CN113283380A - Children motion attitude automatic identification method based on 3D convolution long-term and short-term memory network - Google Patents

Children motion attitude automatic identification method based on 3D convolution long-term and short-term memory network Download PDF

Info

Publication number
CN113283380A
CN113283380A CN202110652217.0A CN202110652217A CN113283380A CN 113283380 A CN113283380 A CN 113283380A CN 202110652217 A CN202110652217 A CN 202110652217A CN 113283380 A CN113283380 A CN 113283380A
Authority
CN
China
Prior art keywords
image
videos
motion
term memory
memory network
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
Application number
CN202110652217.0A
Other languages
Chinese (zh)
Inventor
庄悦阳
李睿
张洁欣
江涛
冯安豫
陈经纬
柳维林
陈宪泽
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Individual
Original Assignee
Individual
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Individual filed Critical Individual
Priority to CN202110652217.0A priority Critical patent/CN113283380A/en
Publication of CN113283380A publication Critical patent/CN113283380A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/20—Movements or behaviour, e.g. gesture recognition
    • G06V40/23—Recognition of whole body movements, e.g. for sport training
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • 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)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Molecular Biology (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Computing Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Psychiatry (AREA)
  • Social Psychology (AREA)
  • Human Computer Interaction (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a children motion posture automatic identification method based on a 3D convolution long-term and short-term memory network, which comprises the following steps: step S1, recording three sections of videos of the inspected person walking naturally indoors by using a non-tip or tip and above camera, wherein the videos are the videos of the natural walking indoors, the videos are fixed at the iliac spine of the inspected person in a height horizontal mode, the videos are arranged 5m away from the inspected person by the camera, and the angles of the three sections of videos are the front side, the back side and the side face respectively; s2, carrying out optimal estimation on the motion target information by using a Kalman filtering technology on the acquired image video, and realizing optimal acquisition on the motion target tracking image; and step S3, preprocessing the optimal moving target tracking image to obtain an enhanced image. The invention detects the diseases of the whole body of the human body through the dynamic video collected by the non-tip or tip camera and the cameras above, has diversified corresponding modes and plays the roles of early warning and correction training.

Description

Children motion attitude automatic identification method based on 3D convolution long-term and short-term memory network
Technical Field
The invention relates to the technical field of intelligent medical treatment, in particular to a children motion posture automatic identification method based on a 3D convolution long-term and short-term memory network.
Background
The Long Short-Term Memory network (LSTM) is a time-cycle neural network, which is specially designed to solve the Long-Term dependence problem of the general RNN (cyclic neural network), and all RNNs have a chain form of repeated neural network modules. In the standard RNN, this repeated structure block has only a very simple structure, e.g. one tanh layer.
At present, the technology of diagnosing human diseases through computer vision is gradually mature, especially for diseases in the aspect of human body, such as humpback, lumbar vertebra arch and the like, the trend of problems of children groups is more and more younger, and at the same time, the technology is mostly used for checking by using a sensor on a human body belt, and the technology is assisted by a wearable system, so that the operation process is complex, time-consuming and labor-consuming. At present, most of people in the prior art are adults, and adult kinematic parameters and models cannot be used for children. Meanwhile, most of the conventional technical network structures are used for checking through inputting the pictures of the checked person, the difficulty in acquiring the human body posture in the picture processing is high, accurate human body posture judgment is not easy to make, so that the error of the diagnosis result is large and inaccurate, and the technology for diagnosing through computer vision becomes unreliable. Therefore, it is necessary to design a method for automatically recognizing the motion posture of the child based on a 3D convolution long-short term memory network.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides the automatic children motion posture recognition method based on the 3D convolution long-short term memory network.
In order to achieve the purpose, the invention provides the following technical scheme: the children motion posture automatic identification method based on the 3D convolution long-term and short-term memory network comprises the following steps:
step S1, recording three sections of videos of the inspected person walking naturally indoors by using a non-tip or tip and above camera, wherein the videos are the videos of the natural walking indoors, the videos are fixed at the iliac spine of the inspected person in a height horizontal mode, the videos are arranged 5m away from the inspected person by the camera, and the angles of the three sections of videos are the front side, the back side and the side face respectively;
s2, carrying out optimal estimation on the motion target information by using a Kalman filtering technology on the acquired image video, and realizing optimal acquisition on the motion target tracking image;
step S3, preprocessing the optimal moving target tracking image to obtain an enhanced image;
step S4, processing the acquired video by adopting a deppose network in python language, and acquiring a human body posture template through processing;
s5, predicting a human posture template by adopting a deep learning framework keras of Google in python language; then, a classification network structure based on ConvLSTM is independently established for each disease by adopting a deep learning framework keras in python language to perform two-classification prediction;
and step S6, after analysis, a possible disease state result report is given, and an early warning or correction method is prompted.
Preferably, the acquiring of the optimal moving object tracking image based on the kalman filtering technique in step S2 includes the following steps:
1) acquiring an initial position and information of a target in an image video by acquiring a target image sequence in the image video, and endowing the information to a Kalman filter for initialization setting;
2) estimating the position of the target in the next frame of image video by using Kalman filtering to obtain an estimated value, and narrowing the range for the detection of the next frame;
3) and comparing the actually detected target value with the estimated value, updating the parameters of the filter, preparing for predicting the next frame, and acquiring the tracking image of the optimal moving target in repeated prediction and updating.
Preferably, the Kalman filter is described by two parts, namely a state equation and an observation equation, and if the Kalman filter is a discrete dynamic system, the two parts are an N-dimensional dynamic system and a P-dimensional (P ≦ N) observation system, wherein U isKIs a 1-dimensional input vector, WKNoise vector of dimension M, Xk(X∈Rn) Is an N-dimensional state vector, and the state of the system is generally described by the following difference equation:
Xk＝AXK-1+BUK+WK-1
the n X n matrix A is the state transition matrix of the system, and is used for measuring the system state Xk-1To XkThe key to the transformation of (1), matrix B is an n × 1 matrix, which is actually the input transformation matrix, WkIs white noiseAcoustic, the system's observation equation is as follows:
Zk＝HkXk+Vk
where H is the observation matrix, the general localization constant, VkIs a white noise sequence with a mean of 0.
Preferably, the step S3 is to pre-process the optimal moving object tracking image, and includes:
assuming that the noise on the image is additive, uncorrelated, and has a mean value of zero, a noisy digital image includes the original image signal f (x, y) and the noise n (x, y), i.e.: g (x, y) ═ f (x, y) + n (x, y), and the smoothed image obtained after the local averaging processing is:
Figure BDA0003111841470000031
wherein, S is the point set in the point (x, y) field, M is the total number of points (x, y) in the point set S, and the noise variance can be reduced to 1/M after the image is smoothed.
Preferably, the 3D convolution calculation formula in step S5 is as follows:
Input：(N，Cin，Din，Hin，Win)；
Output：(N，Cout，Dout，Hout，Wout)where；
Figure BDA0003111841470000032
Figure BDA0003111841470000033
Figure BDA0003111841470000041
wherein N is batch _ size, C is the number of channels, D is the image depth, H is the image height, W is the image width, Padding is the space between the element frame and the element content, kernel is the convolution kernel, and stride is the step length of each dimension of the image;
ConvLSTM is calculated as:
it＝σ(Wxi*Xt+Whi*Ht-1+WcioCt-1+bi)；
ft＝σ(Wxf*Xt+Whf*Ht-1+WcfoCt-1+bf)；
Ct＝ftoCt-1+itotanh(Wxc*Xt+Whc*Ht-1+bc)；
ot＝σ(Wxo*Xt+Who*Ht-1+WcooCt+bo)；
Ht＝ototanh(Ct)；
because the training adopts the condition that a plurality of networks are set up to independently analyze each disease condition, the utilization loss of the training is reduced;
the function is a two-class loss function:
Figure BDA0003111841470000042
wherein m is the number of samples, y is the real label of the training set, t is the last layer Dense output of the network, and Π (t) is the sigmoi activation function as follows:
Figure BDA0003111841470000043
wherein t is the output result of the last layer of the network;
optimization algorithm Adam formula:
Figure BDA0003111841470000044
mt＝β1mt-1+(1-β1)gt；
vt＝β2vt-1+(1-β2)g 2；
Figure BDA0003111841470000051
Figure BDA0003111841470000052
Figure BDA0003111841470000053
wherein g is the gradient of t time step, m is the first-order gradient momentum of time step, v is the second-order gradient momentum of time step, m ^ and v ^ are the gradient correction to the momentum, θ t represents the model parameter at the time of the t iteration, ε is a minimum value to prevent the denominator from being 0, β is the power parameter, and 0.9 and 0.999 are respectively taken.
The invention has the beneficial effects that:
1. the invention detects the diseases of the whole body of the human body through the dynamic video acquired by the non-tip or tip camera and the cameras, has diversified corresponding modes and plays the roles of early warning and correction training, and the technology particularly collects the motion attitude data of children and provides a new method for early warning, auxiliary diagnosis and correction training of diseases of young children;
2. according to the invention, the acquired image video is subjected to optimal estimation on the moving target information by using the Kalman filtering technology, so that the optimal acquisition of the moving target tracking image is realized, and the noise reduction processing is carried out on the optimal moving target tracking image, so that the acquired picture quality can be effectively improved, and the later diagnosis precision is conveniently improved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention. In the drawings:
FIG. 1 is a flow chart of a children motion gesture automatic identification method based on a 3D convolution long-short term memory network;
FIG. 2 is a flow chart of the present invention for optimal moving object tracking image acquisition based on Kalman filtering technique.
Detailed Description
The technical scheme of the invention is clearly and completely described in the following with reference to the accompanying drawings. In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, and are only for convenience of description and simplicity of description, but do not indicate or imply that the device or element being referred to must have a particular orientation, be constructed and operated in a particular orientation, and thus, should not be construed as limiting the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance.
Example one
As shown in fig. 1-2, the present invention provides the following technical solutions: the children motion posture automatic identification method based on the 3D convolution long-term and short-term memory network comprises the following steps:
step S1, recording three sections of videos of the inspected person walking naturally indoors by using a non-tip or tip and above camera, wherein the videos are the videos of the natural walking indoors, the videos are fixed at the iliac spine of the inspected person in a height horizontal mode, the videos are arranged 5m away from the inspected person by the camera, and the angles of the three sections of videos are the front side, the back side and the side face respectively;
s2, carrying out optimal estimation on the motion target information by using a Kalman filtering technology on the acquired image video, and realizing optimal acquisition on the motion target tracking image;
step S3, preprocessing the optimal moving target tracking image to obtain an enhanced image;
step S4, processing the acquired video by adopting a deppose network in python language, and acquiring a human body posture template through processing;
s5, predicting a human posture template by adopting a deep learning framework keras of Google in python language; then, a classification network structure based on ConvLSTM is independently established for each disease by adopting a deep learning framework keras in python language to perform two-classification prediction;
and step S6, after analysis, a possible disease state result report is given, and an early warning or correction method is prompted.
Preferably, the acquisition of the optimal moving object tracking image based on the kalman filtering technique in step S2 includes the following steps:
1) acquiring an initial position and information of a target in an image video by acquiring a target image sequence in the image video, and endowing the information to a Kalman filter for initialization setting;
2) estimating the position of the target in the next frame of image video by using Kalman filtering to obtain an estimated value, and narrowing the range for the detection of the next frame;
3) and comparing the actually detected target value with the estimated value, updating the parameters of the filter, preparing for predicting the next frame, and acquiring the tracking image of the optimal moving target in repeated prediction and updating.
Preferably, the Kalman filter is described by two parts, a state equation and an observation equation, respectively, and in the case of a discrete dynamic system, the two parts are an N-dimensional dynamic system and a P-dimensional (P ≦ N) observation system, respectively, where U isKIs a 1-dimensional input vector, WKNoise vector of dimension M, Xk(X∈Rn) Is an N-dimensional state vector, and the state of the system is generally described by the following difference equation:
Xk＝AXK-1+BUK+WK-1
the n X n matrix A is the state transition matrix of the system, and is used for measuring the system state Xk-1To XkThe key to the transformation of (1), matrix B is an n × 1 matrix, which is actually the input transformation matrix, WkIs white noise, the system's observation equation is as follows:
Zk＝HkXk+Vk
where H is the observation matrix, the general localization constant, VkIs a white noise sequence with a mean of 0.
Preferably, the step S3 is to pre-process the optimal moving object tracking image, and includes:
assuming that the noise on the image is additive, uncorrelated, and has a mean value of zero, a noisy digital image includes the original image signal f (x, y) and the noise n (x, y), i.e.: g (x, y) ═ f (x, y) + n (x, y), and the smoothed image obtained after the local averaging processing is:
Figure BDA0003111841470000081
wherein, S is the point set in the point (x, y) field, M is the total number of points (x, y) in the point set S, and the noise variance can be reduced to 1/M after the image is smoothed.
Preferably, the 3D convolution calculation formula in step S5 is as follows:
Input：(N，Cin，Din，Hin，Win)；
Output：(N，Cout，Dout，Hout，Wout)where；
Figure BDA0003111841470000082
Figure BDA0003111841470000083
Figure BDA0003111841470000084
wherein N is batch _ size, C is the number of channels, D is the image depth, H is the image height, W is the image width, Padding is the space between the element frame and the element content, kernel is the convolution kernel, and stride is the step length of each dimension of the image;
ConvLSTM is calculated as:
it＝σ(Wxi*Xt+Whi*Ht-1+WcioCt-1+bi)；
ft＝σ(Wxf*Xt+Whf*Ht-1+WcfoCt-1+bf)；
Ct＝ftoCt-1+itotanh(Wxc*Xt+Whc*Ht-1+bc)；
ot＝σ(Wxo*Xt+Who*Ht-1+WcooCt+bo)；
Ht＝ototanh(Ct)；
because the training adopts the condition that a plurality of networks are set up to independently analyze each disease condition, the utilization loss of the training is reduced;
the function is a two-class loss function:
Figure BDA0003111841470000091
wherein m is the number of samples, y is the real label of the training set, t is the last layer Dense output of the network, and Π (t) is the sigmoi activation function as follows:
Figure BDA0003111841470000092
wherein t is the output result of the last layer of the network;
optimization algorithm Adam formula:
Figure BDA0003111841470000096
mt＝β1mt-1+(1-β1)gt；
vt＝β2vt-1+(1-β2)g 2；
Figure BDA0003111841470000093
Figure BDA0003111841470000094
Figure BDA0003111841470000095
wherein g is the gradient of t time step, m is the first-order gradient momentum of time step, v is the second-order gradient momentum of time step, m ^ and v ^ are the gradient correction to the momentum, θ t represents the model parameter at the time of the t iteration, ε is a minimum value to prevent the denominator from being 0, β is the power parameter, and 0.9 and 0.999 are respectively taken.
Finally, it should be noted that: although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that changes may be made in the embodiments and/or equivalents thereof without departing from the spirit and scope of the invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (5)

1. The children motion attitude automatic identification method based on the 3D convolution long-term and short-term memory network is characterized by comprising the following steps: the method comprises the following steps:
step S1, recording three sections of videos of the inspected person walking naturally indoors by using a non-tip or tip and above camera, wherein the videos are the videos of the natural walking indoors, the videos are fixed at the iliac spine of the inspected person in a height horizontal mode, the videos are arranged 5m away from the inspected person by the camera, and the angles of the three sections of videos are the front side, the back side and the side face respectively;
s2, carrying out optimal estimation on the motion target information by using a Kalman filtering technology on the acquired image video, and realizing optimal acquisition on the motion target tracking image;
step S3, preprocessing the optimal moving target tracking image to obtain an enhanced image;
step S4, processing the acquired video by adopting a deppose network in python language, and acquiring a human body posture template through processing;
s5, predicting a human posture template by adopting a deep learning framework keras of Google in python language; then, a classification network structure based on ConvLSTM is independently established for each disease by adopting a deep learning framework keras in python language to perform two-classification prediction;
and step S6, after analysis, a possible disease state result report is given, and an early warning or correction method is prompted.
2. The method for automatically recognizing the motion posture of the child based on the 3D convolution long-short term memory network as claimed in claim 1, wherein the method comprises the following steps: the acquisition of the optimal moving target tracking image based on the kalman filtering technique in step S2 includes the following steps:
1) acquiring an initial position and information of a target in an image video by acquiring a target image sequence in the image video, and endowing the information to a Kalman filter for initialization setting;
2) estimating the position of the target in the next frame of image video by using Kalman filtering to obtain an estimated value, and narrowing the range for the detection of the next frame;
3) and comparing the actually detected target value with the estimated value, updating the parameters of the filter, preparing for predicting the next frame, and acquiring the tracking image of the optimal moving target in repeated prediction and updating.
3. The method for automatically recognizing the motion posture of the child based on the 3D convolution long-short term memory network as claimed in claim 2, wherein the method comprises the following steps: the Kalman filter is described by two parts, namely a state equation and an observation equation, if the Kalman filter is a discrete dynamic system, the two parts are an N-dimensional dynamic system and a P-dimensional (P is less than or equal to N) observation system respectively, wherein U isKIs a 1-dimensional input vector, WKNoise vector of dimension M, Xk(X∈Rn) Is an N-dimensional state vector, and the state of the system is generally described by the following difference equation:
Xk＝AXK-1+BUK+WK-1
the n X n matrix A is the state transition matrix of the system, and is used for measuring the system state Xk-1To XkThe key to the transformation of (1), matrix B is an n × 1 matrix, which is actually the input transformation matrix, WkIs white noise, the system's observation equation is as follows:
Zk＝HhXk+Vk
where H is the observation matrix, the general localization constant, VkIs a white noise sequence with a mean of 0.
4. The method for automatically recognizing the motion posture of the child based on the 3D convolution long-short term memory network as claimed in claim 1, wherein the method comprises the following steps: the step S3 is to pre-process the optimal moving object tracking image, and includes:
assuming that the noise on the image is additive, uncorrelated, and has a mean value of zero, a noisy digital image includes the original image signal f (x, y) and the noise n (x, y), i.e.: g (x, y) ═ f (x, y) + n (x, y), and the smoothed image obtained after the local averaging processing is:
Figure FDA0003111841460000021
wherein, S is the point set in the point (x, y) field, M is the total number of points (x, y) in the point set S, and the noise variance can be reduced to 1/M after the image is smoothed.
5. The method for automatically recognizing the motion posture of the child based on the 3D convolution long-short term memory network as claimed in claim 1, wherein the method comprises the following steps: the 3D convolution calculation formula in step S5 is as follows:
Input：(N,Cin,Din,Hin,Win)；
Output：(N,Cout,Dout,Hout,Wout)where；
Figure FDA0003111841460000031
Figure FDA0003111841460000032
Figure FDA0003111841460000033
wherein N is batch _ size, C is the number of channels, D is the image depth, H is the image height, W is the image width, Padding is the space between the element frame and the element content, kernel is the convolution kernel, and stride is the step length of each dimension of the image;
ConvLSTM is calculated as:
it＝σ(Wxi*Xt+Whi*Ht-1+WcioCt-1+bi)；
ft＝σ(Wxf*Xt+Whf*Ht-1+WcfoCt-1+bf)；
Ct＝ftoCt-1+itotanh(Wxc*Xt+Whc*Ht-1+bc)；
ot＝σ(Wxo*Xt+Who*Ht-1+WcooCt+bo)；
Ht＝ototanh(Ct)；
because the training adopts the condition that a plurality of networks are set up to independently analyze each disease condition, the utilization loss of the training is reduced;
the function is a two-class loss function:
Figure FDA0003111841460000034
wherein m is the number of samples, y is the real label of the training set, t is the last layer Dense output of the network, and Π (t) is the sigmoi activation function as follows:
Figure FDA0003111841460000041
wherein t is the output result of the last layer of the network;
optimization algorithm Adam formula:
gt＝▽θJ(θt-1)；
mt＝β1mt-1+(1-β1)gt；
vt＝β2vt-1+(1-β2)g2；
Figure FDA0003111841460000042
Figure FDA0003111841460000043
Figure FDA0003111841460000044
wherein g is the gradient of t time step, m is the first-order gradient momentum of time step, v is the second-order gradient momentum of time step, m ^ and v ^ are the gradient correction to the momentum, θ t represents the model parameter at the time of the t iteration, ε is a minimum value to prevent the denominator from being 0, β is the power parameter, and 0.9 and 0.999 are respectively taken.
CN202110652217.0A 2021-06-11 2021-06-11 Children motion attitude automatic identification method based on 3D convolution long-term and short-term memory network Pending CN113283380A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110652217.0A CN113283380A (en) 2021-06-11 2021-06-11 Children motion attitude automatic identification method based on 3D convolution long-term and short-term memory network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110652217.0A CN113283380A (en) 2021-06-11 2021-06-11 Children motion attitude automatic identification method based on 3D convolution long-term and short-term memory network

Publications (1)

Publication Number Publication Date
CN113283380A true CN113283380A (en) 2021-08-20

Family

ID=77284273

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110652217.0A Pending CN113283380A (en) 2021-06-11 2021-06-11 Children motion attitude automatic identification method based on 3D convolution long-term and short-term memory network

Country Status (1)

Country Link
CN (1) CN113283380A (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106780542A (en) * 2016-12-29 2017-05-31 北京理工大学 A kind of machine fish tracking of the Camshift based on embedded Kalman filter
US20170263005A1 (en) * 2016-03-10 2017-09-14 Sony Corporation Method for moving object detection by a kalman filter-based approach
CN110321937A (en) * 2019-06-18 2019-10-11 哈尔滨工程大学 A kind of moving human hand tracking method of Faster-RCNN combination Kalman filtering
CN111079656A (en) * 2019-12-18 2020-04-28 林晓莹 Children motion attitude automatic identification technology based on 3D convolution long-term and short-term memory network

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170263005A1 (en) * 2016-03-10 2017-09-14 Sony Corporation Method for moving object detection by a kalman filter-based approach
CN106780542A (en) * 2016-12-29 2017-05-31 北京理工大学 A kind of machine fish tracking of the Camshift based on embedded Kalman filter
CN110321937A (en) * 2019-06-18 2019-10-11 哈尔滨工程大学 A kind of moving human hand tracking method of Faster-RCNN combination Kalman filtering
CN111079656A (en) * 2019-12-18 2020-04-28 林晓莹 Children motion attitude automatic identification technology based on 3D convolution long-term and short-term memory network

Similar Documents

Publication Publication Date Title
Chambers et al. Computer vision to automatically assess infant neuromotor risk
US11017547B2 (en) Method and system for postural analysis and measuring anatomical dimensions from a digital image using machine learning
Khokhlova et al. Normal and pathological gait classification LSTM model
CN108305283B (en) Human behavior recognition method and device based on depth camera and basic gesture
CN107666853A (en) Beat signals are determined according to video sequence
CN111339942A (en) Method and system for recognizing skeleton action of graph convolution circulation network based on viewpoint adjustment
CN106846372B (en) Human motion quality visual analysis and evaluation system and method thereof
CN110956141B (en) Human body continuous action rapid analysis method based on local recognition
CN114999646B (en) Newborn exercise development assessment system, method, device and storage medium
CN112101235B (en) Old people behavior identification and detection method based on old people behavior characteristics
CN111079656A (en) Children motion attitude automatic identification technology based on 3D convolution long-term and short-term memory network
CN116597940A (en) Modeling method of movement disorder symptom quantitative evaluation model
Yin et al. Accurate estimation of body height from a single depth image via a four-stage developing network
CN115346272A (en) Real-time tumble detection method based on depth image sequence
Hafeez et al. Multi-Sensor-Based Action Monitoring and Recognition via Hybrid Descriptors and Logistic Regression
Wairagkar et al. A novel approach for modelling and classifying sit-to-stand kinematics using inertial sensors
CN113283380A (en) Children motion attitude automatic identification method based on 3D convolution long-term and short-term memory network
CN116999057A (en) Hemiplegia gait recognition and hemiplegia gait evaluation method based on wearable sensor
WO2024036825A1 (en) Attitude processing method, apparatus and system, and storage medium
Dogra et al. Toward automating Hammersmith pulled-to-sit examination of infants using feature point based video object tracking
CN115331153B (en) Posture monitoring method for assisting vestibule rehabilitation training
CN114469072B (en) Method for automatically predicting psychological development of infants by using camera
Endres et al. Graph-based action models for human motion classification
CN112102358B (en) Non-invasive animal behavior characteristic observation method
Khokhlova et al. Kinematic covariance based abnormal gait detection

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