CN109731302A - Athletic posture recognition methods, device and electronic equipment - Google Patents
Athletic posture recognition methods, device and electronic equipment Download PDFInfo
- Publication number
- CN109731302A CN109731302A CN201910060946.XA CN201910060946A CN109731302A CN 109731302 A CN109731302 A CN 109731302A CN 201910060946 A CN201910060946 A CN 201910060946A CN 109731302 A CN109731302 A CN 109731302A
- Authority
- CN
- China
- Prior art keywords
- data
- action data
- athletic posture
- action
- neural networks
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Abstract
The present invention provides a kind of athletic posture recognition methods, device and electronic equipments, are related to Intelligent sports equipment technical field, which includes the first action data for obtaining the handheld motion instrument that attitude transducer currently acquires；The attitude transducer is set on the handheld motion instrument；Effective action data are determined using difference value-based algorithm based on the second action data acquired before above-mentioned first action data and preset time period；Dimension is carried out to the effective action data to handle, and obtains standard operation data；The standard operation data are input in convolutional neural networks trained in advance and carry out the classification of motion, obtains the athletic posture type of current handheld sports apparatus.Which can more accurately identify that the athletic posture type of handheld motion instrument effectively reduces cost without numerous peripheral apparatus, widened application range, promoted the popularization and use of product.
Description
Technical field
The present invention relates to Intelligent sports equipment technical field, more particularly, to a kind of athletic posture recognition methods, device and
Electronic equipment.
Background technique
With the raising that nationwide fitness programs are realized, body-building ranks are actively added in more and more people, and specification movement and technology are closed
The grasp of key point becomes a difficult point.Sportsman or individual lover can only use for the essential of exercise in motion process
The mode for video playback+actually practice carries out movement correction and movement explanation.This mode causes the waste of plenty of time, efficiency
Under very low, therefore more and more people wish to find a kind of smart machine for being able to reflect out exercise data, so as to directly
The athletic posture of observation oneself is connect, thus Intelligent sports equipment comes into being.
For example, being equipped with two cameras, a high definition throwing in smart table tennis table in the movement monitoring of soldier's pang ball movement
Shadow instrument and several motion sensors.Camera is responsible for shooting the different movement of table tennis, and built-in various different function
Sensor then is used to track the motion profile of table tennis.Meanwhile high definition projector is then responsible for the movement rail of projection table tennis in real time
Mark and a variety of different data.Entire ping pong table board is exactly the touch display screen of a super large, and is equipped with identification
Module, as long as one's own racket is placed on specific region, so that it may show the every exercise data and posture of the sportsman.
However, being acquired to athletic posture above by the mode of camera and monitoring sphere drop point and knowing method for distinguishing, institute
The peripheral apparatus needed is numerous, with high costs, uses using being limited and being unfavorable for product promotion.
Summary of the invention
In view of this, the purpose of the present invention is to provide a kind of athletic posture recognition methods, device and electronic equipments, with more
It accurately identifies the athletic posture type of handheld motion instrument, and is not necessarily to numerous peripheral apparatus, effectively reduce cost, widen
Application range promotes the popularization and use of product.
In a first aspect, being applied to mobile terminal the embodiment of the invention provides a kind of athletic posture recognition methods, comprising:
Obtain the first action data of the handheld motion instrument that attitude transducer currently acquires；Wherein, the posture sensing
Device is set on the handheld motion instrument, and first action data includes acceleration information；
Based on the second action data acquired before first action data and preset time period, calculated using difference value
Method determines effective action data；
Dimension is carried out to the effective action data to handle, and obtains standard operation data；
The standard operation data are input in convolutional neural networks trained in advance and carry out the classification of motion, obtained current
The athletic posture type of the handheld motion instrument.
With reference to first aspect, the embodiment of the invention provides the first possible embodiments of first aspect, wherein base
The second action data acquired before first action data and preset time period is determined effective using difference value-based algorithm
The step of action data, comprising:
Calculate the difference between first action data and second action data；
The difference and predetermined threshold value are compared, determine the initial data and endpoint data of valid window,
Obtain current effective action data.
With reference to first aspect, the embodiment of the invention provides second of possible embodiments of first aspect, wherein institute
The training process for stating convolutional neural networks includes:
The sample action data of each athletic posture type are obtained, and mark the athletic posture class of the sample action data
Type；
Sample valid data are determined using difference value-based algorithm based on the sample action data after mark；
Dimension is carried out to the sample valid data to handle, and obtains master sample data；
The master sample data are input to convolutional neural networks, the convolutional neural networks are trained, are obtained
Convolutional neural networks after training.
With reference to first aspect, the embodiment of the invention provides the third possible embodiments of first aspect, wherein also
Include:
The current athletic posture type of the handheld motion instrument is obtained, and updates the current athletic posture type pair
The amount of action answered；
Obtain the corresponding amount of action of the highest at least one athletic posture type of current discrimination；The wherein knowledge
Rate is obtained in advance in the training process of the convolutional neural networks；
Calculate the corresponding amount of action of at least one athletic posture type and value, it obtains movement total quantity and carries out
Display.
The third possible embodiment with reference to first aspect, the embodiment of the invention provides the 4th kind of first aspect
Possible embodiment, wherein further include:
Obtain the training mode that user currently selects；
When the training mode is individual event training mode, target athletic posture corresponding to the individual event training mode is determined
Type；
Obtain current movement total quantity and the target athletic posture type currently corresponding target action quantity；
According to the target action quantity and the movement total quantity, the movement accuracy rate of the user is determined.
With reference to first aspect, the embodiment of the invention provides the 5th kind of possible embodiments of first aspect, wherein also
Include: acquisition and shows that current kinematic parameter, the kinematic parameter are corresponding including movement duration, the athletic posture type
Amount of action, energy consumption values and maximum movement speed.
The 5th kind of possible embodiment with reference to first aspect, the embodiment of the invention provides the 6th kind of first aspect
Possible embodiment, wherein the acquisition process of the energy consumption values includes:
Obtain the personal information and current movement duration of user belonging to first action data；
According to the personal information, movement duration and the corresponding amount of action of each athletic posture type, institute is determined
State the energy consumption values of user.
With reference to first aspect, the embodiment of the invention provides the 7th kind of possible embodiments of first aspect, wherein also
Include: the handheld motion instrument be table tennis bat.
Second aspect, the embodiment of the present invention also provide a kind of athletic posture identification device, are applied to mobile terminal, comprising:
Data acquisition module, for obtaining the first action data of the handheld motion instrument that attitude transducer currently acquires；
Wherein, the attitude transducer is set on the handheld motion instrument, and first action data includes acceleration information；
Data determining module, for based on the second movement number acquired before first action data and preset time period
According to determining effective action data using difference value-based algorithm；
Dimension processing module is gone, is handled for carrying out dimension to the effective action data, obtains standard operation data；
Classification of motion module is carried out for the standard operation data to be input in convolutional neural networks trained in advance
The classification of motion obtains the athletic posture type of presently described handheld motion instrument.
The third aspect, the embodiment of the present invention also provide a kind of electronic equipment, including memory, processor, the memory
On be stored with the computer program that can be run on the processor, the processor is realized when executing the computer program
State method described in first aspect and its any possible embodiment.
The embodiment of the present invention bring it is following the utility model has the advantages that
In embodiments of the present invention, which includes the hand-held fortune for obtaining attitude transducer and currently acquiring
First action data of dynamic instrument；Wherein, which is set on the handheld motion instrument, above-mentioned first action data
Including acceleration information；Based on the second action data acquired before above-mentioned first action data and preset time period, difference is utilized
Scoring algorithm determines effective action data；Dimension is carried out to the effective action data to handle, and obtains standard operation data；It will
The standard operation data are input in convolutional neural networks trained in advance and carry out the classification of motion, obtain current handheld sports apparatus
Athletic posture type.Acquire data in which by the attitude transducer in handheld motion instrument, mobile terminal is to adopting
The data collected carry out data extraction, are identified to the valid data extracted by convolutional neural networks, so as to more
It accurately identifies that the athletic posture type of handheld motion instrument effectively reduces cost without numerous peripheral apparatus, widens
Application range, promotes the popularization and use of product.
Other features and advantages of the present invention will illustrate in the following description, also, partly become from specification
It obtains it is clear that understand through the implementation of the invention.The objectives and other advantages of the invention are in specification, claims
And specifically noted structure is achieved and obtained in attached drawing.
To enable the above objects, features and advantages of the present invention to be clearer and more comprehensible, preferred embodiment is cited below particularly, and cooperate
Appended attached drawing, is described in detail below.
Detailed description of the invention
It, below will be to specific in order to illustrate more clearly of the specific embodiment of the invention or technical solution in the prior art
Embodiment or attached drawing needed to be used in the description of the prior art be briefly described, it should be apparent that, it is described below
Attached drawing is some embodiments of the present invention, for those of ordinary skill in the art, before not making the creative labor
It puts, is also possible to obtain other drawings based on these drawings.
Fig. 1 is a kind of flow diagram of athletic posture recognition methods provided in an embodiment of the present invention；
Fig. 2 is a kind of schematic diagram of valid window provided in an embodiment of the present invention；
Fig. 3 is the schematic diagram of the content of effective action data provided in an embodiment of the present invention；
Fig. 4 is a kind of structural schematic diagram of convolutional neural networks provided in an embodiment of the present invention；
Fig. 5 is a kind of block schematic illustration of application software provided in an embodiment of the present invention；
Fig. 6 is a kind of structural schematic diagram of athletic posture identification device provided in an embodiment of the present invention；
Fig. 7 is the structural schematic diagram of a kind of electronic equipment provided in an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with attached drawing to the present invention
Technical solution be clearly and completely described, it is clear that described embodiments are some of the embodiments of the present invention, rather than
Whole embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not making creative work premise
Under every other embodiment obtained, shall fall within the protection scope of the present invention.
At present by way of camera and monitoring sphere drop point, athletic posture is acquired and knows method for distinguishing, it is required
Peripheral apparatus is numerous, with high costs, uses using being limited and being unfavorable for product promotion.It is provided in an embodiment of the present invention based on this
A kind of athletic posture recognition methods, device and electronic equipment, acquire data by the attitude transducer in handheld motion instrument,
Mobile terminal carries out data extraction to collected data, is known to the valid data extracted by convolutional neural networks
Not, it so as to more accurately identify the athletic posture type of handheld motion instrument, without numerous peripheral apparatus, effectively drops
Low cost, has widened application range, has promoted the popularization and use of product.
For convenient for understanding the present embodiment, first to a kind of athletic posture identification side disclosed in the embodiment of the present invention
Method describes in detail.
Embodiment one:
Fig. 1 shows a kind of flow diagram of athletic posture recognition methods provided in an embodiment of the present invention.This method can
With but during being not limited to the movement monitoring for being applied to the handheld motions instrument such as soldier's pang ball, shuttlecock and tennis, can be by loading
There are mobile terminal (such as mobile phone, tablet computer, laptop) either other electronic equipments of corresponding software to realize.Such as Fig. 1
It is shown, this method including the following steps:
Step S101 obtains the first action data of the handheld motion instrument that attitude transducer currently acquires.
Wherein, above-mentioned attitude transducer is set on above-mentioned handheld motion instrument, which can be, but not limited to
Select nine axis accelerometer of MPU9250.Above-mentioned first action data is can wrap in the collected exercise data of current sampling point
Acceleration information is included, can also include angular acceleration and corresponding timestamp etc..
Step S102 is utilized based on the second action data acquired before above-mentioned first action data and preset time period
Difference value-based algorithm determines effective action data.
Wherein preset time period may be set according to actual conditions, and such as need the n-th sampling before current sampling point
Point is used as the second action data, then preset time period is N × sampling interval, and wherein N is positive integer.
Step S103 carries out dimension to above-mentioned effective action data and handles, obtains standard operation data.
Nondimensional pure values are converted by obtained effective action data, in order to the not index energy of commensurate or magnitude
It is enough compared and weights.
Above-mentioned standard action data is input in convolutional neural networks trained in advance and carries out movement point by step S104
Class obtains the athletic posture type of current handheld sports apparatus.
Wherein the athletic posture type can be, but not limited to include that smash after drawing, forehand off-table chop, deep backhand chop, forehand are attacked
Ball.
In embodiments of the present invention, data, mobile terminal pair are acquired by the attitude transducer in handheld motion instrument
Collected data carry out data extraction, are identified to the valid data extracted by convolutional neural networks, so as to
It more accurately identifies that the athletic posture type of handheld motion instrument effectively reduces cost without numerous peripheral apparatus, opens up
Wide application range, promotes the popularization and use of product.
It should be noted that method provided in an embodiment of the present invention can be loaded in the form of application software in mobile terminal
Either on the electronic equipments such as computer, and executed in the embodiment of the present invention by the processor of the mobile terminal either computer
Correlation step.
In a possible embodiment, in order to guarantee accuracy that data are extracted, above-mentioned steps S102 includes: that calculate first dynamic
Make the difference between data and the second action data；The difference and predetermined threshold value are compared, determine valid window
Initial data and endpoint data, obtain current effective action data.
For example, carrying out difference value (i.e. the first action data and the of sliding window calculating action with the width of 4 sampled points
Difference between two action datas).Assuming that the first action data at some sampled point is expressed as acc1_n, then before it
The second action data at 4 sampled points can be expressed as acc1_ (n+4), then the first action data and the second action data it
Between difference can indicate are as follows: Delta_acc1=acc1_n-acc1_ (n-4).If Delta_acc1 > threshold value, then it is assumed that
The sampled point is acceleration starting point (initial data of valid window), Delta_acc1 < threshold value, then it is assumed that the sampled point is
Acceleration terminal (endpoint data of valid window), thus using the data between acceleration starting point and acceleration terminal as current
Effective action data, as shown in Figure 2.
Assuming that the acceleration information in above-mentioned first action data includes 3-axis acceleration (X-axis, Y-axis and Z axis), it can be with table
It is shown asThen the difference between the first action data and the second action data can indicate are as follows:Compare ΔaccNWith the size of threshold value.
In a possible embodiment, the training process of above-mentioned convolutional neural networks includes: to obtain each athletic posture type
Sample action data, and mark the athletic posture type of the sample action data；Based on the sample action data after mark, benefit
With difference value-based algorithm, sample valid data are determined；Dimension is carried out to the sample valid data to handle, and obtains master sample number
According to；The master sample data are input to convolutional neural networks, which are trained, the volume after being trained
Product neural network.
It, can be by being embedded in the handheld motion equipment of attitude transducer (such as table tennis in the collection process of sample action data
Pang racket) carry out data acquisition.Data pass to host computer by bluetooth.The host computer can be mobile terminal, or electricity
Brain etc..The preservation of data is carried out using master system.
By acquisition target repeat do corresponding actions, as after forehand off-table chop, deep backhand chop, drawing smash and forehand attack this
Four movements and other movements, obtain the sample action data of a large amount of X, Y, Z axis.User can certainly be made to upload to cloud clothes
The data of business platform directly obtain data from the cloud service platform in training.Then handmarking's sample action number
According to corresponding athletic posture type.In labeling process, since movement to be identified in practical scene can be adulterated with violate-action
Together, and actual interference type of action is excessive, can not be as the movement label for generalizing (extensive), and direct artificial extract
Data time-consuming low efficiency carries out effective action data to the acceleration information in the first action data so using difference value
Interception (being referred to the acquisition process of above-mentioned effective action data), obtain sample valid data.Due to attitude transducer
Sample frequency it is too low, therefore by be arranged threshold value detect effective action window automatically.
It is specifically the movement label of every kind of athletic posture type configuration after to effective action data cutout and label
As follows: 1 is deep backhand chop, and 3 be forehand off-table chop, and 5 be other movements, and 7 is smash after drawing, 9 be forehand attack.By manually carrying out
The integration of data set, in the data set after integration, the time window length of each athletic posture type to be identified is consistent, in number
According to concentration random distribution.In a possible embodiment, to the part effective action data of sample action data cutout, (i.e. sample has
Imitate data) content is referring to Fig. 3.Wherein in Fig. 3, each field respectively indicates movement label, athletic posture type name in every row
The acceleration information of title, timestamp and X, Y, Z axis.
Before sample valid data obtained above are input to convolutional neural networks, auxiliary function can also be defined
The sample valid data are read out, and to the sample valid data standardization, i.e., by the acceleration of above-mentioned X, Y, Z axis
Information is converted into nondimensional pure values, is able to carry out convenient for the index of not commensurate or magnitude and compares and weight.Common number
Have according to standardized method: min-max standardizes (Min-max normalization), log function is converted, atan function is converted,
Z-score standardizes (zero-mena normalization).In the present embodiment, select z-score standardization above-mentioned
Sample valid data obtain master sample data.It is assumed that the sequence of X-axis acceleration information is x1, x2... ... xn, then the sequence
In any X-axis acceleration information xiA can be expressed as after standardizationi:
Wherein,
Similarly, same processing is done to Y-axis acceleration information and Z axis acceleration information, it is nondimensional after obtaining standardization
Y-axis acceleration information and Z axis acceleration information.
After the completion of standardization, X, Y, Z axis is established respectively according to pre-set time window (i.e. sampling period)
Relationship of the acceleration information in time series between the several points in front and back.In order to be adapted to 3-axis acceleration information preferably
Convolutional neural networks, enhancing surrounding time stab the correlation between upper each axle acceleration, this model, which uses sliding window and does, to be rolled over
(sliding window length is w to the mode on folded ground, and sliding step is assumed to be c).After the data of X, Y, Z axis are folded by sliding window,
It is formed the two-dimensional matrix of 3 similar RGB (RGB) triple channels.
The data set that the two-dimensional matrix of 3 obtained above similar RGB triple channels is formed, can be, but not limited to according to 6:
The ratio of 3:1 is divided into training set, test set, verifying collection, is then loaded in batches into memory, is trained to convolutional neural networks.
When applying deep learning algorithm in athletic posture identification field, most of is the angle conduct from computer vision
Point of penetration.However in practical applications, especially there are many limitations in embedded platform.And it is based on attitude transducer,
The overwhelming majority is to add traditional machine learning algorithm (such as KNN (K-Nearest Neighbor, K- using artificial characteristic value of extracting
It is neighbouring), SVM (Support Vector Machine, support vector machines) etc.), the program is for characteristics extraction professional knowledge
It is more demanding.In fact, for the attitude transducer although acceleration easy to accomplish in X, Y, Z axis respectively time series, angle
The analysis of the data such as speed, but the analysis for a moving object track, it is also necessary in combination with the research of spatial relationship.
Therefore with conventional method go solve the time and space on stationary problem will be especially complex.
Based on this, using X, Y, Z axis as the RGB triple channel of Image neighborhood in the embodiment of the present invention, by each time point
The acceleration information of upper three axis regards a pixel as.In addition, in image procossing neighborhood, each pixel and surrounding
Pixel has certain correlation, and in athletic posture identification, three axis at a upper time point and next time point accelerate
Degree information equally has correlation.Therefore, in the athletic posture classification and identification algorithm in inventive embodiments, image can be used
The more mature convolutional neural networks in identification field.
Convolutional neural networks (Convolutional Neutral Network, abbreviation CNN) are a kind of Feedforward Neural Networks
Network is proposed by the mechanism of biologically receptive field (Receptive Field).There are three in structure for convolutional neural networks
Characteristic: part connection, weight be shared and space or temporal sampling.These characteristics make convolutional neural networks have one
Determine translation, scaling and the distortion invariance in degree.Convolutional neural networks pattern-recognition, in terms of obtain extensively
Using.The convolutional neural networks structure used in the embodiment of the present invention, as shown in figure 4, specifically including: channel input layer, convolution
Layer, maximum pond layer, convolutional layer, full articulamentum and Softmax output layer.Above layers are introduced below.
Convolutional layer:
It is real when two-dimensional matrix (X, Y, Z axis acceleration information) divides the data parallel of triple channel to input from channel input layer
The connection of the acceleration of moving object spatially is showed.Initialize neural network parameter weight (weight_variable) simultaneously
It is that the data in each channel execute one-dimensional convolution, the filter of first convolutional layer with bias (bias_variable)
(filter) size and depth are 60, and sliding step (strides) is [1,1,1,1], and filling mode (padding) is complete
0 filling (VALID).Then pass through the transmitting output of ReLU activation primitive.Compared to sigmoid/tanh function, ReLU only needs one
A threshold value can be obtained by activation value, and the operation of a lot of complexity is calculated without spending.The characteristics of ReLU activation primitive: input signal
When<0, output is all 0, and in the case where input signal>0, output is equal to input.
ReLU activation primitive:
Currently, deep learning one specific target is that key factor is dissociateed from data variable.Initial data is (with certainly
Based on right data) in be usually wrapped the feature of highly dense.However, if it is possible to the complex relationship wound between feature is unlocked,
Sparse features are converted to, then feature just has robustness (eliminating unrelated noise).Sparse features do not need network tool
There is very strong processing linearly inseparable mechanism.Therefore in depth network, nonlinear degree of dependence can reduce.Once refreshing
Through being changed to linearly activate between member and neuron, the non-linear partial of network is activated only from neuron partial selective.
It is existing to calculate neural network and biological neural network there is a big difference from the point of view of 95% sparsity for comparing brain work.And
ReLu only have negative value just can by it is sparse fall, that is, the sparsity introduced can train adjusting, be dynamic change.As long as carrying out
Gradient training, network can guarantee that there is fair amounts on activation chain to the direction of error reduction, the sparse ratio of auto-control
Nonzero value.
Pond layer:
General to carry out pondization come the output of understand to convolutional layer and operate, common pond layer operation has maximum pond layer (max_
Pooling) and average pond layer (average), layer available more abstract action data in pond improves action model
Generality (generalization) is observed after many experiments and is more suitable for the present invention in fact using the pondization operation of maximum pond layer
Apply the model in example.So pond layer choosing maximum pond layer.Then using one layer of convolutional layer.
Full articulamentum:
Then the output of convolutional layer will enter fully-connected network, when action data have passed through the convolutional layer and pond layer of front
After processing, it has been abstracted into the higher feature of information content, convolutional layer and pond layer can have been found out into automatic movement at this time
The process of feature extraction, feature extraction is complete can to use full articulamentum execution classification task.
Softmax layers:
Convolutional neural networks model in the embodiment of the present invention can recognize the movement of five classes, before using Softmax function, need
It is added dropout layers and certain hiding neuron nodes is placed in disabled state, prevent model over-fitting.
After obtaining complete convolutional neural networks, which is trained, gradient can be used when training
Descent algorithm.Common gradient algorithm includes comprehensive gradient descent algorithm and stochastic gradient descent algorithm (stochastic
Gradient descent), when comprehensive gradient algorithm each training pattern, can all traverse a sample data, and very consumption calculates
Resource.In order to accelerate the training process of model, stochastic gradient descent algorithm can be used.That is, training pattern mentions every time
Input of the sample action information of a part of table tennis movement as training sample is taken, these samples also become a batch, and
These samples can represent complete training set information to a certain extent.In TensorFlow, a sample is specified
Size (size) come provide every time randomly choose participate in conclude sample size, by the training of iteration (epoch), and to training
Convolutional neural networks model loss late and accuracy rate printed.
After the convolutional neural networks model after being trained, which is disposed, such as portion
Affix one's name to the Android system of mobile terminal.Firstly, the core of TensorFlow is write with c++.In order to construct Android
Mesh, it is necessary to call c++ function, such as loadModel, getPredictions using JNI (Java Native Interface)
Deng.So compiling .so (shared object) file and jar file by bazel, it is a c++ compiling file and one
Jar file, it will be by calling the JAVA API of the machine c++ to form.Compiled .so and jar file are put into Android project
Under libs file, the model file trained is put under assert file.One is finally created in Android Studio
Class is called and obtains output to convolutional neural networks model inside class.
In a possible embodiment, the action data that attitude transducer acquires can be transmitted to by mobile terminal by bluetooth.
For mobile terminal by intercepting to action data, the effective action data that will acquire carry out data normalization processing, as convolution mind
Input through network will finally return to the discrimination of each athletic posture type, and record to the discrimination.
In order to guarantee the accurate fixed of data, in a possible embodiment, the above method further include: obtain above-mentioned handheld motion
The current athletic posture type of instrument, and update the current corresponding amount of action of athletic posture type；Obtain current identification
The corresponding amount of action of the highest at least one athletic posture type of rate；Wherein the discrimination is in convolutional neural networks
It is obtained in advance in training process；Calculate the corresponding amount of action of above-mentioned at least one athletic posture type and value, it is moved
Make total quantity and is shown.
In the specific application process, after mobile terminal judges the athletic posture type of current movement, the fortune is updated
The dynamic corresponding amount of action of posture type, such as when handheld motion equipment is table tennis bat, and judge current athletic posture
Type is forehand off-table chop, then the corresponding amount of action of forehand off-table chop adds 1.Then according to the corresponding identification of each athletic posture type
Rate determines movement total quantity, such as chooses first three highest corresponding amount of action of athletic posture type of discrimination, it is assumed that first three
The amount of action of a athletic posture type be respectively as follows: the amount of action of forehand off-table chop be 35, the amount of action of deep backhand chop is 40,
The amount of action smashed after drawing is 21, then the movement total quantity that mobile terminal is finally shown is 35+40+21=96.
In addition, the above method further include: obtain and show current kinematic parameter, the kinematic parameter include movement duration,
The corresponding amount of action of athletic posture type, energy consumption values and maximum movement speed.
A length of user opens corresponding software when wherein moving, and recognizes initial time and the current time of first element
Difference.The corresponding amount of action of athletic posture type may refer to the description of above-described embodiment, and maximum movement speed can lead to
Following formula is crossed to calculate:
Wherein aiIndicate 3-axis acceleration information a at ith sample pointix、aiy、aizConjunction
At acceleration, can be expressed asv0Initial velocity, k are corresponding sampled point sequence when obtaining peak acceleration
Number, Δ t is sampling time interval.
The acquisition process of above-mentioned energy consumption values includes: the personal information for obtaining user belonging to first action data
And current movement duration；According to the personal information, movement duration and the corresponding amount of action of each athletic posture type, determine
The energy consumption values of the user.
Wherein the personal information of user can be obtained with the registration information of user, and user can be logged in corresponding by mobile terminal
Software, carry out the registration of personal information, such as user name, password, the pet name, age, address, height, weight, gender.
In a possible embodiment, it when above-mentioned handheld motion equipment is table tennis bat, can be calculated according to following formula
The energy consumption values of user:
Table tennis energy consumption values=4.0 (exercise intensity MET) × run duration × weight (unit is kilogram) *
0.0167。
It, can be there are many training mode, in different training modes in order to better meet the diversified demand of user
Under different functions may be implemented.Such as it is segmented into real-time training mode and individual event training mode.
Based on this, in a possible embodiment, the above method further include: obtain the training mode that user currently selects；When
When the training mode is individual event training mode, target athletic posture type corresponding to the individual event training mode is determined；It obtains current
Movement total quantity and the current corresponding target action quantity of target athletic posture type；According to the target action quantity and movement
Total quantity determines the movement accuracy rate of user.
For example, user selects training mode for individual event training mode by mobile terminal, and select to need the movement of training
For deep backhand chop.The data that mobile terminal acquires in real time according to Posture acquisition device determine anti-in the way of above-described embodiment description
The target action quantity of hand off-table chop is 99, current movement total quantity the sum of (various athletic posture types corresponding amount of action)
It is 110, it is determined that user is when carrying out the action training of deep backhand chop, accuracy rate 99/110.
Selection function key can also be acted in the display interface setting of mobile terminal under individual event training mode, can arranged
Unrestricted choice on table.And while showing the movement accuracy rate of user, movement (the target movement trained can also be shown
The corresponding movement of posture type) amount of action, movement duration, energy consumption and it is maximum clap fast (i.e. maximum movement speed), with
It is effectively trained conducive to user, improves play technology.Key is instructed in addition, being also provided under individual event training mode, is passed through a little
Detailed play recommended information can be checked by hitting the key, which can intercept auto-correlation books, and such as " table tennis quick start is complete
Journey diagram ".
Under real-time training mode, can also show in the display interface of mobile terminal: movement total degree moves duration, energy
Amount consumption, maximum clap speed, and (smash, forehand are remote after the corresponding drawing of such as table tennis bat for the athletic posture type that can be accurately identified
Cut, deep backhand chop, forehand attack and other kinds of movement) and corresponding amount of action.
In a further embodiment, in order to promote the mutual exchange between user, the display interface of mobile terminal is also set up
There is community's button, when detecting that user clicks community's button, shows Community Page.User can be by clicking Community Page
Share button and the motion information (such as movement duration, energy consumption, maximum movement speed) of oneself is sent to cloud service platform, and
By cloud service platform share into the Community Page of each user, can also in community with other people exchange and interdynamic, such as thumb up,
Comment etc..
In embodiments of the present invention, pass through Classification and Identification mould of the Tensorflow frame training based on convolutional neural networks
Type, and the model is deployed to mobile terminal (such as Android mobile phone), it is each to realize to comprehensively utilize the recognition result of model output
Item function.In action recognition field, in the prior art by being configured threshold value, sliding window to the time series data that posture senses
Segmentation, it is artificial to extract characteristic value and a series of Supervised classification.And selection is using based on convolutional Neural in the embodiment of the present invention
The model of network is identified.Because convolutional neural networks can be avoided using original signal directly as the input of network
Complicated characteristic extraction procedure, while there is preferable robustness.In addition, for being difficult to early period manually obtain a large amount of different samples
Sample action data the problem of, the present embodiment combines cloud service platform, realizes the remote real-time synchronous of data.With user
The increase of amount, developer can log in rear end and obtain a large amount of sample action data, to provide big data money for training pattern
Source.
In mobile terminal, corresponding application software is developed, above-mentioned model is deployed in mobile terminal, to realize that the present invention mentions
The athletic posture recognition methods of confession.Referring to Fig. 5, user logs in application software, which is provided with moving interface and community
Interface.In moving interface, mobile terminal is connect by bluetooth with attitude transducer, is accelerated with obtaining three axis of handheld motion equipment
Information is spent, then these information are carried out with data cutout, data cleansing and the integration of effective action, as convolutional neural networks mould
The input of type, the model will return to the discrimination of marked movement each early period.While in order to provide friendly human-computer interaction
Interface, the embodiment of the present invention is compared each action recognition rate by programming, and counts to the highest movement of discrimination
It counts and shows the motion state for facilitating user to understand itself on the display interface of the mobile terminal.In addition, may be used also in display interface
With relative motions indexs such as display movement duration, speed of swing (i.e. maximum movement speed), kinergety consumption, in order to promote to use
Family exchange increase exchange between communities function, community interface share dynamic, check dynamic and comment on thumb up, greatly meet user
Diversified demand.
Embodiment two:
It is directed to a kind of athletic posture recognition methods of above-described embodiment description, the embodiment of the invention also provides a kind of fortune
Dynamic gesture recognition device, the device are applied to mobile terminal, and referring to Fig. 6, which includes:
Data acquisition module 11, for obtaining the first movement number of the handheld motion instrument that attitude transducer currently acquires
According to；Wherein, which is set on the handheld motion instrument, which includes acceleration information；
Data determining module 12, for based on the second movement acquired before above-mentioned first action data and preset time period
Data determine effective action data using difference value-based algorithm；
Dimension processing module 13 is gone, is handled for carrying out dimension to above-mentioned effective action data, obtains standard operation number
According to；
Classification of motion module 14, for above-mentioned standard action data is input in advance trained convolutional neural networks into
The row classification of motion obtains the athletic posture type of current handheld sports apparatus.
In embodiments of the present invention, data, mobile terminal pair are acquired by the attitude transducer in handheld motion instrument
Collected data carry out data extraction, are identified to the valid data extracted by convolutional neural networks, so as to
It more accurately identifies that the athletic posture type of handheld motion instrument effectively reduces cost without numerous peripheral apparatus, opens up
Wide application range, promotes the popularization and use of product.
Embodiment three:
Referring to Fig. 7, the embodiment of the present invention also provides a kind of electronic equipment 100, comprising: processor 40, memory 41, bus
42 and communication interface 43, the processor 40, communication interface 43 and memory 41 are connected by bus 42；Processor 40 is for holding
The executable module stored in line storage 41, such as computer program.
Wherein, memory 41 may include high-speed random access memory (RAM, Random Access Memory),
It may further include nonvolatile memory (non-volatile memory), for example, at least a magnetic disk storage.By at least
One communication interface 43 (can be wired or wireless) realizes the communication between the system network element and at least one other network element
Connection, can be used internet, wide area network, local network, Metropolitan Area Network (MAN) etc..
Bus 42 can be isa bus, pci bus or eisa bus etc..The bus can be divided into address bus, data
Bus, control bus etc..Only to be indicated with a four-headed arrow convenient for indicating, in Fig. 7, it is not intended that an only bus or
A type of bus.
Wherein, memory 41 is for storing program, and the processor 40 executes the journey after receiving and executing instruction
Sequence, method performed by the device that the stream process that aforementioned any embodiment of the embodiment of the present invention discloses defines can be applied to handle
In device 40, or realized by processor 40.
Processor 40 may be a kind of IC chip, the processing capacity with signal.During realization, above-mentioned side
Each step of method can be completed by the integrated logic circuit of the hardware in processor 40 or the instruction of software form.Above-mentioned
Processor 40 can be general processor, including central processing unit (Central Processing Unit, abbreviation CPU), network
Processor (Network Processor, abbreviation NP) etc.；It can also be digital signal processor (Digital Signal
Processing, abbreviation DSP), specific integrated circuit (Application Specific Integrated Circuit, referred to as
ASIC), ready-made programmable gate array (Field-Programmable Gate Array, abbreviation FPGA) or other are programmable
Logical device, discrete gate or transistor logic, discrete hardware components.It may be implemented or execute in the embodiment of the present invention
Disclosed each method, step and logic diagram.General processor can be microprocessor or the processor is also possible to appoint
What conventional processor etc..The step of method in conjunction with disclosed in the embodiment of the present invention, can be embodied directly in hardware decoding processing
Device executes completion, or in decoding processor hardware and software module combination execute completion.Software module can be located at
Machine memory, flash memory, read-only memory, programmable read only memory or electrically erasable programmable memory, register etc. are originally
In the storage medium of field maturation.The storage medium is located at memory 41, and processor 40 reads the information in memory 41, in conjunction with
Its hardware completes the step of above method.
Discount coupon provided in an embodiment of the present invention checks and writes off device and electronic equipment, with discount coupon core provided by the above embodiment
Pin method technical characteristic having the same reaches identical technical effect so also can solve identical technical problem.
Discount coupon is carried out provided by the embodiment of the present invention and checks and writes off the computer program product of method, including stores processing
The computer readable storage medium of the executable non-volatile program code of device, the instruction that said program code includes can be used for holding
Row previous methods method as described in the examples, specific implementation can be found in embodiment of the method, and details are not described herein.
It is apparent to those skilled in the art that for convenience and simplicity of description, the device of foregoing description
And the specific work process of electronic equipment, it can refer to corresponding processes in the foregoing method embodiment, details are not described herein.
The flow chart and block diagram in the drawings show multiple embodiment method and computer program products according to the present invention
Architecture, function and operation in the cards.In this regard, each box in flowchart or block diagram can represent one
A part of module, section or code, a part of the module, section or code include it is one or more for realizing
The executable instruction of defined logic function.It should also be noted that in some implementations as replacements, function marked in the box
It can also can occur in a different order than that indicated in the drawings.For example, two continuous boxes can actually be substantially parallel
Ground executes, they can also be executed in the opposite order sometimes, and this depends on the function involved.It is also noted that block diagram
And/or the combination of each box in flow chart and the box in block diagram and or flow chart, it can the function as defined in executing
Can or the dedicated hardware based system of movement realize, or can come using a combination of dedicated hardware and computer instructions real
It is existing.
In the description of the present invention, it should be noted that term " center ", "upper", "lower", "left", "right", "vertical",
The orientation or positional relationship of the instructions such as "horizontal", "inner", "outside" be based on the orientation or positional relationship shown in the drawings, merely to
Convenient for description the present invention and simplify description, rather than the device or element of indication or suggestion meaning must have a particular orientation,
It is constructed and operated in a specific orientation, therefore is not considered as limiting the invention.In addition, term " first ", " second ",
" third " is used for descriptive purposes only and cannot be understood as indicating or suggesting relative importance.Unless specifically stated otherwise, otherwise exist
Component described in these embodiments and opposite step, numerical expression and the numerical value of step are not limit the scope of the invention.
In several embodiments provided herein, it should be understood that disclosed systems, devices and methods, it can be with
It realizes by another way.The apparatus embodiments described above are merely exemplary, for example, the division of the unit,
Only a kind of logical function partition, there may be another division manner in actual implementation, in another example, multiple units or components can
To combine or be desirably integrated into another system, or some features can be ignored or not executed.Another point, it is shown or beg for
The mutual coupling, direct-coupling or communication connection of opinion can be through some communication interfaces, device or unit it is indirect
Coupling or communication connection can be electrical property, mechanical or other forms.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple
In network unit.It can select some or all of unit therein according to the actual needs to realize the mesh of this embodiment scheme
's.
It, can also be in addition, the functional units in various embodiments of the present invention may be integrated into one processing unit
It is that each unit physically exists alone, can also be integrated in one unit with two or more units.
It, can be with if the function is realized in the form of SFU software functional unit and when sold or used as an independent product
It is stored in the executable non-volatile computer-readable storage medium of a processor.Based on this understanding, of the invention
Technical solution substantially the part of the part that contributes to existing technology or the technical solution can be with software in other words
The form of product embodies, which is stored in a storage medium, including some instructions use so that
One computer equipment (can be personal computer, server or the network equipment etc.) executes each embodiment institute of the present invention
State all or part of the steps of method.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory (ROM, Read-
Only Memory), random access memory (RAM, Random Access Memory), magnetic or disk etc. are various can be with
Store the medium of program code.
Finally, it should be noted that embodiment described above, only a specific embodiment of the invention, to illustrate the present invention
Technical solution, rather than its limitations, scope of protection of the present invention is not limited thereto, although with reference to the foregoing embodiments to this hair
It is bright to be described in detail, those skilled in the art should understand that: anyone skilled in the art
In the technical scope disclosed by the present invention, it can still modify to technical solution documented by previous embodiment or can be light
It is readily conceivable that variation or equivalent replacement of some of the technical features；And these modifications, variation or replacement, do not make
The essence of corresponding technical solution is detached from the spirit and scope of technical solution of the embodiment of the present invention, should all cover in protection of the invention
Within the scope of.Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. a kind of athletic posture recognition methods, which is characterized in that be applied to mobile terminal, comprising:
Obtain the first action data of the handheld motion instrument that attitude transducer currently acquires；Wherein, the attitude transducer is set
It is placed on the handheld motion instrument, first action data includes acceleration information；
Based on the second action data acquired before first action data and preset time period, using difference value-based algorithm, really
Determine effective action data；
Dimension is carried out to the effective action data to handle, and obtains standard operation data；
The standard operation data are input in convolutional neural networks trained in advance and carry out the classification of motion, obtained presently described
The athletic posture type of handheld motion instrument.
2. the method according to claim 1, wherein before based on first action data and preset time period
Second action data of acquisition, using difference value-based algorithm, the step of determining effective action data, comprising:
Calculate the difference between first action data and second action data；
The difference and predetermined threshold value are compared, the initial data and endpoint data of valid window is determined, obtains
Current effective action data.
3. the method according to claim 1, wherein the training process of the convolutional neural networks includes:
The sample action data of each athletic posture type are obtained, and mark the athletic posture type of the sample action data；
Sample valid data are determined using difference value-based algorithm based on the sample action data after mark；
Dimension is carried out to the sample valid data to handle, and obtains master sample data；
The master sample data are input to convolutional neural networks, the convolutional neural networks are trained, are trained
Convolutional neural networks afterwards.
4. the method according to claim 1, wherein further include:
The current athletic posture type of the handheld motion instrument is obtained, and it is corresponding to update the current athletic posture type
Amount of action；
Obtain the corresponding amount of action of the highest at least one athletic posture type of current discrimination；The wherein discrimination
It is to be obtained in advance in the training process of the convolutional neural networks；
Calculate the corresponding amount of action of at least one athletic posture type and value, it obtains movement total quantity and is shown
Show.
5. according to the method described in claim 4, it is characterized by further comprising:
Obtain the training mode that user currently selects；
When the training mode is individual event training mode, target athletic posture class corresponding to the individual event training mode is determined
Type；
Obtain current movement total quantity and the target athletic posture type currently corresponding target action quantity；
According to the target action quantity and the movement total quantity, the movement accuracy rate of the user is determined.
6. the method according to claim 1, wherein further include:
Obtain and show that current kinematic parameter, the kinematic parameter are corresponding including movement duration, the athletic posture type
Amount of action, energy consumption values and maximum movement speed.
7. according to the method described in claim 6, it is characterized in that, the acquisition process of the energy consumption values includes:
Obtain the personal information and current movement duration of user belonging to first action data；
According to the personal information, movement duration and the corresponding amount of action of each athletic posture type, the use is determined
The energy consumption values at family.
8. the method according to claim 1, wherein the handheld motion instrument is table tennis bat.
9. a kind of athletic posture identification device, which is characterized in that be applied to mobile terminal, comprising:
Data acquisition module, for obtaining the first action data of the handheld motion instrument that attitude transducer currently acquires；Wherein,
The attitude transducer is set on the handheld motion instrument, and first action data includes acceleration information；
Data determining module, the second action data for being acquired before based on first action data and preset time period,
Using difference value-based algorithm, effective action data are determined；
Dimension processing module is gone, is handled for carrying out dimension to the effective action data, obtains standard operation data；
Classification of motion module is acted for the standard operation data to be input in convolutional neural networks trained in advance
Classification, obtains the athletic posture type of presently described handheld motion instrument.
10. a kind of electronic equipment, including memory, processor, it is stored with and can runs on the processor on the memory
Computer program, which is characterized in that the processor realizes the claims 1 to 8 when executing the computer program
Method described in one.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910060946.XA CN109731302A (en) | 2019-01-22 | 2019-01-22 | Athletic posture recognition methods, device and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910060946.XA CN109731302A (en) | 2019-01-22 | 2019-01-22 | Athletic posture recognition methods, device and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109731302A true CN109731302A (en) | 2019-05-10 |
Family
ID=66365625
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910060946.XA Pending CN109731302A (en) | 2019-01-22 | 2019-01-22 | Athletic posture recognition methods, device and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109731302A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110222634A (en) * | 2019-06-04 | 2019-09-10 | 河海大学常州校区 | A kind of human posture recognition method based on convolutional neural networks |
| CN110879662A (en) * | 2019-11-27 | 2020-03-13 | 云南电网有限责任公司电力科学研究院 | Action recognition device and method based on AHRS algorithm |
| CN112295206A (en) * | 2020-10-29 | 2021-02-02 | 成都方德尔科技有限公司 | Virtual presentation system |
| CN112446407A (en) * | 2019-09-03 | 2021-03-05 | 财团法人资讯工业策进会 | Motion data tagging system, method and non-transitory computer readable medium |
| CN114167984A (en) * | 2021-01-28 | 2022-03-11 | Oppo广东移动通信有限公司 | Device control method, device, storage medium and electronic device |
| CN114259162A (en) * | 2021-12-10 | 2022-04-01 | 麒盛科技股份有限公司 | Air bag pillow capable of identifying sleeping posture and control method thereof |
| CN114788951A (en) * | 2021-01-26 | 2022-07-26 | 王振兴 | Hand-held motion analysis system and method |
| CN115145846A (en) * | 2021-03-31 | 2022-10-04 | 广东高云半导体科技股份有限公司 | Artificial intelligence system and method for gesture recognition |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105929940A (en) * | 2016-04-13 | 2016-09-07 | 哈尔滨工业大学深圳研究生院 | Rapid three-dimensional dynamic gesture recognition method and system based on character value subdivision method |
| CN205721628U (en) * | 2016-04-13 | 2016-11-23 | 哈尔滨工业大学深圳研究生院 | A kind of quick three-dimensional dynamic hand gesture recognition system and gesture data collecting device |
| CN106563260A (en) * | 2016-10-28 | 2017-04-19 | 深圳职业技术学院 | Table tennis intelligent motion system based on attitude sensor and computing method based on table tennis intelligent motion system |
| CN108170274A (en) * | 2017-12-29 | 2018-06-15 | 南京邮电大学 | A kind of action identification method based on wearable device |
| CN109011505A (en) * | 2018-06-22 | 2018-12-18 | 华南理工大学 | A kind of recognition methods of low power consumption high-precision table tennis and device |
-
2019
- 2019-01-22 CN CN201910060946.XA patent/CN109731302A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105929940A (en) * | 2016-04-13 | 2016-09-07 | 哈尔滨工业大学深圳研究生院 | Rapid three-dimensional dynamic gesture recognition method and system based on character value subdivision method |
| CN205721628U (en) * | 2016-04-13 | 2016-11-23 | 哈尔滨工业大学深圳研究生院 | A kind of quick three-dimensional dynamic hand gesture recognition system and gesture data collecting device |
| CN106563260A (en) * | 2016-10-28 | 2017-04-19 | 深圳职业技术学院 | Table tennis intelligent motion system based on attitude sensor and computing method based on table tennis intelligent motion system |
| CN108170274A (en) * | 2017-12-29 | 2018-06-15 | 南京邮电大学 | A kind of action identification method based on wearable device |
| CN109011505A (en) * | 2018-06-22 | 2018-12-18 | 华南理工大学 | A kind of recognition methods of low power consumption high-precision table tennis and device |
Non-Patent Citations (1)
| Title |
|---|
| 石代伟,张若英: "结合手机传感器和卷积神经网络的人体行为识别", 《电子技术与软件工程》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110222634A (en) * | 2019-06-04 | 2019-09-10 | 河海大学常州校区 | A kind of human posture recognition method based on convolutional neural networks |
| CN110222634B (en) * | 2019-06-04 | 2022-11-01 | 河海大学常州校区 | Human body posture recognition method based on convolutional neural network |
| CN112446407A (en) * | 2019-09-03 | 2021-03-05 | 财团法人资讯工业策进会 | Motion data tagging system, method and non-transitory computer readable medium |
| CN110879662A (en) * | 2019-11-27 | 2020-03-13 | 云南电网有限责任公司电力科学研究院 | Action recognition device and method based on AHRS algorithm |
| CN112295206A (en) * | 2020-10-29 | 2021-02-02 | 成都方德尔科技有限公司 | Virtual presentation system |
| CN114788951A (en) * | 2021-01-26 | 2022-07-26 | 王振兴 | Hand-held motion analysis system and method |
| CN114788951B (en) * | 2021-01-26 | 2024-02-20 | 王振兴 | Handheld motion analysis system and method |
| CN114167984A (en) * | 2021-01-28 | 2022-03-11 | Oppo广东移动通信有限公司 | Device control method, device, storage medium and electronic device |
| CN114167984B (en) * | 2021-01-28 | 2024-03-12 | Oppo广东移动通信有限公司 | Equipment control method and device, storage medium and electronic equipment |
| CN115145846A (en) * | 2021-03-31 | 2022-10-04 | 广东高云半导体科技股份有限公司 | Artificial intelligence system and method for gesture recognition |
| CN114259162A (en) * | 2021-12-10 | 2022-04-01 | 麒盛科技股份有限公司 | Air bag pillow capable of identifying sleeping posture and control method thereof |
| CN114259162B (en) * | 2021-12-10 | 2023-12-05 | 麒盛科技股份有限公司 | Airbag pillow capable of recognizing sleeping posture and control method thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109731302A (en) | Athletic posture recognition methods, device and electronic equipment | |
| Zhu et al. | Feature selective anchor-free module for single-shot object detection | |
| CN108197589B (en) | Semantic understanding method, apparatus, equipment and the storage medium of dynamic human body posture | |
| Li et al. | An intelligent optimization method of motion management system based on BP neural network | |
| CN110503077B (en) | Real-time human body action analysis method based on vision | |
| Host et al. | An overview of Human Action Recognition in sports based on Computer Vision | |
| CN108229277A (en) | Gesture identification, control and neural network training method, device and electronic equipment | |
| Zhang et al. | Fast collective activity recognition under weak supervision | |
| CN107851195A (en) | Target detection is carried out using neutral net | |
| US11819734B2 (en) | Video-based motion counting and analysis systems and methods for virtual fitness application | |
| CN112827168B (en) | Target tracking method, device and storage medium | |
| CN113743273B (en) | Real-time rope skipping counting method, device and equipment based on video image target detection | |
| CN109725699A (en) | Recognition methods, device and the equipment of identification code | |
| CN109740567A (en) | Key point location model training method, localization method, device and equipment | |
| CN113822254B (en) | Model training method and related device | |
| Yang et al. | Release the power of online-training for robust visual tracking | |
| US11450010B2 (en) | Repetition counting and classification of movements systems and methods | |
| Ahmad et al. | Optimized deep learning-based cricket activity focused network and medium scale benchmark | |
| Xiao et al. | Recognizing sports activities from video frames using deformable convolution and adaptive multiscale features | |
| Kokul et al. | Gate connected convolutional neural network for object tracking | |
| CN109857878A (en) | Article mask method and device, electronic equipment and storage medium | |
| CN110472537B (en) | Self-adaptive identification method, device, equipment and medium | |
| Ou et al. | Multi-scale spatialtemporal information deep fusion network with temporal pyramid mechanism for video action recognition | |
| CN107851457A (en) | It is configured as realizing the framework and method of the analysis of the technical ability to physical performance for the transmission for including being applied to interaction skill training content | |
| Zhou | Computational Analysis of Table Tennis Games from Real-Time Videos Using Deep Learning |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190510 |