CN112198966B - Stroke identification method and system based on FMCW radar system - Google Patents
Stroke identification method and system based on FMCW radar system Download PDFInfo
- Publication number
- CN112198966B CN112198966B CN202011420741.7A CN202011420741A CN112198966B CN 112198966 B CN112198966 B CN 112198966B CN 202011420741 A CN202011420741 A CN 202011420741A CN 112198966 B CN112198966 B CN 112198966B
- Authority
- CN
- China
- Prior art keywords
- matrix
- stroke
- chinese character
- feature
- algorithm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 67
- 238000000605 extraction Methods 0.000 claims abstract description 32
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 18
- 238000010586 diagram Methods 0.000 claims abstract description 11
- 239000011159 matrix material Substances 0.000 claims description 161
- 238000012549 training Methods 0.000 claims description 40
- 230000008569 process Effects 0.000 claims description 26
- 230000008859 change Effects 0.000 claims description 24
- 238000012360 testing method Methods 0.000 claims description 12
- 238000006243 chemical reaction Methods 0.000 claims description 9
- 238000009432 framing Methods 0.000 claims description 9
- 238000001228 spectrum Methods 0.000 claims description 7
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 238000011176 pooling Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 abstract description 18
- 230000006870 function Effects 0.000 description 23
- 230000003993 interaction Effects 0.000 description 13
- 230000009286 beneficial effect Effects 0.000 description 4
- 238000011161 development Methods 0.000 description 4
- 238000013459 approach Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 239000013598 vector Substances 0.000 description 2
- 206010047571 Visual impairment Diseases 0.000 description 1
- 238000005452 bending Methods 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000012790 confirmation Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000003623 enhancer Substances 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000005530 etching Methods 0.000 description 1
- 230000001771 impaired effect Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 208000029257 vision disease Diseases 0.000 description 1
- 230000004393 visual impairment Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/017—Gesture based interaction, e.g. based on a set of recognized hand gestures
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/02—Systems using the reflection of electromagnetic waves other than radio waves
- G01S17/06—Systems determining position data of a target
- G01S17/08—Systems determining position data of a target for measuring distance only
- G01S17/32—Systems determining position data of a target for measuring distance only using transmission of continuous waves, whether amplitude-, frequency-, or phase-modulated, or unmodulated
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
- G06V40/28—Recognition of hand or arm movements, e.g. recognition of deaf sign language
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Electromagnetism (AREA)
- Human Computer Interaction (AREA)
- General Engineering & Computer Science (AREA)
- Social Psychology (AREA)
- Psychiatry (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Character Discrimination (AREA)
Abstract
The invention discloses a stroke identification method and a system based on an FMCW radar system, wherein the method comprises the following steps: acquiring intermediate frequency signal data of at least one stroke to be recognized contained in handwritten Chinese characters based on an FMCW radar system; preprocessing intermediate frequency signal data of each stroke to be recognized to obtain a characteristic diagram set of each stroke to be recognized; acquiring a trained Chinese character basic stroke recognition model; the Chinese character basic stroke recognition model is a convolutional neural network model which takes a characteristic graph set as an input parameter and takes a basic stroke category as an output parameter; and inputting the characteristic diagram set of each stroke to be recognized into a Chinese character basic stroke recognition model, and acquiring a basic stroke category which is output by the Chinese character basic stroke recognition model and is matched with each stroke to be recognized. The method and the device reduce the data volume for representing the gesture motion trend, improve the characteristic extraction efficiency, and can accurately judge the basic stroke category.
Description
Technical Field
The invention belongs to the technical field of gesture recognition, and particularly relates to a stroke recognition method and system based on an FMCW radar system.
Background
At present, the commonly used human-computer interaction methods include a keyboard, a mouse, a handwriting pad, touch screen input and the like, and all the methods are contact human-computer interaction methods. In many special application scenarios the contact interactive device is subject to environmental restrictions, such as sterile operation of the operating room and operation by visually impaired users. However, with the development of science and technology, many intelligent terminal devices and human-computer interaction devices are on the market, and some smart phones are provided with barrier-free modes to provide voice feedback so that users do not need to look at a screen when using the devices, which can simplify the operation of visually-impaired users, but may have privacy problems. A special human-computer interaction device is designed for a user with visual impairment, the user needs to spend a lot of time to adapt to products, the user does not actively adapt to the user requirements, and the user needs to make changes.
In non-contact man-machine interaction equipment, man-machine interaction equipment based on vision has proposed a method for recognizing Chinese characters through a mobile phone camera: one is a method for framing, scanning and identifying Chinese characters through a camera, and the method requires a user to continuously adjust the vertical distance and the left and right positions of the camera to frame and click confirmation to obtain the Chinese characters to be identified, so that the operation is not well controlled; the other method is a method for identifying the smearing region by firstly shooting a picture containing the required Chinese characters, smearing the picture with the Chinese characters to be identified and then identifying the smearing region, wherein the method has excessive steps and the smearing position is not well grasped. Text input methods based on computer vision gestures have also been proposed: one approach is for people to use paper pens to write normally and then collect the writing ink on the paper through a camera to recognize the writing ink as characters. This method is still limited by external conditions (e.g., paper size, etc.); another approach is to write virtual characters on a desktop or in the air directly by fingers, collect the moving track of the fingers in real time by using a camera or a motion sensor (such as Kinect, etc.), and then recognize the track as characters.
The man-machine interaction equipment based on radio frequency only achieves simple gesture instruction operation, and in view of the published literature data at present, relevant research for recognizing the basic strokes of handwritten Chinese characters by utilizing a millimeter wave radar is not available. If the basic strokes of the handwritten Chinese characters can be accurately identified, the subsequent handwritten Chinese character human-computer interaction equipment can be developed, and the research can be applied to the barrier-free mode and the human-computer interaction fields of games, medical treatment, military and the like.
Disclosure of Invention
Aiming at the problems of the existing non-contact human-computer interaction equipment, the invention provides a stroke identification method and system based on an FMCW radar system.
The technical scheme of the invention is as follows:
in a first aspect, a stroke recognition method based on an FMCW radar system includes:
acquiring intermediate frequency signal data of at least one stroke to be recognized contained in handwritten Chinese characters based on an FMCW radar system;
preprocessing the intermediate frequency signal data of each stroke to be recognized to obtain a characteristic diagram set of each stroke to be recognized; the preprocessing comprises feature extraction and feature enhancement;
acquiring a trained Chinese character basic stroke recognition model; the Chinese character basic stroke recognition model is a convolutional neural network model which takes the feature map set as an input parameter and takes the basic stroke category as an output parameter;
and inputting the feature graph set of each stroke to be recognized into the Chinese character basic stroke recognition model, and acquiring the basic stroke category which is output by the Chinese character basic stroke recognition model and is matched with each stroke to be recognized.
Preferably, the preprocessing the intermediate frequency signal data of each stroke to be recognized to obtain a feature map set of each stroke to be recognized includes:
performing feature extraction on the intermediate frequency signal data of each stroke to be recognized through a first algorithm to obtain a corresponding feature matrix set, wherein the feature matrix set comprises a distance-time matrix and an angle-time matrix;
and performing characteristic enhancement on the characteristic matrix set of each stroke to be recognized through a second algorithm to obtain the corresponding characteristic diagram set.
Preferably, the performing feature extraction on the intermediate frequency signal data of each stroke to be recognized through a first algorithm to obtain a corresponding feature matrix set, where the feature matrix set includes a distance-time matrix and an angle-time matrix, and includes:
acquiring a frequency matrix of the intermediate frequency signal data through a third algorithm according to the format of the intermediate frequency signal data and the trend change characteristic of the stroke waving process, and acquiring a distance-time matrix of which the distance changes along with time according to the change of the frequency information;
and according to the format of the intermediate frequency signal data and the trend change characteristic of the stroke waving process, acquiring a phase data matrix of the distance matrix through a fourth algorithm, converting the phase data matrix into an angle data matrix, and then acquiring an angle-time matrix of which the angle changes along with time.
Preferably, the obtaining a phase data matrix of the distance matrix by a fourth algorithm according to the format of the intermediate frequency signal data and the trend change characteristic of the stroke waving process, and obtaining an angle-time matrix of which the angle changes with time after converting the phase data matrix into an angle data matrix includes:
performing fast Fourier transform on the intermediate frequency signal data to obtain phase information of each matrix element in the distance matrix;
carrying out covariance operation on the phase matrix to obtain a covariance matrix, and traversing a preset spatial spectrum function to obtain the angle matrix;
and carrying out time series combination on the angle matrix to obtain the angle-time matrix.
Preferably, the performing feature enhancement on the feature matrix set of each stroke to be recognized through a second algorithm to obtain the corresponding feature map set includes:
acquiring the position coordinates of the characteristic areas in the image according to the image represented by the two characteristic matrixes, and extracting the trend change characteristics of the stroke waving process in the image according to a preset matrix frame;
obtaining an optimal threshold value through a fifth algorithm, and binarizing the two feature matrixes according to the optimal threshold value to obtain a corresponding binarized image;
and opening the binary image corresponding to the two feature matrixes.
Preferably, before the obtaining of the trained basic stroke recognition model of the chinese character, the method includes:
acquiring first sample data and second sample data of preset basic strokes of the handwritten Chinese characters; the first sample data and the second sample data both contain a feature graph set corresponding to preset basic strokes in a preset quantity;
building a basic stroke recognition model of the Chinese character to be trained;
setting model training parameters and a loss function;
inputting the first sample data into the Chinese character basic stroke recognition model for iterative training, and calculating a loss value between a predicted category result and an actual category result output by the Chinese character basic stroke recognition model according to the loss function;
after each iteration is finished, inputting the second sample data into the Chinese character basic stroke recognition model for testing, and obtaining the prediction accuracy of the Chinese character basic stroke recognition model;
and judging the reduction degree of the loss value after each iteration so as to update the model training parameters or terminate the learning in advance.
Preferably, the Chinese character basic stroke recognition model comprises a convolutional neural network structure and a three-layer fully-connected neural network structure of a plurality of convolutional layers and a plurality of pooling layers.
Preferably, the basic stroke categories are:
wherein,
basic stroke categories output by the Chinese character basic stroke recognition model;
identifying a one-dimensional array of models for said basic strokes of said Chinese character
The feature map set of the middle input belongs to
Probability of class base strokes;
the basic stroke category corresponding to the maximum probability.
In a second aspect, a stroke recognition system based on an FMCW radar system includes:
the data acquisition module is used for acquiring intermediate frequency signal data of at least one stroke to be recognized contained in handwritten Chinese characters based on an FMCW radar system;
the processing module is used for preprocessing the intermediate frequency signal data of each stroke to be recognized and acquiring a characteristic diagram set of each stroke to be recognized; the preprocessing comprises feature extraction and feature enhancement;
the model acquisition module is used for acquiring a trained Chinese character basic stroke recognition model; the Chinese character basic stroke recognition model is a convolutional neural network model which takes the feature map set as an input parameter and takes the basic stroke category as an output parameter;
and the recognition module is used for inputting the feature map set of each stroke to be recognized into the Chinese character basic stroke recognition model and acquiring the basic stroke category which is output by the Chinese character basic stroke recognition model and matched with each stroke to be recognized.
Preferably, the stroke recognition system based on FMCW radar system further includes
The system comprises a sample acquisition module, a first sampling module and a second sampling module, wherein the sample acquisition module is used for acquiring first sample data and second sample data of preset basic strokes of handwritten Chinese characters; the first sample data and the second sample data both contain a feature graph set corresponding to preset basic strokes in a preset quantity;
the building module is used for building a basic stroke recognition model of the Chinese character to be trained, and setting model training parameters and a loss function;
the training module is used for inputting the first sample data into the Chinese character basic stroke recognition model for iterative training, and calculating a loss value between a predicted category result and an actual category result output by the Chinese character basic stroke recognition model according to the loss function;
the output module is used for judging whether the training of the Chinese character basic stroke recognition model is finished according to the loss value;
and the testing module is used for inputting the second sample data to the trained Chinese character basic stroke recognition model for testing to obtain the prediction accuracy of the Chinese character basic stroke recognition model.
By adopting the scheme, the invention has the following beneficial effects:
1) the invention obtains two characteristic graphs by preprocessing the intermediate frequency signal data of the strokes to be recognized contained in the handwritten Chinese characters, such as characteristic extraction, characteristic enhancement and the like, reduces the data volume for representing the gesture movement trend, improves the characteristic extraction efficiency, and saves the model training time;
2) the method identifies two characteristic graphs of the strokes to be identified through the Chinese character basic stroke network model to obtain the basic stroke category, and can accurately judge the basic stroke category. In addition, the invention is beneficial to the subsequent development of the non-contact man-machine interaction equipment.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a schematic flow chart of a stroke recognition method based on an FMCW radar system according to an embodiment of the present invention;
FIG. 2 is a flowchart illustrating the step S20 of the stroke recognition method based on FMCW radar system according to an embodiment of the present invention;
FIG. 3 is a flowchart illustrating a step S201 of a stroke recognition method based on an FMCW radar system according to an embodiment of the present invention;
FIG. 4 is a flowchart illustrating a step S202 of a stroke recognition method based on an FMCW radar system according to an embodiment of the present invention;
FIG. 5 is a schematic flow chart of a model training process based on a stroke recognition method of an FMCW radar system according to another embodiment of the present invention;
FIG. 6 is a schematic structural diagram of a stroke recognition system based on an FMCW radar system in accordance with an embodiment of the present invention;
FIG. 7 is a schematic structural diagram of a stroke recognition system based on FMCW radar system according to another embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
In an embodiment of the present invention, as shown in fig. 1, an embodiment of the present invention provides a stroke recognition method based on an FMCW radar system, the method including:
and step S10, acquiring intermediate frequency signal data of at least one stroke to be recognized contained in the preset handwritten Chinese character based on the FMCW radar system.
Before step S10, in this embodiment, hardware devices of an FMCW Radar (Frequency Modulated Continuous Wave Radar) system need to be configured, and a spatial range of a hand motion region needs to be set.
Taking the handwritten Chinese character 'eight' as an example, the FMCW radar system is enabled to emit electromagnetic waves in a direction facing the hand, data acquisition is carried out on the waving process of two strokes to be recognized of the handwritten Chinese character 'eight', and intermediate frequency signal data of the strokes to be recognized are obtained.
Step S20, preprocessing the intermediate frequency signal data of each stroke to be recognized, and acquiring a feature map set of each stroke to be recognized.
In the embodiment, the preprocessing of the intermediate frequency signal data of the basic strokes of the handwritten Chinese characters comprises feature extraction and feature enhancement, wherein the feature extraction comprises distance estimation for generating a distance-time matrix (RTM) and angle estimation for generating an angle-time matrix (ATM); the characteristic enhancement comprises characteristic region framing, binarization and opening operation. Preferably, as shown in fig. 2, step S20 specifically includes the following steps:
step S201, performing feature extraction on the intermediate frequency signal data of each stroke to be recognized through a first algorithm to obtain a corresponding feature matrix set, wherein the feature matrix set comprises a distance-time matrix and an angle-time matrix.
The first algorithm is a Distance Feature Sequence Extraction (DFSE) and Angle Feature Sequence Extraction (AFSE) combined Feature Extraction algorithm (DA-FSE algorithm) and is used for representing the trend change Feature of the stroke waving process through a Distance-time matrix (RTM) and an angle-time matrix (ATM).
Preferably, as shown in fig. 3, step S201 specifically includes the following steps:
step S2011, estimating a distance, namely, obtaining a frequency matrix of the intermediate frequency signal data through a third algorithm according to the format of the intermediate frequency signal data and the trend change characteristic of the stroke waving process, and obtaining a distance-time matrix of which the distance changes along with time after converting the frequency matrix into a distance matrix.
Wherein the third algorithm is a short time fourier transform (STFT algorithm); the frequency matrix comprises frequency information of each frame of intermediate frequency signal data; the distance matrix includes distance information obtained by converting frequency information, and the distance-time matrix is a two-dimensional matrix combined by the distance matrix according to a time dimension. Preferably, step S2011 specifically includes the following steps:
firstly, acquiring a window function and parameters of the window function; optionally, according to the characteristics and parameter configuration of the intermediate frequency signal data, a window function is selected through multiple comparison experiments, and the size and redundancy of the window function are set.
Then, inputting the window function and the frequency matrix into a matrix conversion model based on a third algorithm to obtain the distance matrix output by the matrix conversion model; optionally, the matrix conversion model based on the third algorithm is as follows:
in the formula (1), the first and second groups,
in order to obtain a distance matrix after the STFT operation,
to output length
The frequency matrix of (a) is determined,
is the length of the window
The window function of (2). As can be seen from formula (1), will
Performing STFT operation Q times to obtain
。
And finally, performing time sequence combination on the distance matrix to obtain a distance-time matrix (RTM).
Step S2012, estimating an angle, obtaining a phase matrix of the intermediate frequency signal data through a fourth algorithm according to the format of the intermediate frequency signal data and the trend change characteristic of the stroke waving process, and obtaining an angle-time matrix of which the angle changes with time after converting the phase matrix into an angle matrix.
Wherein, the fourth algorithm is an algorithm combining Fourier transform and a spatial spectrum function (Range-Capon algorithm); the phase matrix comprises phase information of each frame of intermediate frequency signal data; the angle matrix comprises angle information obtained by converting the phase information; the angle-time matrix refers to a two-dimensional matrix in which angle matrices are combined according to a time dimension. Preferably, step S2012 includes the following steps:
first, a Fast Fourier Transform (FFT) is performed on the intermediate frequency signal data to obtain a phase matrix containing phase information, and it can be understood that each matrix element in the phase matrix represents the phase information of each frame of the intermediate frequency signal data.
And then, carrying out covariance operation on the phase matrix to obtain a covariance matrix, and carrying out one-time traversal on a preset spatial spectrum function to obtain an angle matrix. Wherein the spatial spectrum function is:
in the formula (2), the first and second groups,
in order to be a function of the spatial spectrum,
is the inverse of the covariance matrix,
as a guide vector
The transpose of (a) is performed,
is the angle of arrival.
And finally, carrying out time sequence combination on the angle matrix to obtain an angle-time matrix (ATM).
In this embodiment, a distance-time matrix (RTM) in which the distance changes with time and an angle-time matrix (ATM) in which the angle changes with time are obtained by using a first algorithm (DA-FSE algorithm), and then a trend change characteristic of the stroke waving process is represented by the distance-time matrix (RTM) and the angle-time matrix (ATM), so that the data size can be reduced, and the time for model training can be saved.
Step S202, performing feature enhancement on the feature matrix set of each stroke to be recognized through a second algorithm to obtain the corresponding feature map set.
The second algorithm is a feature enhancement algorithm (FA-FBO algorithm) combining the framing, binarization and opening operations of the specific region, and is used for performing feature enhancement on the image represented by the feature matrix through a series of operations such as framing, binarization and opening operations of the specific region. Preferably, as shown in fig. 4, step S202 specifically includes the following steps:
step S2021, framing a characteristic area, acquiring position coordinates of the characteristic area in the image according to the image represented by the two characteristic matrixes, and extracting trend change characteristics of the stroke waving process in the image according to a preset matrix frame. Wherein, the two characteristic matrixes are a distance-time matrix and an angle-time matrix; the preset matrix frame is set to a fixed size.
And step S2022, carrying out binarization, obtaining an optimal threshold value through a fifth algorithm, and carrying out binarization on the two feature matrixes according to the optimal threshold value to obtain a corresponding binarized image. The fifth algorithm is a maximum inter-class variance method (Otsu algorithm).
That is, the Otsu algorithm is used to obtain the optimal threshold, and when the gray value of the image pixel is greater than or equal to the optimal threshold, the gray value is set to a first value, for example, the first value is 255; on the contrary, when the gray value of the image pixel is smaller than the optimal threshold value, the image pixel is placed at a second numerical value, for example, the second numerical value is 0, and then the two feature matrixes are respectively binarized to obtain a binarized image. Preferably, the binarization of step S2022 is specifically expressed as:
in the formula (3), the first and second groups,
is a gray-scale value of a pixel of the image,
is the optimal threshold.
And step S2023, performing opening operation on the binary images corresponding to the two feature matrixes to smooth the contour of the trend change feature in the binary images corresponding to the two feature matrixes, and breaking narrow gaps and eliminating thin protrusions. Preferably, the opening operation of step S2023 is specifically expressed as:
in the formula (4), the first and second groups,Cis a structural element of the compound and is a structural element,Bare a set. As can be seen from the formula (4), the structural elements are usedCTo the collectionBBy open operation is meant the use of structural elementsCTo the collectionBEtching and reusing structural elementsCThe result is expanded.
In this embodiment, a second algorithm (FA-FBO algorithm) is used to perform feature enhancement on the RTM whose distance changes with time and the ATM whose angle changes with time, so that error interference can be reduced and features can be highlighted.
Step S30, acquiring a trained Chinese character basic stroke recognition model; the Chinese character basic stroke recognition model is a convolutional neural network model which takes the feature map set as an input parameter and takes the basic stroke category as an output parameter.
In this embodiment, the basic stroke recognition model of the chinese character in step S30 is based on the convolutional neural network model obtained by the training of the following steps S501 to S505.
And step S40, inputting the feature map set of each stroke to be recognized into the Chinese character basic stroke recognition model, and acquiring the basic stroke category output by the Chinese character basic stroke recognition model and matched with each stroke to be recognized.
Inputting the characteristic diagram set of each stroke to be recognized into the Chinese character basic stroke recognition model to obtain a one-dimensional array output by the Chinese character basic stroke recognition model, wherein the output array is of the length of
One-dimensional array of
The one-dimensional array
To middle
An elementThe feature map set representing the input belongs to
Probability of class base stroke
Probability of will
And the corresponding basic stroke category at the maximum is taken as the recognition result of the feature graph set, so that the judgment of the basic strokes of the handwritten Chinese characters is completed. Preferably, the basic stroke category may be expressed as:
in the formulas (5) and (6),
basic stroke categories output by a basic stroke recognition model of the Chinese character,
one-dimensional array for Chinese character basic stroke recognition model
The feature map set of the middle input belongs to
The probability of the basic stroke of the class,
the basic stroke category corresponding to the maximum probability. From the formulas (5) and (6), it can be seen thatOne-dimensional array
Has a length of
And is and
one-dimensional array of numbers of basic stroke categories
The total probability of all elements in (a) added is 1.
In summary, in the embodiment, feature extraction, feature enhancement and other preprocessing are performed on the intermediate-frequency signal data of the strokes to be recognized included in the handwritten Chinese characters to obtain two feature maps, so that the data volume for representing the gesture movement trend is reduced, the feature extraction efficiency is improved, and meanwhile, the model training time is saved. And then, identifying two characteristic graphs of the strokes to be identified through the Chinese character basic stroke network model to obtain basic stroke categories, and accurately judging the basic stroke categories. In addition, the invention is beneficial to the subsequent development of the non-contact man-machine interaction equipment.
As still another embodiment of the present invention, as shown in fig. 5, the stroke recognition method based on the FMCW radar system further includes a model training process, which specifically includes the following steps:
step S501, acquiring first sample data and second sample data of preset basic strokes of handwritten Chinese characters; and the first sample data and the second sample data both contain a feature diagram set corresponding to preset basic strokes in a preset quantity.
In this embodiment, the number of the first sample data and the number of the second sample data may be set according to requirements. Taking the hand-written Chinese character 'Yong' as an example, the FMCW radar system is used for collecting the intermediate frequency signal data of the eight basic strokes of horizontal stroke, dot stroke, lifting stroke, left-falling stroke, bending stroke, right-falling stroke, vertical stroke and hook stroke of the hand-written Chinese character 'Yong'. Acquiring 100 groups of sample data of the intermediate frequency signal data of each basic stroke, acquiring 800 groups of sample data of basic strokes of handwritten Chinese characters in total, and preprocessing each sample data, wherein the preprocessing comprises the following steps: firstly, performing feature extraction on each sample data through a first algorithm to obtain a feature matrix set of each data sample, then performing feature enhancement on two feature matrices contained in the feature matrix set through a second algorithm to obtain a corresponding feature map set, and finally performing normalization processing on the two feature maps contained in the feature map set. Alternatively, the feature map may be normalized to a size of 125 × 189.
Further, grouping feature graph sets of basic strokes of the handwritten Chinese characters obtained after preprocessing, and optionally, taking 500 groups of feature graph sets as first sample data of preset basic strokes of the handwritten Chinese characters, and combining 300 groups of feature graph sets as second sample data of the basic strokes of the handwritten Chinese characters. The first sample data is used for training a basic stroke model of the Chinese character, and the second sample data is used for testing the prediction accuracy of the basic stroke model of the Chinese character.
It should be noted that, in other embodiments, the sample data may be grouped first, and then the grouped sample data may be preprocessed.
And step S502, building a basic stroke recognition model of the Chinese character to be trained, and setting model training parameters and a loss function.
In the embodiment, the basic stroke recognition model of Chinese characters comprises a Convolutional Neural Network (CNN) structure with a plurality of convolutional layers and a plurality of pooling layers and a three-layer fully-connected neural network structure. Optionally, the Convolutional Neural Network (CNN) structure comprises three convolutional layers and three pooling layers; the sizes of convolution kernels of three convolution layers of the CNN are all 3 multiplied by 3, the step length is 1, the number of the convolution kernels of each layer is respectively 4, 4 and 8, and all-zero filling operation is adopted during convolution; the pooling layer kernel size is 2 × 2, step size is 2. In the three-layer fully-connected neural network structure, the number of nodes in the hidden layer is 100, and the number of nodes in the output layer is 100
，
Indicating the number of basic stroke categories.
The model training parameters can be set according to requirements, and include but are not limited to batch size, learning rate, iteration times, direction propagation algorithm and the like. Optionally, the batch size may be set to 10, the learning rate may be set to 0.00001, the number of iterations may be set to 100, and the directional propagation algorithm may be set to the random gradient descent algorithm.
The loss function can be set according to requirements, and the loss function can adopt a cross entropy algorithm.
Step S503, inputting the first sample data to the Chinese character basic stroke recognition model for iterative training, and calculating a loss value between a predicted category result and an actual category result output by the Chinese character basic stroke recognition model according to the loss function.
And step S504, judging whether the training of the Chinese character basic stroke recognition model is finished according to the loss value.
In the model training process, inputting first sample data of preset basic strokes of handwritten Chinese characters into a Chinese character basic stroke recognition model to be trained for iterative training, calculating a loss value between a prediction category result and an actual category result output by the model through a loss function, judging whether the loss value is smaller than a preset loss threshold value, if the loss value is smaller than the preset loss threshold value, determining that the model training is finished, storing model training parameters, and stopping learning in advance; otherwise, if the loss value is larger than or equal to the preset loss threshold value, updating the model training parameters, re-training the Chinese character basic stroke recognition model based on the updated model training parameters until the loss value is smaller than the preset loss threshold value, determining that the model training is finished, and storing the updated model training parameters.
And step S505, inputting the second sample data to the trained Chinese character basic stroke recognition model for testing, and obtaining the prediction accuracy of the Chinese character basic stroke recognition model.
It should be noted that, as shown in fig. 5, the steps S501 to S505 are performed before the step S10, and in other embodiments, the steps S501 to S505 may be performed before the step S20 or the step S30, so that the steps S501 to S505 are performed before any one of the steps S30, S20 and S10.
In the model testing process, a second data sample of the preset basic strokes of the handwritten Chinese characters is used as input parameters of a trained Chinese character basic stroke recognition model, the input parameters are extracted to characteristic vectors through a Convolutional Neural Network (CNN) structure in the Chinese character basic stroke recognition model, output results, namely the classes of the basic strokes of the handwritten Chinese characters are obtained through a three-layer fully-connected neural network structure, and then the prediction accuracy of the model is obtained. The method can be obtained through experiments, the model is tested by adopting 300 second sample data, and the prediction accuracy of the model can reach 96%. It can be understood that the stroke identification method based on the FMCW radar system of the embodiment can accurately judge the basic stroke category.
In addition, an embodiment of the present invention further provides a stroke recognition system based on an FMCW radar system, as shown in fig. 7, the system includes a data acquisition module 110, a preprocessing module 120, a model acquisition module 130, and a recognition module 140, and detailed descriptions of the functional modules are as follows:
and the data acquisition module 110 is used for acquiring intermediate frequency signal data of at least one stroke to be recognized contained in the handwritten Chinese character based on the FMCW radar system.
The preprocessing module 120 is configured to preprocess the intermediate frequency signal data of each stroke to be recognized, and obtain a feature map set of each stroke to be recognized; the preprocessing includes feature extraction and feature enhancement.
A model obtaining module 130, configured to obtain a trained basic stroke recognition model of a Chinese character; the Chinese character basic stroke recognition model is a convolutional neural network model which takes the feature map set as an input parameter and takes the basic stroke category as an output parameter.
The recognition module 140 is configured to input the feature map set of each stroke to be recognized into the basic Chinese character stroke recognition model, and obtain a basic stroke category output by the basic Chinese character stroke recognition model and matched with each stroke to be recognized.
Further, as shown in fig. 6, the preprocessing module 120 includes a feature extraction sub-module 121 and a feature enhancement sub-module 122, and the detailed description of each functional sub-module is as follows:
the feature extraction submodule 121 is configured to perform feature extraction on the intermediate frequency signal data of each stroke to be recognized through a first algorithm to obtain a corresponding feature matrix set, where the feature matrix set includes a distance-time matrix and an angle-time matrix.
And the feature enhancer module 122 is configured to perform feature enhancement on the feature matrix set of each stroke to be recognized through a second algorithm to obtain the corresponding feature map set.
Further, as shown in fig. 6, the feature extraction sub-module 121 includes the following units, and the detailed description of each functional unit is as follows:
a distance estimation unit 1211, configured to obtain a frequency matrix of the intermediate frequency signal data through a third algorithm according to a format of the intermediate frequency signal data and a trend change characteristic of a stroke waving process, and obtain a distance-time matrix of a distance change with time according to a change of the frequency information.
The angle estimation unit 1212 is configured to obtain a phase data matrix of the distance matrix through a fourth algorithm according to the format of the intermediate frequency signal data and the trend change characteristic of the stroke waving process, and obtain an angle-time matrix in which an angle changes with time after converting the phase data matrix into an angle data matrix.
Further, the angle estimation unit includes the following units, and the detailed description of each functional unit is as follows:
and the transformation unit is used for performing fast Fourier transformation on the intermediate frequency signal data to obtain phase information of each matrix element in the distance matrix.
And the traversal unit is used for performing covariance operation on the phase matrix to obtain a covariance matrix and traversing a preset spatial spectrum function to obtain the angle matrix.
And the combining unit is used for carrying out time series combination on the angle matrix to obtain the angle-time matrix.
Further, as shown in fig. 6, the feature enhancement sub-module 122 includes the following units, and the detailed description of each functional unit is as follows:
and the feature framing unit 1221 is configured to acquire a feature area position coordinate in the image according to the image represented by the two feature matrices, and extract a trend change feature of the stroke waving process in the image according to a preset matrix frame.
A binarization unit 1222, configured to obtain an optimal threshold through a fifth algorithm, binarize the two feature matrices according to the optimal threshold, and obtain a corresponding binarized image.
And an opening operation unit 1223, configured to perform an opening operation on the binarized image corresponding to the two feature matrices.
Further, as shown in fig. 7, the system further includes a sample obtaining module 151, a building module 152, a training module 153, a testing module 154, and an output module 155, and the detailed description of each functional module is as follows:
the sample acquisition module 151 is configured to acquire first sample data and second sample data of preset basic strokes of a handwritten Chinese character; and the first sample data and the second sample data both contain a feature diagram set corresponding to preset basic strokes in a preset quantity.
And the building module 152 is used for building a basic stroke recognition model of the Chinese character to be trained, and setting model training parameters and a loss function.
The training module 153 is configured to input the first sample data to the chinese character basic stroke recognition model for iterative training, and calculate a loss value between a predicted category result and an actual category result output by the chinese character basic stroke recognition model according to the loss function.
The output module 154 is configured to determine whether the training of the basic stroke recognition model of the Chinese character is completed according to the loss value;
and the testing module 155 is configured to input the second sample data to the trained basic Chinese character stroke recognition model for testing, so as to obtain a prediction accuracy of the basic Chinese character stroke recognition model.
Further, the basic stroke recognition model of the Chinese character in the model obtaining module 130 includes a convolutional neural network structure having a plurality of convolutional layers and a plurality of pooling layers and a three-layer fully-connected neural network structure.
Further, the basic stroke categories in the recognition module 140 are:
wherein,
basic stroke categories output by the Chinese character basic stroke recognition model;
identifying a one-dimensional array of models for said basic strokes of said Chinese character
The feature map set of the middle input belongs to
Probability of class base strokes;
the basic stroke category corresponding to the maximum probability.
The stroke recognition system based on the FMCW radar system provided by the embodiment of the invention can obtain two characteristic graphs by preprocessing the intermediate frequency signal data of the strokes to be recognized contained in the handwritten Chinese characters, such as characteristic extraction, characteristic enhancement and the like, thereby reducing the data volume for representing the gesture movement trend, improving the characteristic extraction efficiency and saving the model training time; and the two characteristic graphs of the strokes to be recognized can be recognized through the Chinese character basic stroke network model to obtain basic stroke categories, and the basic stroke categories can be accurately distinguished. In addition, the invention is beneficial to the subsequent development of the non-contact man-machine interaction equipment.
Those of ordinary skill in the art will understand that: the discussion of any embodiment above is meant to be exemplary only, and is not intended to intimate that the scope of the disclosure, including the claims, is limited to these examples; within the idea of the invention, also technical features in the above embodiments or in different embodiments may be combined and there are many other variations of the different aspects of the invention as described above, which are not provided in detail for the sake of brevity. Therefore, any omissions, modifications, substitutions, improvements and the like that may be made without departing from the spirit and principles of the invention are intended to be included within the scope of the invention.
Claims (7)
1. A stroke identification method based on an FMCW radar system is characterized by comprising the following steps:
acquiring intermediate frequency signal data of at least one stroke to be recognized contained in handwritten Chinese characters based on an FMCW radar system;
performing feature extraction on the intermediate frequency signal data of each stroke to be recognized through a first algorithm to obtain a corresponding feature matrix set; the feature matrix set comprises a distance-time matrix and an angle-time matrix; the first algorithm is a feature extraction algorithm combining distance feature sequence extraction and angle feature sequence extraction; the method comprises the following steps: distance estimation, namely acquiring a frequency matrix of the intermediate frequency signal data through a third algorithm according to the format of the intermediate frequency signal data and the trend change characteristic of the stroke waving process, converting the frequency matrix into a distance matrix, and then acquiring a distance-time matrix of which the distance changes along with time, wherein the distance estimation specifically comprises the following steps:
step one, acquiring a window function and parameters of the window function;
inputting the window function and the frequency matrix into a matrix conversion model based on a third algorithm to obtain the distance matrix output by the matrix conversion model; the matrix conversion model based on the third algorithm is as follows:
wherein,
in order to obtain a distance matrix after the STFT operation,
to output length
The frequency matrix of (a) is determined,
is the length of the window
A window function of;
thirdly, performing time sequence combination on the distance matrix to obtain a distance-time matrix;
angle estimation, namely acquiring a phase matrix of the intermediate frequency signal data through a fourth algorithm according to the format of the intermediate frequency signal data and the trend change characteristic of the stroke waving process, and converting the phase matrix into an angle matrix to obtain an angle-time matrix of which the angle changes along with time;
performing feature enhancement on the feature matrix set of each stroke to be recognized through a second algorithm to obtain a corresponding feature map set; the second algorithm is a feature enhancement algorithm combining framing, binaryzation and opening operation of a specific area; the method comprises the following steps: acquiring the position coordinates of the characteristic areas in the image according to the image represented by the two characteristic matrixes, and extracting the trend change characteristics of the stroke waving process in the image according to a preset matrix frame;
obtaining an optimal threshold value through a fifth algorithm, and binarizing the two feature matrixes according to the optimal threshold value to obtain a corresponding binarized image;
opening the binary images corresponding to the two feature matrixes;
acquiring a trained Chinese character basic stroke recognition model; the Chinese character basic stroke recognition model is a convolutional neural network model which takes the feature map set as an input parameter and takes the basic stroke category as an output parameter;
and inputting the feature graph set of each stroke to be recognized into the Chinese character basic stroke recognition model, and acquiring the basic stroke category which is output by the Chinese character basic stroke recognition model and is matched with each stroke to be recognized.
2. The FMCW radar system-based stroke recognition method as claimed in claim 1, wherein the obtaining of the phase matrix of the if signal data through a fourth algorithm according to the format of the if signal data and the trend change characteristic of the stroke waving process and the angle-time matrix of the angle change with time after converting the phase matrix into the angle matrix comprises: performing fast Fourier transform on the intermediate frequency signal data to obtain a phase matrix containing phase information;
carrying out covariance operation on the phase matrix to obtain a covariance matrix, and traversing a preset spatial spectrum function to obtain the angle matrix;
and carrying out time series combination on the angle matrix to obtain the angle-time matrix.
3. The FMCW radar system-based stroke recognition method of claim 1, further comprising:
acquiring first sample data and second sample data of preset basic strokes of the handwritten Chinese characters; the first sample data and the second sample data both contain a feature graph set corresponding to preset basic strokes in a preset quantity;
building a basic stroke recognition model of the Chinese character to be trained, and setting model training parameters and a loss function;
inputting the first sample data into the Chinese character basic stroke recognition model for iterative training, and calculating a loss value between a predicted category result and an actual category result output by the Chinese character basic stroke recognition model according to the loss function;
judging whether the training of the Chinese character basic stroke recognition model is finished according to the loss value;
and inputting the second sample data to the trained Chinese character basic stroke recognition model for testing to obtain the prediction accuracy of the Chinese character basic stroke recognition model.
4. The FMCW radar system-based stroke recognition method of claim 3, wherein the Chinese character basic stroke recognition model includes a convolutional neural network structure having a plurality of convolutional layers and a plurality of pooling layers and a three-layer fully-connected neural network structure.
5. The FMCW radar system-based stroke recognition method of claim 1, wherein the basic stroke categories are:
wherein,
basic stroke categories output by the Chinese character basic stroke recognition model;
identifying a one-dimensional array of models for said basic strokes of said Chinese character
The feature map set of the middle input belongs to
Probability of class base strokes;
the basic stroke category corresponding to the maximum probability.
6. A stroke recognition system based on an FMCW radar system, the system comprising:
the data acquisition module is used for acquiring intermediate frequency signal data of at least one stroke to be recognized contained in handwritten Chinese characters based on an FMCW radar system;
the characteristic extraction module is used for extracting characteristics of the intermediate frequency signal data of each stroke to be recognized through a first algorithm to obtain a corresponding characteristic matrix set; the feature matrix set comprises a distance-time matrix and an angle-time matrix; the first algorithm is a feature extraction algorithm combining distance feature sequence extraction and angle feature sequence extraction; the feature extraction module comprises: the distance estimation unit is used for acquiring a frequency matrix of the intermediate frequency signal data through a third algorithm according to the format of the intermediate frequency signal data and the trend change characteristic of the stroke waving process, converting the frequency matrix into a distance matrix, and then acquiring a distance-time matrix of which the distance changes along with time, wherein the work flow of the distance estimation unit is as follows:
step one, acquiring a window function and parameters of the window function;
inputting the window function and the frequency matrix into a matrix conversion model based on a third algorithm to obtain the distance matrix output by the matrix conversion model; the matrix conversion model based on the third algorithm is as follows:
wherein,
in order to obtain a distance matrix after the STFT operation,
to output length
The frequency matrix of (a) is determined,
is the length of the window
Window function of
Thirdly, performing time sequence combination on the distance matrix to obtain a distance-time matrix;
the angle estimation unit is used for acquiring a phase matrix of the intermediate frequency signal data through a fourth algorithm according to the format of the intermediate frequency signal data and the trend change characteristic of the stroke waving process, and acquiring an angle-time matrix of which the angle changes along with time after converting the phase matrix into an angle matrix;
the characteristic enhancement module is used for carrying out characteristic enhancement on the characteristic matrix set of each stroke to be recognized through a second algorithm to obtain a corresponding characteristic diagram set; the second algorithm is a feature enhancement algorithm combining framing, binaryzation and opening operation of a specific area; the feature enhancement module includes: the characteristic framing unit is used for acquiring the position coordinates of the characteristic area in the image according to the image represented by the two characteristic matrixes and extracting the trend change characteristics of the stroke waving process in the image according to a preset matrix frame;
the binarization unit is used for obtaining an optimal threshold value through a fifth algorithm, binarizing the two feature matrixes according to the optimal threshold value, and obtaining a corresponding binarized image;
the opening operation unit is used for opening the binary images corresponding to the two feature matrixes;
the model acquisition module is used for acquiring a trained Chinese character basic stroke recognition model; the Chinese character basic stroke recognition model is a convolutional neural network model which takes the feature map set as an input parameter and takes the basic stroke category as an output parameter;
and the recognition module is used for inputting the feature map set of each stroke to be recognized into the Chinese character basic stroke recognition model and acquiring the basic stroke category which is output by the Chinese character basic stroke recognition model and matched with each stroke to be recognized.
7. The FMCW radar system-based stroke recognition system of claim 6, further comprising:
the system comprises a sample acquisition module, a first sampling module and a second sampling module, wherein the sample acquisition module is used for acquiring first sample data and second sample data of preset basic strokes of handwritten Chinese characters; the first sample data and the second sample data both contain a feature graph set corresponding to preset basic strokes in a preset quantity;
the building module is used for building a basic stroke recognition model of the Chinese character to be trained, and setting model training parameters and a loss function;
the training module is used for inputting the first sample data into the Chinese character basic stroke recognition model for iterative training, and calculating a loss value between a predicted category result and an actual category result output by the Chinese character basic stroke recognition model according to the loss function;
the output module is used for judging whether the training of the Chinese character basic stroke recognition model is finished according to the loss value;
and the testing module is used for inputting the second sample data to the trained Chinese character basic stroke recognition model for testing to obtain the prediction accuracy of the Chinese character basic stroke recognition model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011420741.7A CN112198966B (en) | 2020-12-08 | 2020-12-08 | Stroke identification method and system based on FMCW radar system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011420741.7A CN112198966B (en) | 2020-12-08 | 2020-12-08 | Stroke identification method and system based on FMCW radar system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112198966A CN112198966A (en) | 2021-01-08 |
| CN112198966B true CN112198966B (en) | 2021-03-16 |
Family
ID=74033862
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011420741.7A Active CN112198966B (en) | 2020-12-08 | 2020-12-08 | Stroke identification method and system based on FMCW radar system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112198966B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113361367B (en) * | 2021-06-01 | 2022-04-26 | 中南大学 | Underground target electromagnetic inversion method and system based on deep learning |
| CN114092938B (en) * | 2022-01-19 | 2022-04-19 | 腾讯科技（深圳）有限公司 | Image recognition processing method and device, electronic equipment and storage medium |
| CN115497107B (en) * | 2022-09-30 | 2023-04-18 | 江西师范大学 | Zero-sample Chinese character recognition method based on stroke and radical decomposition |
| CN117608399B (en) * | 2023-11-23 | 2024-06-14 | 首都医科大学附属北京天坛医院 | Track fitting method and device based on Chinese character strokes |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108509910A (en) * | 2018-04-02 | 2018-09-07 | 重庆邮电大学 | Deep learning gesture identification method based on fmcw radar signal |
| CN109271838A (en) * | 2018-07-19 | 2019-01-25 | 重庆邮电大学 | A kind of three parameter attributes fusion gesture identification method based on fmcw radar |
| CN109829509A (en) * | 2019-02-26 | 2019-05-31 | 重庆邮电大学 | Radar gesture identification method based on fused neural network |
| CN110348288A (en) * | 2019-05-27 | 2019-10-18 | 哈尔滨工业大学（威海） | A kind of gesture identification method based on 77GHz MMW RADAR SIGNAL USING |
| CN110765974A (en) * | 2019-10-31 | 2020-02-07 | 复旦大学 | Micro-motion gesture recognition method based on millimeter wave radar and convolutional neural network |
| CN111027458A (en) * | 2019-08-28 | 2020-04-17 | 深圳大学 | Gesture recognition method and device based on radar three-dimensional track characteristics and storage medium |
| CN111399642A (en) * | 2020-03-09 | 2020-07-10 | 深圳大学 | Gesture recognition method and device, mobile terminal and storage medium |
| CN111427031A (en) * | 2020-04-09 | 2020-07-17 | 浙江大学 | Identity and gesture recognition method based on radar signals |
| CN111461037A (en) * | 2020-04-07 | 2020-07-28 | 电子科技大学 | End-to-end gesture recognition method based on FMCW radar |
| WO2020176105A1 (en) * | 2019-02-28 | 2020-09-03 | Google Llc | Smart-device-based radar system detecting user gestures in the presence of saturation |
| CN111680539A (en) * | 2020-04-14 | 2020-09-18 | 北京清雷科技有限公司 | Dynamic gesture radar recognition method and device |
| WO2020187397A1 (en) * | 2019-03-19 | 2020-09-24 | HELLA GmbH & Co. KGaA | A method for a detection and classification of gestures using a radar system |
| CN111722706A (en) * | 2019-03-21 | 2020-09-29 | 英飞凌科技股份有限公司 | Method and system for recognizing aerial written characters based on radar network |
| CN111813222A (en) * | 2020-07-09 | 2020-10-23 | 电子科技大学 | Terahertz radar-based fine dynamic gesture recognition method |
| CN111814743A (en) * | 2020-07-30 | 2020-10-23 | 深圳壹账通智能科技有限公司 | Handwriting recognition method and device and computer readable storage medium |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109085548B (en) * | 2018-09-26 | 2021-03-26 | 湖南时变通讯科技有限公司 | Surface penetrating radar hyperbolic target detection method and device |
| CN110579746A (en) * | 2019-10-11 | 2019-12-17 | 湖南时变通讯科技有限公司 | Echo signal processing method, device, equipment and storage medium |
| CN111157988B (en) * | 2020-02-27 | 2023-04-07 | 中南大学 | Gesture radar signal processing method based on RDTM and ATM fusion |
-
2020
- 2020-12-08 CN CN202011420741.7A patent/CN112198966B/en active Active
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108509910A (en) * | 2018-04-02 | 2018-09-07 | 重庆邮电大学 | Deep learning gesture identification method based on fmcw radar signal |
| CN109271838A (en) * | 2018-07-19 | 2019-01-25 | 重庆邮电大学 | A kind of three parameter attributes fusion gesture identification method based on fmcw radar |
| CN109829509A (en) * | 2019-02-26 | 2019-05-31 | 重庆邮电大学 | Radar gesture identification method based on fused neural network |
| WO2020176105A1 (en) * | 2019-02-28 | 2020-09-03 | Google Llc | Smart-device-based radar system detecting user gestures in the presence of saturation |
| WO2020187397A1 (en) * | 2019-03-19 | 2020-09-24 | HELLA GmbH & Co. KGaA | A method for a detection and classification of gestures using a radar system |
| CN111722706A (en) * | 2019-03-21 | 2020-09-29 | 英飞凌科技股份有限公司 | Method and system for recognizing aerial written characters based on radar network |
| CN110348288A (en) * | 2019-05-27 | 2019-10-18 | 哈尔滨工业大学（威海） | A kind of gesture identification method based on 77GHz MMW RADAR SIGNAL USING |
| CN111027458A (en) * | 2019-08-28 | 2020-04-17 | 深圳大学 | Gesture recognition method and device based on radar three-dimensional track characteristics and storage medium |
| CN110765974A (en) * | 2019-10-31 | 2020-02-07 | 复旦大学 | Micro-motion gesture recognition method based on millimeter wave radar and convolutional neural network |
| CN111399642A (en) * | 2020-03-09 | 2020-07-10 | 深圳大学 | Gesture recognition method and device, mobile terminal and storage medium |
| CN111461037A (en) * | 2020-04-07 | 2020-07-28 | 电子科技大学 | End-to-end gesture recognition method based on FMCW radar |
| CN111427031A (en) * | 2020-04-09 | 2020-07-17 | 浙江大学 | Identity and gesture recognition method based on radar signals |
| CN111680539A (en) * | 2020-04-14 | 2020-09-18 | 北京清雷科技有限公司 | Dynamic gesture radar recognition method and device |
| CN111813222A (en) * | 2020-07-09 | 2020-10-23 | 电子科技大学 | Terahertz radar-based fine dynamic gesture recognition method |
| CN111814743A (en) * | 2020-07-30 | 2020-10-23 | 深圳壹账通智能科技有限公司 | Handwriting recognition method and device and computer readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112198966A (en) | 2021-01-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112198966B (en) | Stroke identification method and system based on FMCW radar system | |
| CN109359538B (en) | Training method of convolutional neural network, gesture recognition method, device and equipment | |
| US8977059B2 (en) | Integrating feature extraction via local sequential embedding for automatic handwriting recognition | |
| Lahiani et al. | Hand gesture recognition method based on HOG-LBP features for mobile devices | |
| CN108027876B (en) | System for recognizing multiple object inputs, method and product thereof | |
| KR101354663B1 (en) | A method and apparatus for recognition of handwritten symbols | |
| Malassiotis et al. | A gesture recognition system using 3D data | |
| Nair et al. | Hand gesture recognition system for physically challenged people using IOT | |
| CN109903314A (en) | A kind of method, the method for model training and the relevant apparatus of image-region positioning | |
| Elpeltagy et al. | Multi‐modality‐based Arabic sign language recognition | |
| CN110472506B (en) | Gesture recognition method based on support vector machine and neural network optimization | |
| CN108198159A (en) | A kind of image processing method, mobile terminal and computer readable storage medium | |
| Hsieh et al. | Air-writing recognition based on deep convolutional neural networks | |
| Wu et al. | Depth-based hand gesture recognition | |
| CN112749646A (en) | Interactive point-reading system based on gesture recognition | |
| CN111242109A (en) | Method and device for manually fetching words | |
| Mali et al. | Indian sign language recognition using SVM classifier | |
| Choudhury et al. | A CNN-LSTM based ensemble framework for in-air handwritten Assamese character recognition | |
| WO2018178228A1 (en) | System and method for recognition of objects from ink elements | |
| Hamdan et al. | Deep learning based handwriting recognition with adversarial feature deformation and regularization | |
| CN111160308B (en) | Gesture recognition method, device, equipment and readable storage medium | |
| Lahiani et al. | Hand pose estimation system based on Viola-Jones algorithm for android devices | |
| Arafat et al. | Two stream deep neural network for sequence-based Urdu ligature recognition | |
| Lahiani et al. | Real Time Static Hand Gesture Recognition System for Mobile Devices. | |
| Singh et al. | A Temporal Convolutional Network for modeling raw 3D sequences and air-writing recognition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |