CN110502107A - Wearable real-time action instructs system and method - Google Patents
Wearable real-time action instructs system and method Download PDFInfo
- Publication number
- CN110502107A CN110502107A CN201910683451.2A CN201910683451A CN110502107A CN 110502107 A CN110502107 A CN 110502107A CN 201910683451 A CN201910683451 A CN 201910683451A CN 110502107 A CN110502107 A CN 110502107A
- Authority
- CN
- China
- Prior art keywords
- real
- data
- sensor
- time
- wearable
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06F18/24155—Bayesian classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/017—Gesture based interaction, e.g. based on a set of recognized hand gestures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- 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
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
- G16H20/30—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to physical therapies or activities, e.g. physiotherapy, acupressure or exercising
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2203/00—Indexing scheme relating to G06F3/00 - G06F3/048
- G06F2203/01—Indexing scheme relating to G06F3/01
- G06F2203/011—Emotion or mood input determined on the basis of sensed human body parameters such as pulse, heart rate or beat, temperature of skin, facial expressions, iris, voice pitch, brain activity patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2203/00—Indexing scheme relating to G06F3/00 - G06F3/048
- G06F2203/01—Indexing scheme relating to G06F3/01
- G06F2203/012—Walk-in-place systems for allowing a user to walk in a virtual environment while constraining him to a given position in the physical environment
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Human Computer Interaction (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physical Education & Sports Medicine (AREA)
- Probability & Statistics with Applications (AREA)
- Public Health (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Multimedia (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses a kind of wearable real-time actions to instruct system and method, wherein, the system uses distributed artificial intelligence framework, by hardware (sensor), cell phone application (part, in real time) and Cloud Server is (global, it forms afterwards), local intelligence Real-time Feedback and global intelligence is constantly updated into study and combined, computing capability is pushed to hardware and mobile phone by cloud.The invention has the beneficial effects that: wearable real-time action provided by the invention instructs system that can realize the comprehensive real-time acquisition of various information, the Intelligent treatment of motion information in motion process and targetedly carry out real-time action director, it is also greatly reduced data delay simultaneously, also gets rid of the dependence to cloud and network.
Description
Technical field
The present invention relates to a kind of action director's system and method, and in particular to a kind of wearable real-time action instruct system and
Method belongs to field of computer technology.
Background technique
Increase with people to healthy attention degree, sports have been increasingly becoming the life side of more and more people
Formula, but sports fan is if be easy to body without scientific, reasonable, complete training guidance plan and if blindly taking exercise
Body causes biggish damage.Therefore, the correctness for how rationally scientifically evaluating athletic performance, improves the accurate of sporter's training
Property and science, and targetedly provide training guidance plan have become the key subject there is an urgent need to research and development.
In exercise and training, movement is correct most important, and the movement of mistake often becomes the killer of health, not only rises
Less than exercise effect, cause somatic damage instead.For example one seem the movement of simple deep-knee-bend, if movement is incorrect, can hurt
Waist or knee, especially in the case where body bears weights, this injury can be more serious.
In rehabilitation and rehabilitation training, correctly movement is also critically important.Rehabilitation training is not intended to the stir yourself of elephant
Arm draws leg so simple, it is necessary under the guidance of the professionals such as physiatrician, rehabilitation therapist, according to each patient
Different situations, targetedly formulate therapeutic scheme, and the controlling by formulation by professionals such as physiatrician, rehabilitation therapists
Treatment scheme being trained step by step specifically can be as accurate as training of each piece of muscle, each movement, each training
It is not arbitrarily, if not being trained by therapeutic scheme, it is likely that can go wrong.Such as in the training of hemiplegia stroke rehabilitation
In, there is the problems such as shoulder subluxation, shoulder arthralgia, shoulder-hand syndrome in a lot of patients, and here it is not according to rehabilitation
Caused by the requirement of the professionals such as doctor, rehabilitation therapist is done.So rehabilitation not act on one's own, it is self-righteous into
Row will be done according to correct guidance.
Wearable device directly wears or is integrated into one kind on the clothes or accessory of user and portable sets
It is standby.Wearable device is not only a kind of hardware device, can be by software support and data interaction, cloud interaction come real
Existing powerful function, wearable device will life, perception to us bring very big transformation.
Wearable device have developed rapidly in recent years, be the consumer technological product renovated with most fast speed.Currently, can wear
The product form for wearing equipment mainly has Intelligent bracelet, smartwatch, intelligent glasses etc..There is representative in this few class products whole world
Company, such as: Intelligent bracelet occupies most of share of world market using Fitbit, China meter Yun Dong and Garmin as representative；
Smartwatch occupies more than half the market share of the whole world using Apple Watch as representative；Intelligent glasses be then at present with
The HoloLens of Google Glass and Microsoft are representative, and Apple is also in the intelligent glasses for researching and developing oneself.Intelligent glasses
Practical application at present in movement is also seldom.Many sports items all begin trying wearable technology, such as: NFL, NBA,
NHL, MLB, football, tennis, even college football.The development and innovation of wearable technology, by the thorough cloth for changing sporting world
Office.
Wearable technology can collect information, improve bean vermicelli participation, and reduction sportsman is sick and wounded, optimizes training system.But
For sports, these Intelligent bracelets and smartwatch can only provide very limited and part information at present.
The development of wearable product is unsatisfactory, presently mainly based on heart rate and step counting, and heart rate when movement
The current mark post product Apple Watch that connects fails to survey standard, so wearable product can be described as data acquisition inaccurately now,
Also, the problem of most critical, is, user only sees data, and effect is limited, and user needs to know how to do, and how to correct, most
It can find the problem in time well, real-time correction.
In many fields, such as in sports field and health care rehabilitation field, human body and other objects are captured
The movement of (animal, plant or any non-living body equipment) (can be the movement of object as a whole, be also possible to one of object
Point movement, can also be the combination of both) be it is vital, once capture movement, so that it may determine different movements
Characteristic, such as time location, speed, acceleration, distance, time, spinning rate.
Currently, the capture system for being used to capture the movement of human body and other objects is broadly divided into two major classes type: optics movement
Capture system, inertial sensor motion capture system.
1, optics motion capture system
In optics motion capture system, object pastes one or more optical markings objects in predetermined position, by one or
Multiple video cameras record the position coordinates of these optical markings objects simultaneously, then all with the method identification and tracking of image procossing
Optical markings object rebuilds the movement of object.The VICON optics motion capture system of Britain and domestic love victory motion image solution
Analysis system, the two are all the different parts in body, such as head, body, arm and leg, stick optical markings object, then catch
The movement that a people moves before video camera is caught, extracts optical markings object with identification and track algorithm data processing means later
Coordinate data, to provide motion characteristics.
In optics motion capture system, when optical markings object is blocked or is moved out of the video camera for recording its movement
When, video camera will be unable to tracking optical markings object, and motion profile will become imperfect, will go wrong.This is solved to ask
One of topic possible solution: using multiple video cameras, but is intended to be fully solved this problem, expense will be held high very much
It is expensive, while the complexity of motion capture system is also added, in addition, existing motion capture system is not portable, three dimension system
It needs multiple cameras while connecting computer, can only be wired connection, video analysis system video is not handled simultaneously.
Certainly, the optics movement acquisition equipment with depth perception of low price, such as Kinect, Leap are had recently emerged
Motion has caused body-sensing interaction upsurge, so that body-sensing interaction enters mainstream.This kind of optical acquisition device mainly has sense of depth
Know, use structure light, flight time or binocular vision technology based on optical profile type capturing technology, proposes complete movement
Capture, processing and virtual acting reconfiguration scheme.But these low price optics movement capture device generally all by use scope,
There is the limitation of unobstructed, accuracy etc., a fixed place can only be erected at as the peripheral hardware of computer, can only be made indoors
With cannot use outdoors, daily can not wear the activity data for acquiring true life and work.
As it can be seen that existing optics motion capture system use is excessively complicated, expensive, and fortune can only be collected indoors
Dynamic data, can not routine use.
2, inertial sensor motion capture system
In inertial sensor motion capture system, motion sensor is by bandage or pastes fixed on object or being placed on object
Inside body, motion sensor can provide the acceleration signal for representing three mutually orthogonal direction X, Y of different directions, Z-direction, top
Spiral shell instrument signal is used to measure motion sensor around the rotation speed of X, Y, Z axis and three-dimensional magnetometer.
Using motion sensor captures object movement when, can not be quasi- if lacking reference or calibrating position for a long time
Motion sensor position is determined, then motion capture will go wrong.Even if the initial position to motion sensor carries out
Calibration, in motion sensor moving process, very big error will soon occur in position and direction, so that motion-sensing
The exercise data of device becomes unreliable.
Xsens company, the U.S., which opens, is proposed a double capture system based on inertial sensor, and Beijing promise is also risen
Company is also proposed a double recently and captures product.This two product has the shortcomings that common:
(1) use multiple sensors, whole body is a set of to work together, occupy fixed space, it is not portable, can not routine use,
And it is expensive；
(2) there was only motion capture and operation of recording, based on data collection, without real-time motion analysis, can not to
The posture at family and movement provide feedback in real time；
(3) WIFI is used, power consumption is high, needs to carry on the back a large size battery and transmitting base station, influences to move, and is not suitable for making outdoors
With；
(4) angle is calculated with magnetometer, anti-interference is very poor, and slightly bigger metal will lead to hand leg and float
Ease, also will appear various displacement differences over time；
It (5) must be with PC work.
ZEPP company, the U.S. is proposed a motion capture system using single sensor, and what is collected is single-point number
According to, such as: shank is connected the sensor to, for recording club speed after sportsman swings, club face, rhythm, backward
It swings distance etc., or connects the sensor on tennis racket, swing the bat the exercise data of front and back tennis racket for recording sportsman,
But this motion capture system can be only applied in equipment, can not accurately capture and track the movement of human body itself.
Summary of the invention
To solve the deficiencies in the prior art, the purpose of the present invention is to provide various information in a kind of achievable motion process
Comprehensive real-time acquisition, the Intelligent treatment of motion information and the wearable reality for targetedly carrying out real-time action director
When action director's system and method.
In order to achieve the above objectives, the present invention adopts the following technical scheme that:
A kind of wearable real-time action instructs system, which is characterized in that use distributed artificial intelligence framework, by hardware,
Local intelligence Real-time Feedback and global intelligence are constantly updated study and combined, meter by cell phone application and Cloud Server composition
Calculation ability pushes hardware and mobile phone to by cloud, in which:
Aforementioned hardware is the external member for the easy wearing being made of multiple sensor nodes, and multiple sensor nodes collect number
According to as a result calculating in real time and fusion pass through Bluetooth transmission to cell phone application；
Aforementioned cell phone application displays and saves the data received from multiple sensors, and utilize by Bluetooth control sensor
The data combination human anatomy of sensor and the principle of sport dynamics are constructed the valid model of description human motion and are therefrom mentioned
Motion feature is taken, when movement deviates with motion model, according to the position that movement lack of standardization occurs, to body part, fortune
Dynamic method carries out error analysis, provides feedback, and the guidance command corrected with sound to sportsman's proposition movement, data knot in real time
Fruit uploads to Cloud Server by network；
The data result that aforementioned Cloud Server long-time storage cell phone application uploads, and carry out machine learning and analysis.
Wearable real-time action above-mentioned instructs system, which is characterized in that needs to complete number inside foregoing sensor node
According to the work of analysis and processing comprising four modules: sensor module, processor module, communication module and power supply, in which:
Sensor module is made of sensing element and AD conversion unit, and sensing element is responsible for the motion information of test object,
AD conversion unit is responsible for motion information being transformed into electric signal；
Processor module is made of processor and memory, and processor is responsible for controlling the normal of other each modules of sensor node
Work, and the relevant treatment of each signal is carried out, memory is responsible for storing related data；
Communication module is made of network protocol and transceiver, is responsible for the transmission of signal, wirelessly by node data
Send other equipment to；
To sensor node, other each modules provide the energy operated normally to power supply.
Wearable real-time action above-mentioned instructs system, which is characterized in that aforementioned processor is using ARM Cortex-
M4 processor.
Wearable real-time action above-mentioned instructs system, which is characterized in that foregoing sensor node is passed using nine axis
Sensor.
Wearable real-time action above-mentioned instructs system, which is characterized in that aforementioned cell phone application can according to training goal and
The difference of movement technique acquires sensing data to different motive positions, constructs movement mould in conjunction with the theory of motion analysis engineering
Type realizes the electronic data of motion model, prepares for skill analysis.
Wearable real-time action above-mentioned instructs system, which is characterized in that including user, sports team, research institution
Personnel and organizations can by cell phone application apply and website shared data.
Wearable real-time action above-mentioned instructs system, which is characterized in that sensor as aforementioned and cell phone application with it is offline and
Internuncial mode of having a rest is run, and when they are offline or can only be connected intermittently to cloud, once reconnecting, they are just
Meeting its last state of automatic synchronization, and run with continuing Non-intermittent, irrespective of whether connecting.
Wearable real-time action above-mentioned instructs system, which is characterized in that sensor as aforementioned and cell phone application only send us
The data that need to store simultaneously are analyzed beyond the clouds.
Wearable real-time action above-mentioned instructs system, which is characterized in that aforementioned Cloud Server is Microsoft Azure cloud platform
Or private services device.
Wearable real-time action above-mentioned instructs system, which is characterized in that uses Microsoft Azure cloud platform, and uses
Azure machine learning carrys out the model and scheme of creation analysis data, with the side including binary Bayes, multivariate logistic regression
Method carries out classification and pattern-recognition, while assisting from cloud to the deployment of edge APP and the AI learning outcome of sensor.
The invention has the beneficial effects that:
(1) wearable real-time action provided by the invention instructs system globe area from the data of multiple sensors, often
A sensor is capable of real-time acquisition 9 axle exercise datas, based on these data amount and calculation amount, which being capable of the inspection of Comprehensive items
It surveys index and obtains more accurately and reliably result；
(2) wearable real-time action provided by the invention instructs system there was only data acquisition, can during practice
To be analyzed according to the data that acquire in real time these data, then based on the analysis results the movement of mistake remind and
It corrects, correct muscle memory is cultivated in help；
(3) wearable real-time action provided by the invention instructs system using international pioneering distributed artificial intelligence frame
Structure --- sensor+cell phone application (part, in real time)+Cloud Server (global, subsequent), not only by local intelligence Real-time Feedback and
Global intelligence is constantly updated study and is combined, in this way it is contemplated that the personalized and reasonable diversity of movement, rather than
Carry out rigid error correction according to the template of standard, and computing capability is pushed to sensor and mobile phone by cloud, greatly reduces data and prolong
Late, and dependence to cloud and network is got rid of；
(4) wearable real-time action provided by the invention instructs system that will carry out long-term on privately owned or publicly-owned server
Data accumulation and analysis can help the training and raising of intelligent algorithm, from the point of view of long-term trend, can really realize
Personalized comprehensive analysis.
Detailed description of the invention
Fig. 1 is the schematic diagram of distributed artificial intelligence framework；
Fig. 2 is the composition schematic diagram of sensor node；
Fig. 3 is the placement schematic diagram of three inertial sensors when calculating the joint angles and acceleration of human motion；
Fig. 4 is the flow chart that joint angles are calculated using inertial sensor direction estimation value；
Fig. 5 is the flow chart of the human body attitude identification based on inertial sensor；
Fig. 6 is the conditional probability table of each feature in simple Bayesian network；.
Specific embodiment
Specific introduce is made to the present invention below in conjunction with the drawings and specific embodiments.
Referring to Fig.1, wearable real-time action provided by the invention instructs system using distributed artificial intelligence framework, by hard
Local intelligence Real-time Feedback and global intelligence are constantly updated study and combined by part, cell phone application and Cloud Server composition,
Computing capability is pushed to hardware and mobile phone by cloud.
1, hardware
By the external member for the easy wearing that multiple sensor nodes form.Multiple sensor nodes collect data, calculate in real time
And fusion, as a result pass through Bluetooth transmission to cell phone application.
Need to complete the work of data analysis and process inside sensor node, so the sensor node packet that we design
Containing multiple modules, modules are used cooperatively the work requirements that could complete system, it is generally the case that referring to Fig. 2, a sensing
Device node includes four modules: sensor module, processor module, communication module and power supply, in which:
(1) sensor module: being made of sensing element and AD conversion unit, and sensing element is responsible for the movement letter of test object
Breath, AD conversion unit are responsible for motion information being transformed into electric signal；
(2) processor module: being made of ARM Cortex-M4 processor and memory, and ARM Cortex-M4 processor is responsible for
The normal work of other each modules of sensor node is controlled, and carries out the relevant treatment of each signal, memory is responsible for storing dependency number
According to；
(3) communication module: being made of network protocol and transceiver, is responsible for the transmission of signal, node data is passed through wireless
Mode sends other equipment to；
(4) power supply: to sensor node, other each modules provide the energy operated normally.
Sensor (sensor) is a kind of detection device, can experience measured information, and can be by information by a set pattern
Rule is converted into electric signal, is transmitted, handled, stored, show, recorded.
Nine axle sensors are the combinations of three kinds of sensors: 3 axis acceleration sensors, 3 axis gyroscopes and 3 axis electronic compass (
Magnetic Sensor).These three partial actions are different, cooperate, are in the electronic products such as our mobile phone, tablet computer, game machines
Common motion sensing tracks element, applied to the interactive controlling in all kinds of softwares, game.
Acceleration sensor is all directions acceleration in measurement space, it utilizes the inertia of one " gravity block ", sensor
When movement, " gravity block " can generate pressure to X, Y, Z-direction (front and rear, left and right, up and down), recycle a kind of piezo-electric crystal
This pressure conversion at electric signal, with the variation of movement, all directions pressure is different, and electric signal is also changing, to judge
The acceleration direction and velocity magnitude of equipment.
Gyroscope is a kind of for measurement angle and the equipment for maintaining direction, in flying games, sport game and the
In the game such as one visual angle class shooting, the displacement of player hand can be completely monitored, to realize various game operation effects.Gyro
Instrument is as acceleration sensor and a microelectronic element, and using Coriolis force, the oscillator by continuous vibration is rotating
Motion excursion in system changes circuit state, causes the variation of related electrical parameters, so as to reflect equipment lateral tilting
Tiltedly, the motion conditions such as tilt forward and back and be swung left and right.
Using acceleration sensor and gyroscope, the entire motion state of equipment can be described substantially, but with for a long time
Movement, acceleration sensor and gyroscope can all generate Accumulated deviation, be unable to accurate description athletic posture, for example operation control picture occurs
Inclination.
Electronic compass (geomagnetic sensor) is modified compensation by absolute direction-pointing function using measurement earth magnetic field, can
Effectively to solve Accumulated deviation, to correct the direction of motion of human body, attitude angle, movement dynamics and speed etc..
Nine axle sensors reduce circuit board and overall space as integrated transducer, more suitable for use in light and handy portable
Electronic equipment and wearable product in.
The data precision of integrated transducer is not only related with the precision of device itself, after also relating to assemble welding
Correction and for different application mating algorithm.Suitable algorithm can make up the data fusion from multiple sensors
Single deficiency of the sensor when calculating accurate position and direction, to realize high-precision motion detection.
By taking the joint angles for calculating human motion as an example, need using three inertial sensors (IMU1, IMU2, IMU3),
Each inertial sensor includes: 3 axis acceleration sensors, 3 axis gyroscopes and 3 axis electronic compass (geomagnetic sensor), and 3 axis accelerate
Sensor, 3 axis gyroscopes and 3 axis electronic compass measure acceleration, angular speed and magnetic in itself three-dimensional local coordinate system respectively
Field vector.
Referring to Fig. 3, first inertial sensor IMU1 is placed on thigh, second inertial sensor IMU2 is placed on small
On leg, third inertial sensor IMU3 is placed on instep, wherein IMU1 and IMU2 constitutes knee joint, IMU3 and IMU2 structure
At ankle-joint.
Data from IMU1 and IMU2 must be converted into coordinate system relevant to joint, i.e. wherein one or two axis
With the coordinate system of joints axes and/or the longitudinal axes coincident of this section.
Referring to Fig. 4, the method for calculating joint angles using inertial sensor direction estimation value is specific as follows:
IMU1 and IMU2 is provided to be estimated relative to the accurate three-dimensional perspective in common fixed reference frame orientation, then
It is converted into the vector of local joint axial coordinate, angle, that is, accurate joint angles of two vectors in three-dimensional.
We assume that IMU1 and IMU2 is relative to common fixed reference system, (i.e. referential must for each inertial sensor
Must be identical) direction provided by spin matrix, the two spin matrixs use R respectively1(t) and R2(t) it indicates, they should
Them are defined to by the vector median filters of local measurement into reference frame, i.e., we have R1(t) j1=R2(t) j2 ω t, this
In the case of, bend and stretch angle [alpha]acc+gyr+mag(t) it can be easily calculated as:
Wherein,Indicate R3In two vectors between signed angle, and c ε R3Can be it is any make to
(such as c=[1,0,0] T can be used the vector that amount product is not zero, unless j1Or j2Exactly [± 1,0,0] T).
2, cell phone application
Our cell phone application can display and save the data received from multiple sensors by Bluetooth control sensor,
And the principle of human anatomy and sport dynamics is combined to construct according to (acceleration, angular speed, earth magnetism) using 9 number of axle of sensor
The valid model of description human motion (variation such as limbs direction, dynamics) simultaneously therefrom extracts motion feature.
According to the difference of training goal and movement technique, sensing data is acquired to different motive positions, in conjunction with movement point
The theory (such as: method of having an effect, bone and the driving in joint process etc.) for analysing engineering constructs motion model, realizes motion model
Electronic data is prepared for skill analysis.
The coordination variation at the positions such as skeleton and joint has important influence to sports achievement is improved, we are special to movement
Polynary quantification treatment is applied in levies in kind, provides the data that can quickly handle for real time kinematics characteristic evaluating.Motion feature evaluation can be to fortune
Real time data tracking is implemented in the movement of mobilization, when movement deviates with motion model, according to the position of movement generation lack of standardization
It sets, error analysis, and the guidance command corrected with sound to sportsman's proposition movement is carried out to body part, movement technique.
Referring to Fig. 5, the human body attitude identification based on inertial sensor is broadly divided into 5 stages: data acquisition, data are located in advance
Reason, data sectional, feature extraction and classifier training.
(1) data acquire
The stage is mainly the physical signal that human body is acquired by sensor, such as: acceleration, angular speed and joint angles
Etc. information.
(2) data prediction
The stage mainly carries out the noise processing such as synchronous with alignment to data.
(3) data sectional
The stage mainly carries out data extraction in time domain and domain space respectively to individual part, carries out independent analysis.
(4) feature extraction
The stage mainly analyzes unit movement, calculates and extracts relevant attributive character as sample data.
(5) classifier training
The stage is realized pair mainly by the sample action of acquisition according to different principle of classification structural classification models
The division of unknown sample.
Cell phone application handles data, and intelligent algorithm provides feedback in real time (form of feedback includes vibration, sound, figure etc.)
And it saves as a result, data result uploads to Cloud Server by network.
User, sports team, research institution etc. can be applied by cell phone application and website shared data.
3, Cloud Server
Cell phone application upload data result can with long-time storage in Microsoft's Azure cloud platform or private services device, and
Carry out machine learning and analysis.
We store data using Microsoft's Azure cloud platform, while using Azure machine learning (Azure Machine
Learning) come construct we analysis data model and scheme, carried out with the methods of binary Bayes, multivariate logistic regression
Classification and pattern-recognition, while assisting from cloud to the deployment of edge APP and the AI learning outcome of sensor.
Classification based training is the important step of human body attitude identification, and by carrying out data acquisition to different movements, feature mentions
It takes with after selection, the attribute vector of description human body attitude information will be obtained, a large amount of attribute vector collective combinations are to together, just
The set of eigenvectors of individual part is constituted, as constructing the sample set of disaggregated model.
The research of gesture recognition is carried out by sensing data, generally uses statistical pattern recognition method, this needs a large amount of
Sample set carry out classifier training.And a core as human body attitude identification, classifier training need to select
Different sorting algorithms is simultaneously assessed, and optimal classifier is finally selected.
Common sorting algorithm has very much, be set forth below two we.
(1) decision tree (Decision Tree, DT)
Decision tree is the machine learning method of supervised, is commonly used for data classification and recurrence.As a kind of decision tree,
Essence is to simplify challenge, and neutralizing is solved at hierarchical structure, so it is also a kind of multilevel policy decision model.
Decision tree can be counted as the tree structure being made of node and directed edge, be a kind of statistical theory model,
Wherein, node includes internal node and leaf node, and internal node indicates the detection of certain attribute of multisample, the branch extended
A kind of testing result is represented, leaf node indicates a kind of specific classification results.
Building decision tree is a complicated process, and feature selecting and feature division are to construct the committed step of decision tree,
Wherein, feature selecting has specific information reference, it is main according to certain index relevant to this feature, relatively conventional index
There are ratio of profit increase, information gain, entropy etc., feature division can also can be regarded as a kind of subdivided method, by different classes of screening therein
Out, in order to each data for being directed toward side be allowed to belong to same type as much as possible.
The developing algorithm of common decision tree has C4.5, ID3, CART etc..
The principle of decision tree is relatively easy, and construction process is simultaneously uncomplicated, and the building time is relatively short.But this method is for number
It according to missing problem and is not suitable for, is easy to appear over-fitting.
(2) Naive Bayes Classifier
Naive Bayes Classifier is a kind of based on Bayesian Weak Classifier, and all Naive Bayes Classifiers are all
It is assumed that each feature of sample is uncorrelated to other features.For example, if a kind of fruit has red, round, general 3 English of diameter
The features such as very little, the fruit can be judged as apple.Although these features interdepend or some features are determined by other features
It is fixed, however Naive Bayes Classifier think these attributes determine the fruit whether in the probability distribution for being apple it is independent.
Naive Bayes Classifier is easily set up, particularly suitable for large data collection, it is well known that this is a kind of victory
Cross the efficient classification method of many complicated algorithms.
Bayesian formula provides the mode for calculating posterior probability P (X | Y):
In Bayesian network, the relationship that influences each other between feature, we are indicated with directed acyclic graph (DAG).In number
On, we indicate the correlation between object with figure (Graph).Figure is by node (representing object) and side (representation relation) group
At.If side has direction, obtained figure is known as digraph.For a digraph, if from any node
It sets out, all cannot pass through several sides and return to the node, then this digraph is known as directed acyclic graph.In Bayesian network
In, we indicate influence of some feature to another feature with the direction on the side of directed acyclic graph, acyclic, guarantee between feature
The relationship of influencing each other will not fall into inexhaustible circulation.For feature to the specific influence degree of given feature, we use condition
Probability tables (CPT) indicate that this concept of conditional probability table is it will be understood that Fig. 6 just illustrates a simple Bayesian network
The conditional probability table of each feature in network.The relationship that influences each other between feature, that is, the building of DAG depend on experience or field
Knowledge.Conditional probability table can then be improved with naive Bayesian.
It being instructed in system in wearable real-time action provided by the invention, sensor and cell phone application belong to edge device, and two
Person runs in such a way that offline and interval is internuncial, when they are offline or can only be connected intermittently to cloud, once weight
New connection, they will automatic synchronization its last state, and run with continuing Non-intermittent, irrespective of whether connecting, in addition,
The two only sends the data that we need to store and analyzes beyond the clouds, can reduce the data that cloud is got forwarded to from equipment in this way
Amount.
To sum up, wearable real-time action provided by the invention instructs system to have three spotlights:
(1) real-time data analysis is carried out while collecting data, the posture and movement to user provide Real-time Feedback；
(2) entire to count from the initial data of sensor node to significant modelling of human body motion, then to feedback in real time
According to process using the feature extraction and sorting algorithm of intelligence, intelligent personalized data accumulation and analysis are realized；
(3) distributed artificial intelligence framework is used --- sensor+cell phone application (part, in real time)+Cloud Server (it is global,
Afterwards), this is that our system can accomplish the basic of real-time instruction, and computing capability is pushed to sensor and mobile phone by cloud,
Data delay is greatly reduced, the dependence to cloud and network is also got rid of.
Wearable real-time action provided by the invention instructs system to energize wearable sensing equipment kimonos threading row AI,
Can solve various industries and the pain spot problem of age level crowd movement, can quantization movement in real time and amount of exercise, in real time
Movement is corrected in analysis, help, and correct muscle memory is cultivated in training or exercise, and data analysis helps prevent damage, passes through
Wearable sensing equipment collects the instantaneous variation (acceleration, angular speed, the orientation etc. in each portion of limbs) of human action, counts in real time
Motion feature is calculated, analysis acts deviation, identifies nonstandard action, and carry out movement error correction guidance to sporter with voice.In this way
It can not only help sportsman to improve sports achievement, and outstanding fortune can be gone out by the guided teaching fast culture of standard operation
It mobilizes, and in human body recovery field (such as in terms of limb rehabilitating), rehabilitation training guidance is according to pathological analysis
With doctor's diagnosis and treatment as a result, the limb motion energy that used integrated operation specification and sports medical science are paid attention to when with rehabilitation
On the basis of the recovery purpose of power, with motion analysis technique come training action model required when constructing limb rehabilitation training, pass through
Quantification technique makes " rehabilitation training " to have the data evaluation means that can see, and this digitizing technique fundamentally solves state
Interior rehabilitation training guidance without standardized administration, house rehabilitation can be promoted, become the medical services body of low cost, high quality
System, brings benefit to the people.
Wearable real-time action provided by the invention instructs system, by mutually tying artificial intelligence technology with wearable technology
Conjunction is applied on daily exercise guidance, motion exercise target can be allowed more clear, mode is apparent, with following some spies
Point and advantage:
(1) motion detection is more acurrate, and detection accuracy is domestically leading
Wearable real-time action provided by the invention instructs system globe area from the data of multiple sensors, Mei Gechuan
Sensor is capable of real-time acquisition 9 axle exercise datas, based on these data amount and calculation amount, the system can Comprehensive items detection refer to
Mark obtains more accurately and reliably result.
(2) Real-time Feedback instructs, and synchronous remind is corrected
Wearable real-time action provided by the invention instructs system there was only data acquisition, can root during practice
The data acquired when factually analyze these data, and then the movement of mistake is reminded and entangled based on the analysis results
Just, it helps to form correct muscle memory.
(3) distributed artificial intelligence combines local intelligence Real-time Feedback with global intelligent updating study
Wearable real-time action provided by the invention instructs system using international pioneering distributed artificial intelligence framework, will
Local intelligence Real-time Feedback and global intelligence constantly update the combination of study, in this way it is contemplated that personalized and movement is reasonable
Diversity, rather than carry out rigid error correction according to the template of standard.
(4) exercise data accumulates, and realizes historical data comprehensive analysis
Wearable real-time action provided by the invention instructs system that will carry out long-term number on privately owned or publicly-owned server
According to accumulation and analysis, the training and raising of intelligent algorithm can be helped, from the point of view of long-term trend, can really be realized a
Comprehensive analysis of property.
Exercise makes one strong, and correct motion mode is particularly important to healthy and reduction damage is improved.It is provided by the invention
Wearable real-time action instructs system that can be widely used in each age-colony and each field industry.
In body-building field, these data can be divided according to the data acquired in real time during body-building, running etc.
Then analysis based on the analysis results reminds and corrects the movement of mistake, help forms correct muscle memory.
In Sports Field, when carrying out the movement such as basketball, football, baseball, tennis, skating, skiing can Comprehensive it is each
Item Testing index obtains more accurately and reliably result.
In human body recovery field, the rehabilitation training of quantization technological guidance the elderly, patient or disabled person can be passed through.
In the more demanding professional domain of action criteria, such as military field, medical domain, soldier can be helped to be formed just
True trail helps doctor to form correctly operation gimmick.
It should be noted that the above embodiments do not limit the invention in any form, it is all to use equivalent replacement or equivalent change
The mode changed technical solution obtained, falls within the scope of protection of the present invention.
Claims (10)
1. a kind of wearable real-time action instructs system, which is characterized in that distributed artificial intelligence framework is used, by hardware, hand
Machine APP and Cloud Server composition, constantly update study with global intelligence for local intelligence Real-time Feedback and combine, calculating
Ability pushes hardware and mobile phone to by cloud, in which:
The hardware is the external member for the easy wearing being made of multiple sensor nodes, and multiple sensor nodes collect data, real
When calculate and fusion, as a result by Bluetooth transmission to cell phone application；
The cell phone application displays and saves the data received from multiple sensors by Bluetooth control sensor, and utilizes sensing
The data combination human anatomy of device and the principle of sport dynamics construct the valid model of description human motion and therefrom extract fortune
Dynamic feature, when movement deviates with motion model, according to the position that movement lack of standardization occurs, to body part, movement side
Method carries out error analysis, provides feedback in real time, and the guidance command corrected with sound to sportsman's proposition movement, data result are logical
It crosses network and uploads to Cloud Server；
The data result that the Cloud Server long-time storage cell phone application uploads, and carry out machine learning and analysis.
2. wearable real-time action according to claim 1 instructs system, which is characterized in that inside the sensor node
Need to complete the work of data analysis and process comprising four modules: sensor module, processor module, communication module and
Power supply, in which:
Sensor module is made of sensing element and AD conversion unit, and sensing element is responsible for the motion information of test object, and AD turns
Unit is changed to be responsible for motion information being transformed into electric signal；
Processor module is made of processor and memory, and processor is responsible for controlling the normal work of other each modules of sensor node
Make, and carry out the relevant treatment of each signal, memory is responsible for storing related data；
Communication module is made of network protocol and transceiver, is responsible for the transmission of signal, node data is wirelessly transmitted
To other equipment；
To sensor node, other each modules provide the energy operated normally to power supply.
3. wearable real-time action according to claim 2 instructs system, which is characterized in that the processor using
ARM Cortex-M4 processor.
4. wearable real-time action according to claim 2 instructs system, which is characterized in that the sensor node uses
Be nine axle sensors.
5. wearable real-time action according to claim 1 instructs system, which is characterized in that the cell phone application being capable of root
According to the difference of training goal and movement technique, sensing data is acquired to different motive positions, in conjunction with the reason of motion analysis engineering
By motion model is constructed, realizes the electronic data of motion model, prepare for skill analysis.
6. wearable real-time action according to claim 1 instructs system, which is characterized in that including user, sports team, grind
Studying carefully the personnel and organizations including mechanism can be applied and website shared data by cell phone application.
7. wearable real-time action according to claim 1 instructs system, which is characterized in that the sensor and mobile phone
APP is run in such a way that offline and interval is internuncial, when they are offline or can only be connected intermittently to cloud, once weight
New connection, they will automatic synchronization its last state, and run with continuing Non-intermittent, irrespective of whether connecting.
8. wearable real-time action according to claim 7 instructs system, which is characterized in that the sensor and mobile phone
APP only sends the data that we need to store and analyzes beyond the clouds.
9. wearable real-time action according to claim 1 instructs system, which is characterized in that the Cloud Server is Microsoft
Azure cloud platform or private services device.
10. wearable real-time action according to claim 1 instructs system, which is characterized in that flat using Microsoft Azure cloud
Platform, and come using Azure machine learning the model and scheme of creation analysis data, with including that binary Bayes, polynary logic are returned
Method including returning carries out classification and pattern-recognition, while assisting from cloud to the portion of edge APP and the AI learning outcome of sensor
Administration.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910683451.2A CN110502107A (en) | 2019-07-26 | 2019-07-26 | Wearable real-time action instructs system and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910683451.2A CN110502107A (en) | 2019-07-26 | 2019-07-26 | Wearable real-time action instructs system and method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110502107A true CN110502107A (en) | 2019-11-26 |
Family
ID=68587513
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910683451.2A Pending CN110502107A (en) | 2019-07-26 | 2019-07-26 | Wearable real-time action instructs system and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110502107A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111097142A (en) * | 2019-12-19 | 2020-05-05 | 武汉西山艺创文化有限公司 | Motion capture motion training method and system based on 5G communication |
| CN111488861A (en) * | 2020-05-13 | 2020-08-04 | 吉林建筑大学 | Ski athlete gesture recognition system based on multi-feature value fusion |
| CN112691357A (en) * | 2020-12-29 | 2021-04-23 | 深圳市艾利特医疗科技有限公司 | Cardiopulmonary function exercise system and method |
| CN115527264A (en) * | 2022-01-24 | 2022-12-27 | 温州医科大学 | Intelligent evaluation method and system for cheering exercise training effect |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105797350A (en) * | 2016-03-18 | 2016-07-27 | 深圳大学 | Intelligent method and system for body building posture recognition, evaluation, early-warning and intensity estimation |
| CN106153077A (en) * | 2016-09-22 | 2016-11-23 | 苏州坦特拉自动化科技有限公司 | A kind of initialization of calibration method for M IMU human motion capture system |
| CN108379815A (en) * | 2018-02-02 | 2018-08-10 | 梦卓科技（深圳）有限公司 | The automation training system with Real-time Feedback based on elastic intelligent sensor node |
| CN109011508A (en) * | 2018-07-30 | 2018-12-18 | 三星电子（中国）研发中心 | A kind of intelligent coach system and method |
| WO2019021315A1 (en) * | 2017-07-27 | 2019-01-31 | Tarun Kumar | Motion sense technology system |
| US20190054347A1 (en) * | 2015-08-18 | 2019-02-21 | Michael Saigh | Wearable sports guidance communication system and developers tool kit |
| CN109480858A (en) * | 2018-12-29 | 2019-03-19 | 中国科学院合肥物质科学研究院 | It is a kind of for quantify detect disturbances in patients with Parkinson disease bradykinesia symptom wearable intelligence system and method |
-
2019
- 2019-07-26 CN CN201910683451.2A patent/CN110502107A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190054347A1 (en) * | 2015-08-18 | 2019-02-21 | Michael Saigh | Wearable sports guidance communication system and developers tool kit |
| CN105797350A (en) * | 2016-03-18 | 2016-07-27 | 深圳大学 | Intelligent method and system for body building posture recognition, evaluation, early-warning and intensity estimation |
| CN106153077A (en) * | 2016-09-22 | 2016-11-23 | 苏州坦特拉自动化科技有限公司 | A kind of initialization of calibration method for M IMU human motion capture system |
| WO2019021315A1 (en) * | 2017-07-27 | 2019-01-31 | Tarun Kumar | Motion sense technology system |
| CN108379815A (en) * | 2018-02-02 | 2018-08-10 | 梦卓科技（深圳）有限公司 | The automation training system with Real-time Feedback based on elastic intelligent sensor node |
| CN109011508A (en) * | 2018-07-30 | 2018-12-18 | 三星电子（中国）研发中心 | A kind of intelligent coach system and method |
| CN109480858A (en) * | 2018-12-29 | 2019-03-19 | 中国科学院合肥物质科学研究院 | It is a kind of for quantify detect disturbances in patients with Parkinson disease bradykinesia symptom wearable intelligence system and method |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111097142A (en) * | 2019-12-19 | 2020-05-05 | 武汉西山艺创文化有限公司 | Motion capture motion training method and system based on 5G communication |
| CN111488861A (en) * | 2020-05-13 | 2020-08-04 | 吉林建筑大学 | Ski athlete gesture recognition system based on multi-feature value fusion |
| CN112691357A (en) * | 2020-12-29 | 2021-04-23 | 深圳市艾利特医疗科技有限公司 | Cardiopulmonary function exercise system and method |
| CN115527264A (en) * | 2022-01-24 | 2022-12-27 | 温州医科大学 | Intelligent evaluation method and system for cheering exercise training effect |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Rana et al. | Wearable sensors for real-time kinematics analysis in sports: A review | |
| US10352962B2 (en) | Systems and methods for real-time data quantification, acquisition, analysis and feedback | |
| US11679300B2 (en) | Systems and methods for real-time data quantification, acquisition, analysis, and feedback | |
| US11006856B2 (en) | Method and program product for multi-joint tracking combining embedded sensors and an external sensor | |
| CN110502107A (en) | Wearable real-time action instructs system and method | |
| CN102567638B (en) | A kind of interactive upper limb healing system based on microsensor | |
| US8944939B2 (en) | Inertial measurement of sports motion | |
| JP2016208516A (en) | Method and device for associating frames in video of activity of person with event | |
| Ma et al. | Basketball movements recognition using a wrist wearable inertial measurement unit | |
| EP4021297B1 (en) | Systems for wearable devices that determine balance indices | |
| CN107533806A (en) | It is configured as realizing framework, the apparatus and method of the transmission to the interaction skill training content including the content with multiple selectable expertise changes | |
| KR20220086166A (en) | smart golf exercise equipment, system method thereof | |
| Bakr | Digital Jewellery and Machine Learning Methods for Analysing Player Performance in sports | |
| Hu et al. | Application of intelligent sports goods based on human-computer interaction concept in training management | |
| Zhang et al. | The relationship between measurement and evaluation in physical education teaching based on intelligent analysis and sensor data mining | |
| US20230302325A1 (en) | Systems and methods for measuring and analyzing the motion of a swing and matching the motion of a swing to optimized swing equipment | |
| Yao et al. | Research on position optimization of badminton doubles based on wireless sensor and human-computer interactive training | |
| Chiramba et al. | Action Tracer: An Extendable, Low-Cost and Open-Source Motion Tracking Kit Based on 6-Axes Inertial Measurement Units | |
| Link | Wearables in Sports: From Experimental Validation to Deep Learning Applications | |
| Yao et al. | Research on Position Optimization of Badminton Doubles Based on Wireless Network Sensor and Human-computer Interactive Training | |
| Chen | Precise Recognition Method of Local Fuzzy Features in Sprint Video Images | |
| Jia | Artificial Intelligence Of Things For Ubiquitous Sports Analytics | |
| AbdEl_Azeem | Building a Smartphone application to pair inertial sensors (Notch) for mechanical motion analysis | |
| KR20230102022A (en) | system and control method for acquiring and analyzing swing data by classifying golf swing stages through the use of MEMS sensors |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20191126 |
|
| WD01 | Invention patent application deemed withdrawn after publication |