments.
The specific embodiment is as follows:
in specific implementation, the handwritten letters are set as 26 English letters; handwriting 26 letters by holding a pen with hands, and repeating handwriting 40 times for each letter; setting the current of a 4200 semiconductor feature analyzer to be 0.01A, and acquiring resistance change by the 4200 semiconductor feature analyzer through receiving voltage change of a flexible sensor so as to acquire 40 handwriting strain signals corresponding to each handwritten letter, wherein the 40 handwriting strain signals corresponding to each handwritten letter form a signal data set;
as shown in fig. 1, resistance change maps of different english letters, such as "a", "B", "C", "D", "E", "N", "O", "R", and "S", are respectively obtained; as shown in fig. 2, the resistance change diagrams are written in succession with "S", "E", "N", "S", "O", and "R".
The sensitivity of the flexible sensor can be obtained from the resistance change, and the sensitivity formula GF is as follows:
wherein, beta represents a fitting function for describing the contact resistance between the copper electrode and the liquid metal conductive circuit; ε represents the strain of the flexible sensor; Δ R represents a resistance change value of the flexible sensor; Δ L represents a liquid metal conductive circuit path length variation value; r0Representing an initial resistance of the flexible sensor; l is0Indicating a length of a conductive path of the liquid metal conductive circuit;
and (3) carrying out standardization processing on the signal data set, and selecting 60% of handwriting strain signals as a training set and the remaining 40% of handwriting strain signals as a testing set. And inputting the test set into the trained multi-stream convolution neural network for verification, and obtaining a verification classification result of the handwritten letters corresponding to each handwritten strain signal in the test set. Fig. 3 is a schematic structural diagram of the multi-stream convolutional neural network. And a plurality of handwriting strain signals can be acquired again to form a target signal data set, the target signal data set is input into the trained multi-stream convolution neural network for processing, and the classification result of the handwriting letters corresponding to each handwriting strain signal in the target signal data set is acquired.
The sensitivity of the system is 1.03, the final recognition accuracy acc is 60.75%, and as shown in fig. 4, the final recognition accuracy acc is the recognition accuracy of a single letter.
Claims (7)
1. A handwritten letter signal recognition system based on flexible sensors is characterized in that:
the handwriting strain signal acquisition device is electrically connected with the signal processor; the handwriting strain signal acquisition device comprises a flexible sensor and a 4200 semiconductor characteristic analyzer, wherein the flexible sensor is electrically connected with the 4200 semiconductor characteristic analyzer, and the 4200 semiconductor characteristic analyzer is electrically connected with a signal processor;
the flexible sensor is arranged at a finger joint, senses voltage change generated by strain under bending of the finger joint in the process of writing letters and sends the voltage change to the 4200 semiconductor characteristic analyzer; 4200 the semiconductor characteristic analyzer receives the voltage change from the flexible sensor and obtains the real-time resistance signal of the flexible sensor, and transmits the signal to the signal processor;
the signal processor is internally provided with a multi-stream convolution neural network, receives a signal of the 4200 semiconductor characteristic analyzer, and further obtains a classification result of a handwritten letter corresponding to the output handwritten strain signal through analysis processing of the multi-stream convolution neural network.
2. A handwritten alphabet signal recognition method of the handwritten alphabet signal recognition system as claimed in claim 1, wherein:
the method comprises the following steps:
1) preparing a flexible sensor;
2) disposing a flexible sensor at the finger joint and electrically connecting 4200 the flexible sensor to the semiconductor feature analyzer;
3) handwriting a plurality of letters by using a pen, wherein each letter is repeatedly handwritten for a plurality of same times, the bending strain change of the flexible sensor during handwriting is received through a 4200 semiconductor characteristic analyzer, strain resistance signals during handwriting of each letter are calculated to be used as handwriting strain signals, a plurality of handwriting strain signals corresponding to each handwritten letter are further obtained, and the handwriting strain signals corresponding to each handwritten letter form a signal data set;
4) carrying out standardization processing on the signal data set, and selecting a plurality of handwriting strain signals as a training set;
5) building a multi-stream convolution neural network, and arranging the multi-stream convolution neural network in a signal processor; inputting a training set into a multi-stream convolution neural network in a signal processor for training;
6) and repeating the step 3), obtaining a plurality of handwriting strain signals again to form a target signal data set, inputting the target signal data set into the trained multi-stream convolution neural network for processing, and obtaining a classification result of the handwritten letters corresponding to each handwriting strain signal in the target signal data set.
3. A handwritten-letter signal recognition method according to claim 2, characterized in that:
in the step 1), the preparation of the flexible sensor comprises the following steps:
1.1) adding acrylamide and acrylic acid into deionized water for dissolving, and then sequentially adding zirconium oxychloride octahydrate and photoinitiator 2-hydroxy-4' - (2-hydroxyethoxy) -2-methyl propiophenone to obtain a mixed solution;
1.2) placing the mixed solution on a magnetic stirrer for stirring until the mixed solution becomes colorless transparent liquid, and then introducing argon to obtain transparent solution;
1.3) injecting the transparent solution into a prepared hydrogel mould, and placing the hydrogel mould under an ultraviolet lamp for illumination polymerization for 5min to obtain semi-polymerized poly (acrylamide-co-acrylic acid)/zirconium ion hydrogel; placing half of the semi-polymerized poly (acrylamide-co-acrylic acid)/zirconium ion hydrogel in deionized water to be soaked for 7 days to obtain a poly (acrylamide-co-acrylic acid)/zirconium ion hydrogel substrate;
1.4) placing the poly (acrylamide-co-acrylic acid)/zirconium ion hydrogel substrate in a glass dish, printing a liquid metal conductive circuit on the poly (acrylamide-co-acrylic acid)/zirconium ion hydrogel substrate by using a 3D printer, and respectively bonding and electrically connecting two ends of the liquid metal conductive circuit with one end of two copper electrodes; simultaneously transferring the poly (acrylamide-co-acrylic acid)/zirconium ion hydrogel substrate, the liquid metal conductive circuit and the two copper electrodes in the glass dish into another hydrogel mold prepared in advance, and coating the other half of the semi-polymerized poly (acrylamide-co-acrylic acid)/zirconium ion hydrogel on the copper electrodes and the liquid metal conductive circuit until the semi-polymerized poly (acrylamide-co-acrylic acid)/zirconium ion hydrogel completely wraps the liquid metal conductive circuit and the two copper electrodes to obtain a semi-polymerized conductive hydrogel strain sensor;
1.5) placing the semi-polymerized conductive hydrogel strain sensor under an ultraviolet lamp for illumination polymerization to obtain a liquid metal/hydrogel strain sensor, namely a flexible sensor.
4. A handwritten character signal recognition method according to claim 3, characterized in that:
in the step 1.1), 0.11 to 3.73g of acrylamide, 0.54 to 6.81g of acrylic acid, 0.48 to 4.83g of zirconium oxychloride octahydrate and 0.03 to 0.24g of photoinitiator are added by mass.
5. A handwritten character signal recognition method according to claim 3, characterized in that:
in the step 1.4), the other ends of the two copper electrodes protrude out of the flexible sensor;
in the step 2), the other ends of the two electrode protrusions are respectively electrically connected with the red and black wires of the 4200 semiconductor characteristic analyzer.
6. A handwritten-letter signal recognition method according to claim 2, characterized in that:
in the step 4), for each handwriting strain signal in the signal data set, the following operations are performed:
and (3) carrying out standardization processing on the handwriting strain signal, scaling the original data of the handwriting strain signal between 0 and 1, and keeping the distribution of the original data of the handwriting strain signal.
7. A handwritten-letter signal recognition method according to claim 2, characterized in that:
in the step 5), the multi-stream convolutional neural network is mainly composed of a multi-stream decomposition stage and a fusion stage which are sequentially performed;
the multi-stream decomposition stage comprises four same decomposition parts which are sequentially arranged, wherein each decomposition part comprises a 32 convolution kernel convolution layer, a 64 convolution kernel convolution layer, a 128 convolution kernel convolution layer and a 128 convolution kernel local connection layer; the fusion stage comprises a 512 convolution kernel full link layer, a 256 convolution kernel full link layer, a 128 convolution kernel full link layer and a 64 convolution kernel full link layer;
the input of the multi-stream convolutional neural network is averagely divided into four input parts with equal length, the four input parts respectively correspond to four same decomposition parts which are sequentially arranged in a multi-stream decomposition stage, and the four input parts are respectively and correspondingly input into the four decomposition parts; for each input part and corresponding decomposition part, the following operations are performed:
the input part is sequentially input into the 32 convolution kernel convolutional layer, the 64 convolution kernel convolutional layer, the 128 convolution kernel convolutional layer and the 128 convolution kernel local connecting layer of the decomposition part for processing, and a linear rectification function and a BN algorithm which are sequentially performed are connected between the 32 convolution kernel convolutional layer and the 64 convolution kernel convolutional layer, between the 64 convolution kernel convolutional layer and the 128 convolution kernel convolutional layer, between the 128 convolution kernel convolutional layer and the 128 convolution kernel local connecting layer and behind the 128 convolution kernel local connecting layer;
the outputs of the 32 convolution kernel convolutional layer, the 64 convolution kernel convolutional layer, the 128 convolution kernel convolutional layer and the 128 convolution kernel local connecting layer are input into a linear rectification function for processing, and the output of the linear rectification function is processed by a BN algorithm to be respectively used as the input of the 64 convolution kernel convolutional layer, the 128 convolution kernel convolutional layer and the 128 convolution kernel local connecting layer and the final output of the decomposition part;
a linear rectification function is connected between the 512 convolution kernel full connection layer and the 256 convolution kernel full connection layer, between the 256 convolution kernel full connection layer and the 128 convolution kernel full connection layer, and between the 128 convolution kernel full connection layer and the 64 convolution kernel full connection layer, and a SoftMax classifier is connected behind the 64 convolution kernel full connection layer;
and finally outputting the decomposition parts, performing characteristic fusion in a fusion stage, sequentially processing a fusion result by a 512 convolution kernel full-link layer, a 256 convolution kernel full-link layer, a 128 convolution kernel full-link layer and a 64 convolution kernel full-link layer, inputting the outputs of the 512 convolution kernel full-link layer, the 256 convolution kernel full-link layer and the 128 convolution kernel full-link layer into a linear rectification function for processing, respectively serving as the inputs of the 256 convolution kernel full-link layer, the 128 convolution kernel full-link layer and the 64 convolution kernel full-link layer, and finally processing the output of the 64 convolution kernel full-link layer by a SoftMax classifier to output a classification result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210137709.0A CN114519374A (en) | 2022-02-15 | 2022-02-15 | Handwritten letter signal recognition system and method based on flexible sensor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210137709.0A CN114519374A (en) | 2022-02-15 | 2022-02-15 | Handwritten letter signal recognition system and method based on flexible sensor |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114519374A true CN114519374A (en) | 2022-05-20 |
Family
ID=81595985
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210137709.0A Pending CN114519374A (en) | 2022-02-15 | 2022-02-15 | Handwritten letter signal recognition system and method based on flexible sensor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114519374A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115028861A (en) * | 2022-07-26 | 2022-09-09 | 深圳微检无忧科技有限公司 | High-conductivity double-network hydrogel and preparation method and flexible sensing application thereof |
-
2022
- 2022-02-15 CN CN202210137709.0A patent/CN114519374A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115028861A (en) * | 2022-07-26 | 2022-09-09 | 深圳微检无忧科技有限公司 | High-conductivity double-network hydrogel and preparation method and flexible sensing application thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102207415B (en) | Conductive-rubber-based flexible array clip pressure sensor and manufacturing method | |
| Wen et al. | Wearable multimode sensor with a seamless integrated structure for recognition of different joint motion states with the assistance of a deep learning algorithm | |
| CN114519374A (en) | Handwritten letter signal recognition system and method based on flexible sensor | |
| CN208998966U (en) | A kind of capacitance type touch sensor based on zero Poisson's ratio structure | |
| CN111722723B (en) | Bidirectional bending flexible sensor, sign language recognition system and method | |
| CN108613758A (en) | A kind of capacitance type touch sensor based on zero Poisson's ratio structure | |
| Xie et al. | A Deep Learning-Enabled Skin-Inspired Pressure Sensor for Complicated Recognition Tasks with Ultralong Life | |
| CN110414633B (en) | System and method for recognizing handwritten fonts | |
| Chen et al. | Fabrication of wearable hydrogel sensors with simple ionic-digital conversion and inherent water retention | |
| CN109141696A (en) | A kind of flexible touch sensation sensor and its signal processing system based on piezoelectric membrane | |
| CN111443816A (en) | Gesture recognition system based on flexible antibacterial biological membrane multi-channel data acquisition module | |
| CN108711346A (en) | Flexible wearable blind person's reading apparatus | |
| Guo et al. | An intelligent dual-sensing e-skin system for pressure and temperature detection using laser-induced graphene and polydimethylsiloxane | |
| CN105509937A (en) | Pressure sensor, pressure detection method and manufacturing process | |
| Chen et al. | A bio-impedance analysis method based on human hand anatomy for hand gesture recognition | |
| Chen et al. | Wearable resistive-based gesture-sensing interface bracelet | |
| CN112985649B (en) | Mechanical information detection system based on flexible distributed capacitive touch sensor | |
| CN205562090U (en) | Pressure sensor | |
| Lu et al. | Artificial Intelligence–Enabled Gesture‐Language‐Recognition Feedback System Using Strain‐Sensor‐Arrays‐Based Smart Glove | |
| CN114699082A (en) | Flexible wearable surface electromyography sensor | |
| CN213582081U (en) | Gesture recognition system based on flexible antibacterial biological membrane multi-channel data acquisition module | |
| Atoche-Enseñat et al. | A Smart Tactile Sensing System Based on Carbon Nanotube/Polypropylene Composites for Wearable Applications | |
| CN209070491U (en) | A kind of pliable pressure sensing hand language recognition device | |
| Sun et al. | High‐accuracy dynamic gesture recognition: A universal and self‐adaptive deep‐learning‐assisted system leveraging high‐performance ionogels‐based strain sensors | |
| Jiang et al. | A wearable Braille recognition system based on high density tactile sensors |
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 |