CN108211268A - Exercise load monitoring and sports fatigue method for early warning and system based on training data - Google Patents
Exercise load monitoring and sports fatigue method for early warning and system based on training data Download PDFInfo
- Publication number
- CN108211268A CN108211268A CN201810074759.2A CN201810074759A CN108211268A CN 108211268 A CN108211268 A CN 108211268A CN 201810074759 A CN201810074759 A CN 201810074759A CN 108211268 A CN108211268 A CN 108211268A
- Authority
- CN
- China
- Prior art keywords
- fatigue
- data
- training
- energy consumption
- exercise load
- 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.)
- Granted
Links
Classifications
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
- A63B24/00—Electric or electronic controls for exercising apparatus of preceding groups; Controlling or monitoring of exercises, sportive games, training or athletic performances
- A63B24/0062—Monitoring athletic performances, e.g. for determining the work of a user on an exercise apparatus, the completed jogging or cycling distance
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
- A63B24/00—Electric or electronic controls for exercising apparatus of preceding groups; Controlling or monitoring of exercises, sportive games, training or athletic performances
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
- A63B24/00—Electric or electronic controls for exercising apparatus of preceding groups; Controlling or monitoring of exercises, sportive games, training or athletic performances
- A63B24/0062—Monitoring athletic performances, e.g. for determining the work of a user on an exercise apparatus, the completed jogging or cycling distance
- A63B2024/0065—Evaluating the fitness, e.g. fitness level or fitness index
Landscapes
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Physical Education & Sports Medicine (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provide exercise load monitoring based on training data and with sports fatigue method for early warning and system, comprising two stages, exercise load monitoring stage and sports fatigue early warning stage.Data acquisition used in the exercise load monitoring stage of the present invention is simple, using the neural network concurrent optimization algorithm based on big data, the high-precision that exercise load and intensity are carried out according to determining motor pattern and acceleration information is estimated, realizes the real-time monitoring of training；The sports fatigue early warning stage, according to different motion item characteristic, divide rational sports fatigue time measurement unit, and according to the different mode of fatigue accumulation, drop mode of the exercise load in unit interval unit is monitored in real time, early warning is carried out to sports fatigue with Bayesian Classification Arithmetic, prevents over training from leading to injury gained in sports, it is possible to prevente effectively from injury gained in sports caused by over training.
Description
Technical field
The invention belongs to exercise data analysis field, more particularly to the exercise load monitoring based on training data with
Sports fatigue method for early warning and system.
Background technology
With the rapid development of computer hardware technique and network, video camera, sensor and wireless sensor network are utilized
Network etc. can acquire the sports training data of magnanimity in real time.In face of the sports training data of magnanimity, traditional data
Processing mode is faced with new severe challenge, the big quantization of training big data, diversification, rapid low with value density etc.
Data characteristics allow traditional data processing method and tool that " data " can only be hoped to heave a sigh.How effectively to build and mutually fitted with exercise data
Magnanimity training data variation is really sports load Analysis for valuable information by the mathematical model and tool answered
Field urgent problem to be solved.
Exercise load refers to body during movement to physiological function caused by the training burden and the intensity stimulation that are applied stress
The size of the amount reflected under state, including load, load intensity.The index of evaluation exercise load mainly has at present：Heart rate,
RPE, maximal oxygen uptake, Urine proteins, blood lactase acid, calorie and METS etc..
It is main at present still by Physiology and biochemistry data both at home and abroad in exercise load monitoring and sports fatigue early warning, it is main
The Monitoring Indexes such as heart rate, RPE, maximal oxygen uptake, Urine proteins, blood lactase acid, calorie and METS that will be by detecting sportsman are instructed
Whether suitable practice intensity, if overtrain and produce sports fatigue.The acceleration of U.S. NFL acquisition sportsmen, vertical spring
The data such as power and heart rate, to assist formulating reasonable training plan.The scholar of Beijing Sport University is disappeared using heart rate, RPE and energy
The method of consumption value is monitored volleyballer's training strength.The researcher of Shanghai University Of Sport is with video analysis, heart rate
Monitoring and fatigue scale, the exercise load of comprehensive analysis football referee.
1. Physiology and biochemistry data are depended in existing exercise load monitoring unduly, data acquisition is complex；2. Physiology and biochemistry
There are timeliness for the acquisition of data, if acquisition is not in time, there will be large errors for acquired data；3. existing training
There are hysteresis qualitys for monitoring method, it is impossible to which the generation and evolution of reflection sports fatigue in real time can not carry out over fatigue in time
With injury gained in sports early warning.4. existing exercise load computational methods are applied to small-scale exercise data, movement big data is not applied to
Analysis.
Invention content
In view of the deficiencies of the prior art, the method for the present invention is directed to the kinematics acquired in sports training and moves the present invention
The data such as mechanics by big data management and data mining technology depth analysis, are monitored applied to exercise load, are realized based on big
The exercise load of data calculates in real time and sports fatigue early warning, reaches the target of scientific guidance training.
In order to achieve the above object, technical solution provided by the invention is a kind of exercise load based on training data
Monitoring and sports fatigue method for early warning, include the following steps：
Step 1, the exercise load monitoring stage calculates the energy consumption of subject using neural network concurrent optimization method, described
Rule that energy consumption refers to subject motion's intensity and amount of exercise changes with time, including following sub-step；
Step 1.1, Estimation of energy consumption training sample database, each sample data are established according to different types of motor pattern
Age, gender, height, weight comprising players, the acceleration (a of three dimensionsx,ay,az), totally 7 components；
Step 1.2, Estimation of energy consumption BP neural network is established for each motor pattern, and carries out netinit, initially
Change comprising maximum frequency of training, learn precision, Hidden nodes, initial weight, threshold value, initial learning rate；
Step 1.3, input Estimation of energy consumption training sample X1,X2,……,Xk, Xk=[xk1,xk2,xk3,xk4,xk5,xk6,
xk7], wherein k represents training sample number, 7 components correspond respectively to age, gender, height, weight and three dimensions plus
Speed, and training sample is evenly dispersed in each Map nodes；
Step 1.4, the network weight preserved in Hadoop distributed file systems is read in using Map functions to record, i.e., just
Beginning weights instantiate the neural network on each Map nodes, and obtain the power of each Map nodes according to the network weight of reading
It is worth and exports；
Step 1.5, the network weight preserved in Hadoop distributed file systems is read in using Reduce functions to record, and
Receive Map output weights, and according to each Map nodes output weights arithmetic average, as new network weight into
Row update, while calculate between updated network weight and the network weight read in from Hadoop distributed file systems
Difference judges whether to need to recycle next time according to difference；If network weight no longer updates, Estimation of energy consumption BP neural network
Algorithm training terminates；
Step 1.6, the consumption information of subject is obtained using trained Estimation of energy consumption BP neural network algorithm；
Step 2, the sports fatigue early warning stage carries out subject sports fatigue early warning, packet using Bayesian Classification Arithmetic
Include following sub-step；
Step 2.1, if x={ a1, a2 .., am } is sportsman's data to be sorted, wherein ai is special i-th of x
Levy attribute, common m characteristic attribute；The acquisition modes of characteristic attribute are：Time series in the exercise load monitoring stage is divided
For m section, the corresponding fatigue data in each section is obtained；
Step 2.2, it is known that tired category set C={ (x1, y1), (x2, y2) .., (xn, yn) }, wherein x1, x2 ..,
Xn is the tired correlation properties data of n sportsman, and y1, y2 .., yn are classified for corresponding fatigue；
Step 2.3, the training dataset in sports fatigue early warning stage is obtained, i.e., data set known to tired classification；
Step 2.4, according to known tired category set, the condition that statistics obtains each characteristic attribute under of all categories is general
Rate estimation, i.e.,：P(a1|y1),P(a2|y1),...,P(am|y1)；P(a1|y2),P(a2|y2),...,P(am|y2)；...；P
(a1|yn),P(a2|yn),...,P(am|yn)；
If each characteristic attribute is conditional sampling, following derivation is had according to Bayes' theorem：Some sportsman's
Data x to be sorted may be that probability P (yi | x)=P (x | yi) P (yi)/P (x), wherein P (yi) of yi level of fatigue is fatigue etc.
The probability that grade occurs can obtain according to known tired category set data statistics, and P (x | yi) is x in the data of yi level of fatigue
The probability of appearance；
For arbitrary x, denominator P (x) is fixed value, P (x | yi) P (yi)=P (a1 | yi) P (a2 | yi) ... P (am | yi) P
(yi), i.e.,
Finally calculate the probability that x is each classification：P(y1|x),P(y2|x),..,P(yn|x)；
Step 2.5, if P (yk | x)=max { P (y1 | x), P (y2 | x) .., P (yn | x) }, then the classification of x is for yk
Given level of fatigue.
Further, the number of Map nodes described in step 1.3 is determined according to the number of training sample.
It is special the present invention also provides a kind of exercise load monitoring based on training data and sports fatigue early warning system
Sign is, including following module：
Exercise load monitoring modular, for exercise load monitor the stage, using neural network concurrent optimization method calculate by
The energy consumption of examination person, rule that the energy consumption refers to subject motion's intensity and amount of exercise changes with time, including following submodule；
Estimation of energy consumption training sample setting up submodule, for establishing Estimation of energy consumption training according to different types of motor pattern
Sample database, each sample data include ages of players, gender, height, weight, three dimensions acceleration (ax,
ay,az), totally 7 components；
Netinit submodule establishes Estimation of energy consumption BP neural network, and carry out net for being directed to each motor pattern
Network initializes, and initialization package contains maximum frequency of training, learns precision, Hidden nodes, initial weight, threshold value, initial learning rate；
Training sample distribution sub module, for inputting Estimation of energy consumption training sample X1,X2,……,Xk, Xk=[xk1,xk2,
xk3,xk4,xk5,xk6,xk7], wherein k represents training sample number, and 7 components correspond respectively to age, gender, height, weight
With the acceleration of three dimensions, and training sample is evenly dispersed in each Map nodes；
Map node weights output sub-modules, for reading in what is preserved in Hadoop distributed file systems using Map functions
Network weight records, i.e. initial weight, and the neural network on each Map nodes is instantiated, and obtain according to the network weight of reading
Obtain the weights of each Map nodes and output；
Network weight updates submodule, for reading in what is preserved in Hadoop distributed file systems using Reduce functions
Network weight records, and receives the weights of Map outputs, and the arithmetic average of the output weights according to each Map nodes, as
New network weight is updated；Updated network weight is calculated simultaneously and is read in from Hadoop distributed file systems
Difference between network weight judges whether to need to recycle next time according to difference；If network weight no longer updates, energy consumption
The algorithm training of estimation BP neural network terminates；
Energy consumption testing submodule, for obtaining the energy consumption of subject using trained Estimation of energy consumption BP neural network algorithm
Information；
Sports fatigue warning module for the sports fatigue early warning stage, carries out subject using Bayesian Classification Arithmetic
Sports fatigue early warning, including following submodule；
Characteristic attribute acquisition submodule, if x={ a1, a2 .., am } is sportsman's data to be sorted, wherein ai
For the ith feature attribute of x, common m characteristic attribute；The acquisition modes of characteristic attribute are：It will be in the exercise load monitoring stage
Time series is divided into m section, obtains the corresponding fatigue data in each section；
Fatigue classification submodule, it is known that tired category set C={ (x1, y1), (x2, y2) .., (xn, yn) }, wherein
X1, x2 .., xn are the tired correlation properties data of n sportsman, and y1, y2 .., yn are classified for corresponding fatigue；
Sports fatigue training dataset acquisition submodule, for according to known tired category set, statistics to be obtained each
The conditional probability estimation of each characteristic attribute under classification, i.e.,：P(a1|y1),P(a2|y1),...,P(am|y1)；P(a1|y2),P
(a2|y2),...,P(am|y2)；...；P(a1|yn),P(a2|yn),...,P(am|yn)；
If each characteristic attribute is conditional sampling, following derivation is had according to Bayes' theorem：Some sportsman's
Data x to be sorted may be that probability P (yi | x)=P (x | yi) P (yi)/P (x), wherein P (yi) of yi level of fatigue is fatigue etc.
The probability that grade occurs can obtain according to known tired category set data statistics, and P (x | yi) is x in the data of yi level of fatigue
The probability of appearance；
For arbitrary x, denominator P (x) is fixed value, P (x | yi) P (yi)=P (a1 | yi) P (a2 | yi) ... P (am | yi) P
(yi), i.e.,
Finally calculate the probability that x is each classification：P(y1|x),P(y2|x),..,P(yn|x)；
Level of fatigue decision sub-module, if P (yk | x)=max { P (y1 | x), P (y2 | x) .., P (yn | x) }, then x
Classification is given level of fatigue for yk.
Further, the number of Map nodes described in training sample distribution sub module is determined according to the number of training sample.
Compared with prior art, advantages of the present invention：Data acquisition used in the exercise load monitoring stage is simple, using base
In the neural network concurrent optimization algorithm of big data, according to determining motor pattern and acceleration information carry out exercise load and
The high-precision estimation of intensity, realizes the real-time monitoring of training；The sports fatigue early warning stage, according to different motion project spy
Sign divides rational sports fatigue time measurement unit, and according to the different mode of fatigue accumulation, i.e., monitors exercise load in real time
The drop mode in unit interval unit is measured, early warning is carried out to sports fatigue with Bayesian Classification Arithmetic, prevents from excessively instructing
White silk leads to injury gained in sports, it is possible to prevente effectively from injury gained in sports caused by over training.
Description of the drawings
Fig. 1 is the exercise load algorithm flow chart based on neural network in the embodiment of the present invention.
Fig. 2 is the sports fatigue early warning bayesian algorithm flow chart based on big data in the embodiment of the present invention.
Specific embodiment
Technical scheme of the present invention is described further with reference to the accompanying drawings and examples.
In the embodiment of the present invention it is a kind of based on training data exercise load monitoring and with sports fatigue method for early warning
Comprising two stages, exercise load monitoring stage and sports fatigue early warning stage：
1. the exercise load computational algorithm based on big data：It is moved according to determining motor pattern and acceleration information
The high-precision evaluation method of load and intensity, and realize that the exercise load calculates in embedded systems.
The characteristics of simpler according to neural network structure, but amount of training data is larger and complex, using based on training
The strategy of the parallelization of data set realizes the neural network concurrent optimization algorithm based on big data.The parallelization of data, specifically
There are one complete network in each calculate node, and network initial state is consistent.Algorithm parallelization is embodied in progress
During training, each calculate node is trained with part sample data, after some convergent requirement is reached in calculate node
Summarized again, decided whether to carry out iteration next time by summarized results.Algorithm flow is shown in attached drawing 2, and specific algorithm flow is such as
Under：
(1) Estimation of energy consumption training sample database is established according to different types of motor pattern.In order to quantifying greatly, it is various
The training data of change are preferably managed, and the present invention has carried out structure optimization to movement big data, it is proposed that based on more
The training data organizational structure of matrix is tieed up, each training data attribute is defined as a dimension, training
N number of attribute will form a N-dimensional vector；Each of which property value is expressed with M dimensional vectors again, then the information of every record can
To be represented with a M × N matrix, and its training information in different time then forms a three-dimensional matrice, team's instruction of more people
Practicing information can then be represented with the matrix of more higher-dimension；By building High Dimensional Data Set, the effective of the correlation rule that extracts is promoted
Property.Each sample data should the age comprising players, gender, height, weight, (ax,ay,az) three-dimensional acceleration, training rank
Section, the corresponding consumption information of subject it is known that energy consumption therein represent be exercise intensity and amount of exercise changes with time rule
Rule.
(2) Estimation of energy consumption BP neural network is established for each motor pattern, and carries out netinit, instructed comprising maximum
Practice number, learn precision, Hidden nodes, initial weight, threshold value, initial learning rate etc..Wherein weight initialization is：It is random to assign
Give input, implicit and output layer weight wmi,wij,wjnOne group of smaller non-zero values.
(3) input Estimation of energy consumption training sample X1,X2,……,Xk, and be evenly dispersed in each Map nodes.
(4) Map functions read in the network weight record preserved in Hadoop distributed file systems, according to the network weight example of reading
Change a neural network.Extract training sample, input quantity Xk=[xk1,xk2,xk3,xk4,xk5,xk6,xk7], wherein k is represented
Training sample number, 7 components correspond respectively to the acceleration of age, gender, height, weight and three dimensions；Output quantity Dk
For the power consumption values in sample database, i.e., desired power consumption values.To inputting, exporting and be normalized, using sample into
Row training, according to given input training sample Xi, output quantity Y is gone out by neural computingi(n)；After wherein n represents n times iteration
Output valve, n=1 when calculating for the first time, and by Yi(n) and DiCompare (i=1,2 ... ..., k), if error is more than setting value, then
According to the total error criteria function of k training sample, the layer modified weight amount of each layer is corrected successively with error gradient descent method
With layer threshold value correction amount, read in new sample and be trained；If error is less than setting value, then the weights in this node are obtained, and
Output, idiographic flow are as shown in Figure 1.
(5) Reduce functions read in the network weight record preserved in Hadoop distributed file systems, and it is defeated to receive Map
The weights gone out, and the arithmetic average of the output weights according to each Map nodes, are updated as new network weight.Than
More updated network weight and the network weight read in from Hadoop distributed file systems, calculating difference, according to difference
Judge whether to need to recycle next time, which determines according to algorithm precision prescribed.If network weight no longer updates, then energy consumption is estimated
The training of BP neural network algorithm is calculated to terminate.
(6) consumption information of subject is obtained using trained Estimation of energy consumption BP neural network algorithm；
2. the sports fatigue warning algorithm based on big data：According to different motion item characteristic, division reasonably moves tired
Labor time measurement unit, and according to the different mode of fatigue accumulation, i.e., exercise load is monitored in real time in unit interval unit
Drop mode and the drop mode of level of decision-making during the games, sports fatigue is carried out with Bayesian Classification Arithmetic pre-
It is alert, prevent over training from leading to injury gained in sports.
Within cycle of training, according to the exercise intensity and amount of exercise of RPE questionnaires and exercise load monitoring gained at any time
Changing rule (the exercise intensity peak value caused by fatigue after training and in the unit interval amount of exercise decline), by coach, Dui Yihe
It is discriminated whether with team scientific research personnel in fatigue phase and level of fatigue, and the data acquisition system of known level of fatigue is established with this.
It is using the specific steps of Bayesian Classification Arithmetic prediction fatigue below, algorithm flow is shown in attached drawing 2,：
(1) x={ a1, a2 .., am } is set as sportsman's data to be sorted, and the feature category that each ai is x
Property, characteristic attribute here can be tired questionnaire statistical value, and fatigue data can be obtained according to RPE questionnaires, between 6-20, number
Value is bigger, and representative is more tired.
(2) fatigue category set C={ (x1, y1), (x2, y2) .., (xn, yn) }, wherein x1, x2 .., xn n known to
The tired correlation properties data of a sportsman, y1, y2 .., yn are classified for corresponding fatigue, and the reply of different fatigue phase is arranged
Difference is applied, such as the value of yi can be 1,2,3,4, wherein 1 corresponds to fine tuning drill program, 2 correspond to reduction strenuous exercise instruction
Practice, 3 correspond to the amount and intensity for substantially reducing training, and 4 correspond to deconditioning.
(3) training dataset is obtained, i.e., data set known to tired classification.
(4) according to known tired category set, the conditional probability that statistics obtains each characteristic attribute under of all categories is estimated
Meter, i.e.,：P(a1|1),P(a2|1),...,P(am|1)；P(a1|2),P(a2|2),...,P(am|2)；P(a1|3),P(a2|
3),...,P(am|3)；P(a1|4),P(a2|4),...,P(am|4).
(5) if each characteristic attribute is conditional sampling, following derivation is had according to Bayes' theorem：Some sportsman
Data x to be sorted may be probability P (yi | x)=P (x | yi) P (yi)/P (x), wherein P (yi) of yi level of fatigue be fatigue
The probability that grade occurs can obtain according to known tired category set data statistics, and P (x | yi) in the data of yi level of fatigue
The probability that x occurs；
For Mr. Yu x, denominator P (x) is fixed, as long as so finding out the maximum as conditional probability maximum of molecule
's.Again because each characteristic attribute is conditional sampling, have：P (x | yi) P (yi)=P (a1 | yi) P (a2 | yi) ... P (am |
Yi) P (yi) is i.e.
(6) probability that x is each classification is calculated：P(y1|x),P(y2|x),..,P(yn|x)；
(7) if P (yk | x)=max { P (y1 | x), P (y2 | x) .., P (yn | x) }, then the classification of x is given for yk
Level of fatigue.
In addition, the embodiment of the present invention also provide it is a kind of based on training data exercise load monitoring it is pre- with sports fatigue
Alert system, including following module：
Exercise load monitoring modular, for exercise load monitor the stage, using neural network concurrent optimization method calculate by
The energy consumption of examination person, rule that the energy consumption refers to subject motion's intensity and amount of exercise changes with time, including following submodule；
Estimation of energy consumption training sample setting up submodule, for establishing Estimation of energy consumption training according to different types of motor pattern
Sample database, each sample data include ages of players, gender, height, weight, three dimensions acceleration (ax,
ay,az), totally 7 components；
Netinit submodule establishes Estimation of energy consumption BP neural network, and carry out net for being directed to each motor pattern
Network initializes, and initialization package contains maximum frequency of training, learns precision, Hidden nodes, initial weight, threshold value, initial learning rate；
Training sample distribution sub module, for inputting Estimation of energy consumption training sample X1,X2,……,Xk, Xk=[xk1,xk2,
xk3,xk4,xk5,xk6,xk7], wherein k represents training sample number, and 7 components correspond respectively to age, gender, height, weight
With the acceleration of three dimensions, and training sample is evenly dispersed in each Map nodes；
Map node weights output sub-modules, for reading in what is preserved in Hadoop distributed file systems using Map functions
Network weight records, i.e. initial weight, and the neural network on each Map nodes is instantiated, and obtain according to the network weight of reading
Obtain the weights of each Map nodes and output；
Network weight updates submodule, for reading in what is preserved in Hadoop distributed file systems using Reduce functions
Network weight records, and receives the weights of Map outputs, and the arithmetic average of the output weights according to each Map nodes, as
New network weight is updated；Updated network weight is calculated simultaneously and is read in from Hadoop distributed file systems
Difference between network weight judges whether to need to recycle next time according to difference；If network weight no longer updates, energy consumption
The algorithm training of estimation BP neural network terminates；
Energy consumption testing submodule, for obtaining the energy consumption of subject using trained Estimation of energy consumption BP neural network algorithm
Information；
Sports fatigue warning module for the sports fatigue early warning stage, carries out subject using Bayesian Classification Arithmetic
Sports fatigue early warning, including following submodule；
Characteristic attribute acquisition submodule, if x={ a1, a2 .., am } is sportsman's data to be sorted, wherein ai
For the ith feature attribute of x, common m characteristic attribute；The acquisition modes of characteristic attribute are：It will be in the exercise load monitoring stage
Time series is divided into m section, obtains the corresponding fatigue data in each section；
Fatigue classification submodule, it is known that tired category set C={ (x1, y1), (x2, y2) .., (xn, yn) }, wherein
X1, x2 .., xn are the tired correlation properties data of n sportsman, and y1, y2 .., yn are classified for corresponding fatigue；
Sports fatigue training dataset acquisition submodule, for according to known tired category set, statistics to be obtained each
The conditional probability estimation of each characteristic attribute under classification, i.e.,：P(a1|y1),P(a2|y1),...,P(am|y1)；P(a1|y2),P
(a2|y2),...,P(am|y2)；...；P(a1|yn),P(a2|yn),...,P(am|yn)；
If each characteristic attribute is conditional sampling, following derivation is had according to Bayes' theorem：Some sportsman's
Data x to be sorted may be that probability P (yi | x)=P (x | yi) P (yi)/P (x), wherein P (yi) of yi level of fatigue is fatigue etc.
The probability that grade occurs can obtain according to known tired category set data statistics, and P (x | yi) is x in the data of yi level of fatigue
The probability of appearance；
For arbitrary x, denominator P (x) is fixed value, P (x | yi) P (yi)=P (a1 | yi) P (a2 | yi) ... P (am | yi) P
(yi), i.e.,
Finally calculate the probability that x is each classification：P(y1|x),P(y2|x),..,P(yn|x)；
Level of fatigue decision sub-module, if P (yk | x)=max { P (y1 | x), P (y2 | x) .., P (yn | x) }, then x
Classification is given level of fatigue for yk.
The number of Map nodes is determined according to the number of training sample wherein described in training sample distribution sub module.
Each module specific implementation is corresponding with each step, and it will not go into details by the present invention.
Specific embodiment described herein is only an example for the spirit of the invention.Technology belonging to the present invention is led
The technical staff in domain can do various modifications or additions to described specific embodiment or replace in a similar way
In generation, however, it does not deviate from the spirit of the invention or beyond the scope of the appended claims.
Claims (4)
1. exercise load monitoring and sports fatigue method for early warning based on training data, which is characterized in that including walking as follows
Suddenly：
Step 1, the exercise load monitoring stage calculates the energy consumption of subject, the energy consumption using neural network concurrent optimization method
Refer to subject motion's intensity and amount of exercise changes with time rule, including following sub-step；
Step 1.1, Estimation of energy consumption training sample database is established according to different types of motor pattern, each sample data includes
Ages of players, gender, height, weight, three dimensions acceleration (ax,ay,az), totally 7 components；
Step 1.2, Estimation of energy consumption BP neural network is established for each motor pattern, and carries out netinit, initialization package
Containing maximum frequency of training, learn precision, Hidden nodes, initial weight, threshold value, initial learning rate；
Step 1.3, input Estimation of energy consumption training sample X1,X2,……,Xk, Xk=[xk1,xk2,xk3,xk4,xk5,xk6,xk7],
Middle k represents training sample number, and 7 components correspond respectively to the acceleration of age, gender, height, weight and three dimensions, and
Training sample is evenly dispersed in each Map nodes；
Step 1.4, the network weight preserved in Hadoop distributed file systems is read in using Map functions to record, i.e., initial power
Value instantiates the neural network on each Map nodes, and obtain the weights of each Map nodes simultaneously according to the network weight of reading
Output；
Step 1.5, the network weight preserved in Hadoop distributed file systems is read in using Reduce functions to record, and receive
The weights of Map outputs, and the arithmetic average of the output weights according to each Map nodes, carry out more as new network weight
Newly, while the difference between updated network weight and the network weight read in from Hadoop distributed file systems is calculated
Value, judges whether to need to recycle next time according to difference；If network weight no longer updates, Estimation of energy consumption BP neural network is calculated
Method training terminates；
Step 1.6, the consumption information of subject is obtained using trained Estimation of energy consumption BP neural network algorithm；
Step 2, the sports fatigue early warning stage, using Bayesian Classification Arithmetic to subject carry out sports fatigue early warning, including with
Lower sub-step；
Step 2.1, if x={ a1, a2 .., am } is sportsman's data to be sorted, wherein ai is the ith feature category of x
Property, common m characteristic attribute；The acquisition modes of characteristic attribute are：Time series in the exercise load monitoring stage is divided into m
Section obtains the corresponding fatigue data in each section；
Step 2.2, it is known that tired category set C={ (x1, y1), (x2, y2) .., (xn, yn) }, wherein x1, x2 .., xn are
The tired correlation properties data of n sportsman, y1, y2 .., yn are classified for corresponding fatigue；
Step 2.3, the training dataset in sports fatigue early warning stage is obtained, i.e., data set known to tired classification；
Step 2.4, according to known tired category set, the conditional probability that statistics obtains each characteristic attribute under of all categories is estimated
Meter, i.e.,：P(a1|y1),P(a2|y1),...,P(am|y1)；P(a1|y2),P(a2|y2),...,P(am|y2)；...；P(a1|
yn),P(a2|yn),...,P(am|yn)；
If each characteristic attribute is conditional sampling, following derivation is had according to Bayes' theorem：Some sportsman's treats point
Class data x may be that probability P (yi | x)=P (x | yi) P (yi)/P (x), wherein P (yi) of yi level of fatigue goes out for level of fatigue
Existing probability can obtain according to known tired category set data statistics, and P (x | yi) occur for x in the data of yi level of fatigue
Probability；
For arbitrary x, denominator P (x) is fixed value, P (x | yi) P (yi)=P (a1 | yi) P (a2 | yi) ... P (am | yi) P
(yi), i.e.,
Finally calculate the probability that x is each classification：P(y1|x),P(y2|x),..,P(yn|x)；
Step 2.5, if P (yk | x)=max { P (y1 | x), P (y2 | x) .., P (yn | x) }, then the classification of x is i.e. given for yk
Level of fatigue.
2. the exercise load monitoring based on training data and sports fatigue method for early warning as described in claim 1, special
Sign is：The number of Map nodes is determined according to the number of training sample described in step 1.3.
3. exercise load monitoring and sports fatigue early warning system based on training data, which is characterized in that including such as lower die
Block：
Exercise load monitoring modular, the stage is monitored for exercise load, and subject is calculated using neural network concurrent optimization method
Energy consumption, rule that the energy consumption refers to subject motion's intensity and amount of exercise changes with time, including following submodule；
Estimation of energy consumption training sample setting up submodule, for establishing Estimation of energy consumption training sample according to different types of motor pattern
Database, each sample data include ages of players, gender, height, weight, three dimensions acceleration (ax,ay,
az), totally 7 components；
Netinit submodule establishes Estimation of energy consumption BP neural network for being directed to each motor pattern, and at the beginning of carrying out network
Beginningization, initialization package contain maximum frequency of training, learn precision, Hidden nodes, initial weight, threshold value, initial learning rate；
Training sample distribution sub module, for inputting Estimation of energy consumption training sample X1,X2,……,Xk, Xk=[xk1,xk2,xk3,
xk4,xk5,xk6,xk7], wherein k represents training sample number, and 7 components correspond respectively to age, gender, height, weight and three
The acceleration of a dimension, and training sample is evenly dispersed in each Map nodes；
Map node weights output sub-modules, for reading in the network preserved in Hadoop distributed file systems using Map functions
Weights record, i.e. initial weight, instantiate the neural network on each Map nodes according to the network weight of reading, and obtain every
The weights of a Map nodes and output；
Network weight updates submodule, for reading in the network preserved in Hadoop distributed file systems using Reduce functions
Weights record, and receive the weights of Map outputs, and the arithmetic average of the output weights according to each Map nodes, as new
Network weight is updated；Calculate updated network weight and the network read in from Hadoop distributed file systems simultaneously
Difference between weights judges whether to need to recycle next time according to difference；If network weight no longer updates, Estimation of energy consumption
The training of BP neural network algorithm terminates；
Energy consumption testing submodule is believed for being obtained the energy consumption of subject using trained Estimation of energy consumption BP neural network algorithm
Breath；
Sports fatigue warning module for the sports fatigue early warning stage, moves subject using Bayesian Classification Arithmetic
Giving fatigue pre-warning, including following submodule；
Characteristic attribute acquisition submodule, if x={ a1, a2 .., am } is sportsman's data to be sorted, wherein ai is x's
Ith feature attribute, common m characteristic attribute；The acquisition modes of characteristic attribute are：By the time sequence in the exercise load monitoring stage
Row are divided into m section, obtain the corresponding fatigue data in each section；
Fatigue classification submodule, it is known that tired category set C={ (x1, y1), (x2, y2) .., (xn, yn) }, wherein x1,
X2 .., xn are the tired correlation properties data of n sportsman, and y1, y2 .., yn are classified for corresponding fatigue；
Sports fatigue training dataset acquisition submodule, for according to known tired category set, statistics to be obtained of all categories
Under each characteristic attribute conditional probability estimation, i.e.,：P(a1|y1),P(a2|y1),...,P(am|y1)；P(a1|y2),P(a2|
y2),...,P(am|y2)；...；P(a1|yn),P(a2|yn),...,P(am|yn)；
If each characteristic attribute is conditional sampling, following derivation is had according to Bayes' theorem：Some sportsman's treats point
Class data x may be that probability P (yi | x)=P (x | yi) P (yi)/P (x), wherein P (yi) of yi level of fatigue goes out for level of fatigue
Existing probability can obtain according to known tired category set data statistics, and P (x | yi) occur for x in the data of yi level of fatigue
Probability；
For arbitrary x, denominator P (x) is fixed value, P (x | yi) P (yi)=P (a1 | yi) P (a2 | yi) ... P (am | yi) P
(yi), i.e.,
Finally calculate the probability that x is each classification：P(y1|x),P(y2|x),..,P(yn|x)；
Level of fatigue decision sub-module, if P (yk | x)=max { P (y1 | x), P (y2 | x) .., P (yn | x) }, then the classification of x
It is given level of fatigue for yk.
4. the exercise load monitoring based on training data and sports fatigue method for early warning as claimed in claim 3, special
Sign is：The number of Map nodes is determined according to the number of training sample described in training sample distribution sub module.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810074759.2A CN108211268B (en) | 2018-01-25 | 2018-01-25 | exercise load monitoring and exercise fatigue early warning method and system based on exercise training data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810074759.2A CN108211268B (en) | 2018-01-25 | 2018-01-25 | exercise load monitoring and exercise fatigue early warning method and system based on exercise training data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108211268A true CN108211268A (en) | 2018-06-29 |
| CN108211268B CN108211268B (en) | 2019-12-10 |
Family
ID=62669000
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810074759.2A Active CN108211268B (en) | 2018-01-25 | 2018-01-25 | exercise load monitoring and exercise fatigue early warning method and system based on exercise training data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108211268B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109126101A (en) * | 2018-10-30 | 2019-01-04 | 重庆邮电大学 | Cycling energy consumption calculation system and method |
| CN111061952A (en) * | 2019-12-17 | 2020-04-24 | 广东工业大学 | Intelligent automatic counterweight method, system and equipment based on deep learning |
| CN112037884A (en) * | 2020-09-14 | 2020-12-04 | 成都拟合未来科技有限公司 | Fitness exercise training plan generation evaluation method, system, terminal and medium |
| CN113082665A (en) * | 2021-05-06 | 2021-07-09 | 山东体育学院 | Infant's early education is experienced device with sports teaching |
| CN113297898A (en) * | 2021-03-12 | 2021-08-24 | 李涛 | Automatic mental state identification method based on Bayesian analysis method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106022521A (en) * | 2016-05-19 | 2016-10-12 | 四川大学 | Hadoop framework-based short-term load prediction method for distributed BP neural network |
| CN106650224A (en) * | 2016-10-31 | 2017-05-10 | 华南理工大学 | Remote monitoring available bionic rehabilitation exoskeleton system and control method thereof |
| CN106897802A (en) * | 2017-04-07 | 2017-06-27 | 华为技术有限公司 | Data processing method, device and body-building machine people |
| CN107007291A (en) * | 2017-04-05 | 2017-08-04 | 天津大学 | Intense strain intensity identifying system and information processing method based on multi-physiological-parameter |
| CN107334466A (en) * | 2017-08-08 | 2017-11-10 | 西安交通大学 | A kind of apparatus and method of wearable chronic disease intelligent monitoring and early warning |
-
2018
- 2018-01-25 CN CN201810074759.2A patent/CN108211268B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106022521A (en) * | 2016-05-19 | 2016-10-12 | 四川大学 | Hadoop framework-based short-term load prediction method for distributed BP neural network |
| CN106650224A (en) * | 2016-10-31 | 2017-05-10 | 华南理工大学 | Remote monitoring available bionic rehabilitation exoskeleton system and control method thereof |
| CN107007291A (en) * | 2017-04-05 | 2017-08-04 | 天津大学 | Intense strain intensity identifying system and information processing method based on multi-physiological-parameter |
| CN106897802A (en) * | 2017-04-07 | 2017-06-27 | 华为技术有限公司 | Data processing method, device and body-building machine people |
| CN107334466A (en) * | 2017-08-08 | 2017-11-10 | 西安交通大学 | A kind of apparatus and method of wearable chronic disease intelligent monitoring and early warning |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109126101A (en) * | 2018-10-30 | 2019-01-04 | 重庆邮电大学 | Cycling energy consumption calculation system and method |
| CN111061952A (en) * | 2019-12-17 | 2020-04-24 | 广东工业大学 | Intelligent automatic counterweight method, system and equipment based on deep learning |
| CN112037884A (en) * | 2020-09-14 | 2020-12-04 | 成都拟合未来科技有限公司 | Fitness exercise training plan generation evaluation method, system, terminal and medium |
| CN112037884B (en) * | 2020-09-14 | 2023-12-05 | 成都拟合未来科技有限公司 | Exercise training plan generation evaluation method, system, terminal and medium |
| CN113297898A (en) * | 2021-03-12 | 2021-08-24 | 李涛 | Automatic mental state identification method based on Bayesian analysis method |
| CN113082665A (en) * | 2021-05-06 | 2021-07-09 | 山东体育学院 | Infant's early education is experienced device with sports teaching |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108211268B (en) | 2019-12-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108211268A (en) | Exercise load monitoring and sports fatigue method for early warning and system based on training data | |
| CN106980899A (en) | The deep learning model and system of flow characteristic on prediction vascular tree blood flow paths | |
| CN111986811A (en) | Disease prediction system based on big data | |
| CN108304970A (en) | The method for quick predicting and system of Apple, air conditioned storage monitoring system | |
| CN114822763A (en) | Personalized exercise prescription recommendation method driven by exercise data | |
| CN110070116A (en) | Segmented based on the tree-shaped Training strategy of depth selects integrated image classification method | |
| CN110163131A (en) | Mix the human action classification method of convolutional neural networks and the optimization of microhabitat grey wolf | |
| CN116110597B (en) | Digital twinning-based intelligent analysis method and device for patient disease categories | |
| CN109567313B (en) | Intelligent insole with biological characteristic recognition function | |
| Lin et al. | Breast cancer prediction based on K-Means and SOM Hybrid Algorithm | |
| CN110287896A (en) | A kind of Human bodys' response method based on heterogeneous layering PSO and SVM | |
| CN109583272B (en) | Footprint system capable of acquiring living state of human body | |
| Yu et al. | Evaluation of sports training effect based on GABP neural network and artificial intelligence | |
| Chu et al. | [Retracted] Image Recognition of Badminton Swing Motion Based on Single Inertial Sensor | |
| CN105138835B (en) | Human body composition Forecasting Methodology based on physiologic information entropy | |
| CN111863187A (en) | Method, system, terminal and storage medium for recommending sports scheme | |
| CN107273262A (en) | The Forecasting Methodology and system of a kind of hardware event | |
| Li et al. | Construction and simulation of a multiattribute training data mining model for basketball players based on big data | |
| CN115147768A (en) | Fall risk assessment method and system | |
| CN107563327A (en) | It is a kind of that the pedestrian fed back recognition methods and system again are walked based on oneself | |
| Jing et al. | Optimization of track and field training methods based on SSA-BP and its effect on athletes' explosive power | |
| CN110335680A (en) | A kind of national physique health analysis method based on ɑ-RIPPER classifier | |
| Bülbül | Performance of different membership functions in stress classification with fuzzy logic | |
| Wang et al. | An intelligent data analysis-based medical management method for lower limb health of football athletes | |
| Li et al. | Discrete dynamic modeling analysis of data mining algorithm under the background of big data in the strategic goal of sustainable development of college physical training |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |