CN109276255A - A kind of limb tremor detection method and device - Google Patents
A kind of limb tremor detection method and device Download PDFInfo
- Publication number
- CN109276255A CN109276255A CN201811422696.1A CN201811422696A CN109276255A CN 109276255 A CN109276255 A CN 109276255A CN 201811422696 A CN201811422696 A CN 201811422696A CN 109276255 A CN109276255 A CN 109276255A
- Authority
- CN
- China
- Prior art keywords
- identification model
- limbs
- signal
- identified
- limbs signal
- 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
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
- A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
- A61B5/1101—Detecting tremor
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Abstract
The embodiment of the invention provides a kind of limb tremor detection method and device, and the present invention relates to field of artificial intelligence, method includes: building and training identification model, and whether limbs signal has feature of trembling to identification model for identification；Obtain the limbs signal to be identified of the first user of the sensor acquisition in mobile terminal；The limbs signal to be identified is inputted into trained identification model；It obtains identification model to identify limbs signal to be identified according to trained network parameter and export recognition result, and sends recognition result to mobile terminal.Technical solution provided in an embodiment of the present invention is able to solve the problem that limb tremor accuracy in detection is low in the prior art.
Description
[technical field]
The present invention relates to field of artificial intelligence more particularly to a kind of limb tremor detection method and device.
[background technique]
Currently, as aging of population is gradually increasing, some people with hyperthyroidism, Parkinson or some other special disease
The symptom that will appear the limb tremors such as hand, head, leg is brought the daily life, work, social interaction etc. of patient many
It is inconvenient.Existing limb tremor detection method is generally required and is detected by the medical treatment detection device of profession, or passes through one
A little aided-detection devices are detected, and testing result accuracy is low.
Therefore, limb tremor accuracy in detection how is improved as current urgent problem to be solved.
[summary of the invention]
In view of this, the embodiment of the invention provides a kind of limb tremor detection method and device, to solve existing skill
The low problem of limb tremor accuracy in detection in art.
To achieve the goals above, according to an aspect of the invention, there is provided a kind of limb tremor detection method, described
Method includes:
Identification model is constructed and trains, whether limbs signal has feature of trembling to the identification model for identification；It obtains
The limbs signal to be identified of first user of the sensor acquisition in mobile terminal；Into trained identification model described in input
Limbs signal to be identified；It obtains the identification model and the limbs signal to be identified and defeated is identified according to trained network parameter
Recognition result out, and the recognition result is sent to the mobile terminal.
Further, it is described building and training identification model, comprising: obtain multiple healthy human bodies the first limbs signal and
Second limbs signal of multiple limb tremor patient with sympotoms；According to preset format to the multiple first limbs signal and described more
A second limbs signal carries out sample production respectively, obtains the training set including multiple training samples；Construct the identification model；
The convolutional neural networks that the training set is inputted to the identification model obtain the forward direction output of the convolutional neural networks；Benefit
Weight and the biasing that the convolutional neural networks are updated with reverse transmittance nerve network algorithm obtain the trained identification mould
Type, and save the network parameter of the trained identification model.
Further, it is described using reverse transmittance nerve network algorithm update the convolutional neural networks weight and partially
It sets, the method for obtaining the trained identification model, comprising:
Loss function, the table of the loss function are constructed according to the legitimate reading of forward direction output and the training sample
It is up to formulaWherein, ElossThe loss function is indicated, described in n expression
Training sample sum, xiIndicate the forward direction output of i-th of training sample, yiExpression and xiCorresponding i-th of training sample it is true
Real result；Weight and the biasing that convolutional neural networks are updated using the back-propagation algorithm declined based on batch gradient, to described
Loss function, which optimizes, to be minimized it, and the trained identification model is obtained.
Further, it is described obtain mobile terminal in sensor acquisition the first user limbs signal to be identified it
Afterwards, and before inputting the limbs signal to be identified into the trained identification model, the method also includes:
The limbs signal to be identified is pre-processed, the pretreatment is low including being filtered out using Kalman filter algorithm
The human motion signal of frequency.
Further, the sensor includes polyaxial acceleration transducer, polyaxial gyroscope, in polyaxial inclinator
At least one；The mobile terminal is any one in mobile phone, iPad, smartwatch or wearable smart machine.
Further, it is described obtain mobile terminal in sensor acquisition the first user limbs signal to be identified it
Before, the method also includes:
The preset sample frequency of the sensor is set, so that the sensor is adopted according to the preset sample frequency
Sample obtains the actual samples data of default sampling duration；Actually adopting for the sensor is calculated according to the actual samples data
Sample frequency；Judge whether the frequency error value between the actual samples frequency and the preset sample frequency exceeds default error
Range；When the frequency error value is in the default error range, the preset sample frequency is improved step by step, until measuring
The actual samples frequency and the preset sample frequency between frequency error value exceed the default error range, will work as
Maximum sample frequency of the preceding preset sample frequency as the sensor；It is used with the maximum sample frequency acquisition described first
The limbs signal to be identified at family.
To achieve the goals above, according to an aspect of the invention, there is provided a kind of limb tremor detection device, described
Device includes: construction unit, and for constructing and training identification model, whether limbs signal has the identification model for identification
It trembles feature；Acquiring unit, the limbs signal to be identified of the first user for obtaining the acquisition of the sensor in mobile terminal；It is defeated
Enter unit, for inputting the limbs signal to be identified into trained identification model；Transmission unit, for obtaining the knowledge
Other model identifies the limbs signal to be identified according to trained network parameter and exports recognition result, and sends the identification
As a result to the mobile terminal.
Further, the construction unit includes: acquisition subelement, and the first limbs for obtaining multiple healthy human bodies are believed
Number and multiple limb tremor patient with sympotoms the second limbs signal；Subelement is made, is used for according to preset format to the multiple
First limbs signal and the multiple second limbs signal carry out sample production respectively, obtain the training including multiple training samples
Collection；Subelement is constructed, for constructing the identification model；Training subelement, for the training set to be inputted the identification mould
The convolutional neural networks of type obtain the forward direction output of the convolutional neural networks；Subelement is updated, for utilizing backpropagation mind
Weight and the biasing that the convolutional neural networks are updated through network algorithm obtain the trained identification model, and save instruction
The network parameter for the identification model perfected.
To achieve the goals above, according to an aspect of the invention, there is provided a kind of computer non-volatile memories are situated between
Matter, the storage medium include the program of storage, control equipment where the storage medium in described program operation and execute
The limb tremor detection method stated.
To achieve the goals above, according to an aspect of the invention, there is provided a kind of server, including memory and place
Device is managed, the memory is used to control the execution of program instruction, institute for storing the information including program instruction, the processor
State the step of above-mentioned limb tremor detection method is realized when program instruction is loaded and executed by processor.
In the present solution, limbs signal to be identified by obtaining the sensor acquisition in customer mobile terminal, and by depth
The identification model of degree study identifies limbs signal to be identified, to judge to identify whether limbs signal has feature of trembling, to remind
Whether user suffers from limb tremor disease.Limbs signal to be identified is identified by the identification model of deep learning, and whole process is simple
Quickly, limb tremor accuracy in detection is improved.
[Detailed description of the invention]
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below will be to needed in the embodiment attached
Figure is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for this field
For those of ordinary skill, without any creative labor, it can also be obtained according to these attached drawings other attached
Figure.
Fig. 1 is a kind of flow chart of limb tremor detection method according to an embodiment of the present invention；
Fig. 2 is a kind of schematic diagram of limb tremor detection device according to an embodiment of the present invention；
Fig. 3 is a kind of schematic diagram of server according to an embodiment of the present invention.
[specific embodiment]
For a better understanding of the technical solution of the present invention, being retouched in detail to the embodiment of the present invention with reference to the accompanying drawing
It states.
It will be appreciated that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.Base
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts it is all its
Its embodiment, shall fall within the protection scope of the present invention.
The term used in embodiments of the present invention is only to be not intended to be limiting merely for for the purpose of describing particular embodiments
The present invention.In the embodiment of the present invention and the "an" of singular used in the attached claims, " described " and "the"
It is also intended to including most forms, unless the context clearly indicates other meaning.
It should be appreciated that term "and/or" used herein is only a kind of incidence relation for describing affiliated partner, indicate
There may be three kinds of relationships, for example, A and/or B, can indicate: individualism A, exist simultaneously A and B, individualism B these three
Situation.In addition, character "/" herein, typicallys represent the relationship that forward-backward correlation object is a kind of "or".
It will be appreciated that though terminal may be described using term first, second, third, etc. in embodiments of the present invention,
But these terminals should not necessarily be limited by these terms.These terms are only used to for terminal being distinguished from each other out.For example, not departing from the present invention
In the case where scope of embodiments, first terminal can also be referred to as second terminal, and similarly, second terminal can also be referred to as
One terminal.
Depending on context, word as used in this " if " can be construed to " ... when " or " when ...
When " or " in response to determination " or " in response to detection ".Similarly, depend on context, phrase " if it is determined that " or " if detection
(condition or event of statement) " can be construed to " when determining " or " in response to determination " or " when the detection (condition of statement
Or event) when " or " in response to detection (condition or event of statement) ".
Fig. 1 is a kind of flow chart of limb tremor detection method according to an embodiment of the present invention, as shown in Figure 1, this method
Include:
Step S101 constructs and trains identification model, and whether limbs signal has feature of trembling to identification model for identification.
Step S102 obtains the limbs signal to be identified of the first user of the sensor acquisition in mobile terminal.
Wherein, mobile terminal includes any one in mobile phone, ipad, smartwatch or wearable smart machine.It passes
Sensor includes at least one of polyaxial acceleration transducer, polyaxial gyroscope, polyaxial inclinator.It is to be appreciated that
It is just detected whether by the limb tremor signal of the portable mobile terminal acquisition of human body with limb tremor symptom, Ke Yisui
When detect everywhere, help user to understand itself physical condition in real time.
Step S103 inputs limbs signal to be identified into trained identification model.
Step S104 obtains identification model and identifies limbs signal to be identified according to trained network parameter and export identification
As a result, and sending recognition result to mobile terminal.It allows the user to rapidly and accurately know whether oneself suffers from disease of trembling.
In the present solution, limbs signal to be identified by obtaining the sensor acquisition in customer mobile terminal, and by depth
The identification model of degree study identifies limbs signal to be identified, to judge to identify whether limbs signal has feature of trembling, to remind
Whether user suffers from limb tremor disease.Limbs signal to be identified is identified by the identification model of deep learning, and whole process is simple
Quickly, limb tremor accuracy in detection is improved.
Optionally, it constructs and trains identification model, comprising:
Obtain the first limbs signal of multiple healthy human bodies and the second limbs signal of multiple limb tremor patient with sympotoms.This
In embodiment, the first limbs signal is the normal signal for being diagnosed as the human body of health, and the second limbs signal is to be diagnosed as suffering from
There is the signal that trembles of the human body of limb tremor symptom.Using two different limbs signals as the basic number of training identification model
According to making identification model deep learning.
Limbs signal includes the parameters such as shake frequency, amplitude, frequency spectrum, and tremor tardus (3 can be divided into according to the difference of its frequency by trembling
~5Hz) and tremble fastly (6~12Hz), amplitude indicates the amplitude trembled, and frequency spectrum is the distribution curve of frequency.Limbs signal number energy
It is enough in whether characterization human body has limb tremor feature.
According to preset format sample production is carried out to multiple first limbs signals and multiple second limbs signals respectively, obtained
Training set including multiple training samples.Specifically, classified by different labels to limbs signal (normal and morbid state)
Mark.
Identification model is constructed, identification model includes convolutional neural networks；
By the convolutional neural networks of training set input identification model, the forward direction output of convolutional neural networks is obtained；
Weight and the biasing that convolutional neural networks are updated using reverse transmittance nerve network algorithm, obtain trained identification
Model, and save the network parameter of trained identification model.
Specifically, loss function is constructed to the legitimate reading (label) of output and training sample according to preceding, loss function
Expression formula isWherein, ElossIndicate that loss function, n indicate training sample
This sum, xiIndicate the forward direction output of i-th of training sample, yiExpression and xiThe true knot of corresponding i-th of training sample
Fruit；
Weight and the biasing that convolutional neural networks are updated using the back-propagation algorithm declined based on batch gradient, to loss
Function, which optimizes, to be minimized it, and trained identification model is obtained.
Optionally, identification model is constructed, identification model includes convolutional neural networks and long Memory Neural Networks in short-term；
By the convolutional neural networks and long Memory Neural Networks in short-term of training set input identification model, identification model is obtained
Export result；
Loss function is constructed using the output result of identification model and the label of limbs signal, and iteration loss function is to most
Smallization obtains trained identification model and its network parameter.
Specifically, loss function intersects entropy loss (binary cross entropyloss) function, tool using two classification
Body, the probability distribution of two categories is p and q in training set, and wherein p is true distribution, the non-genuine distribution of q.Using cross entropy
Loss function measures the similitude between two probability distribution.For a stochastic variable X, its all possible values
The expectation (E [I (x)]=log (1/p)) of information content is known as entropy.
Two classification cross entropies are as follows:Wherein, p (x) is really to be distributed, and q (x) is logical
Cross the estimated probability of data calculating.It should be noted that the value of cross entropy is bigger, this difference degree is also bigger.
Carry out iteration two using Adam gradient descent method and classify cross entropy loss function to minimizing, obtains trained identification
Model；Obtain the network parameter of the convolutional neural networks and long Memory Neural Networks in short-term in trained identification model.
Optionally, after obtaining limbs signal to be identified, and limbs to be identified are being inputted into trained identification model
Before signal, method further include:
Limbs signal to be identified is pre-processed, pretreatment includes the human body that low frequency is filtered out using Kalman filter algorithm
Motor message.
Optionally, the limbs to be identified acquired by mobile device built-in sensors are inputted into trained identification model
It trembles before signal, method further include:
The preset sample frequency of sensor is set, so that sensor is sampled according to preset sample frequency, is preset
Sample the actual samples data of duration；The actual samples frequency of sensor is calculated according to actual samples data；Judge actual samples
Whether the frequency error value between frequency and preset sample frequency exceeds default error range；
When frequency error value is in default error range, preset sample frequency is improved step by step, until what is measured actually adopts
Frequency error value between sample frequency and preset sample frequency exceeds default error range, using current preset sample frequency as
The maximum sample frequency of sensor；The limbs signal to be identified of the first user is acquired with maximum sample frequency.It is to be appreciated that logical
The maximum sample frequency of sensor is crossed to acquire limb tremor signal to be identified, recognition efficiency can be effectively improved.
The embodiment of the invention provides a kind of limb tremor detection device, the limb tremor detection device is above-mentioned for executing
Limb tremor detection method, as shown in Fig. 2, the device includes: construction unit 10, acquiring unit 20, input unit 30, sends list
Member 40.
Construction unit 10, for constructing and training identification model, whether limbs signal has shake to identification model for identification
It quivers feature.
Acquiring unit 20, the limbs signal to be identified of the first user for obtaining the acquisition of the sensor in mobile terminal；
Wherein, mobile terminal includes any one in mobile phone, ipad, smartwatch or wearable smart machine.It passes
Sensor includes at least one of polyaxial acceleration transducer, polyaxial gyroscope, polyaxial inclinator.It is to be appreciated that
It is just detected whether by the limb tremor signal of the portable mobile terminal acquisition of human body with limb tremor symptom, Ke Yisui
When detect everywhere, help user to understand itself physical condition in real time.
Input unit 30, for inputting limbs signal to be identified into trained identification model.
Transmission unit 40 identifies limbs signal to be identified and defeated according to trained network parameter for obtaining identification model
Recognition result out, and recognition result is sent to mobile terminal.It allows the user to rapidly and accurately know whether oneself suffers from shake
It quivers disease.
In the present solution, limbs signal to be identified by obtaining the sensor acquisition in customer mobile terminal, and by depth
The identification model of degree study identifies limbs signal to be identified, to judge to identify whether limbs signal has feature of trembling, to remind
Whether user suffers from limb tremor disease.Limbs signal to be identified is identified by the identification model of deep learning, and whole process is simple
Quickly, limb tremor accuracy in detection is improved.
Optionally, construction unit 10 includes obtaining subelement, production subelement, the first building subelement, the first training
Unit, first update subelement.
Subelement is obtained, for obtaining the first limbs signal and multiple limb tremor patient with sympotoms of multiple healthy human bodies
Second limbs signal.
In the present embodiment, the first limbs signal is the normal signal for being diagnosed as the human body of health, and the second limbs signal is
It is diagnosed as the signal that trembles of the human body with limb tremor symptom.Using two different limbs signals as training identification model
Master data, make identification model deep learning.
Limbs signal includes the parameters such as shake frequency, amplitude, frequency spectrum, and tremor tardus (3 can be divided into according to the difference of its frequency by trembling
~5Hz) and tremble fastly (6~12Hz), amplitude indicates the amplitude trembled, and frequency spectrum is the distribution curve of frequency.Limbs signal number energy
It is enough in whether characterization human body has limb tremor feature.
Make subelement, for according to preset format to multiple first limbs signals and multiple second limbs signals respectively into
The production of row sample, obtains the training set including multiple training samples.Specifically, (normal to limbs signal by different labels
And morbid state) carry out classification annotation.
First building subelement, for constructing identification model, identification model includes convolutional neural networks；
First training subelement, for obtaining convolutional Neural net for the convolutional neural networks of training set input identification model
The forward direction of network exports；
First updates subelement, for the weight using reverse transmittance nerve network algorithm update convolutional neural networks and partially
It sets, obtains trained identification model, and save the network parameter of trained identification model.
Specifically, by the convolutional neural networks of training set input identification model, the forward direction output of convolutional neural networks is obtained；
Weight and the biasing that convolutional neural networks are updated using reverse transmittance nerve network algorithm, obtain trained identification
Model, and save the network parameter of trained identification model.
Specifically, loss function is constructed to the legitimate reading (label) of output and training sample according to preceding, loss function
Expression formula isWherein, ElossIndicate that loss function, n indicate training sample
Sum, xiIndicate the forward direction output of i-th of training sample, yiExpression and xiThe legitimate reading of corresponding i-th of training sample；
Weight and the biasing that convolutional neural networks are updated using the back-propagation algorithm declined based on batch gradient, to loss
Function, which optimizes, to be minimized it, and trained identification model is obtained.
Optionally, construction unit 10 further includes the second building subelement, the second training subelement, the second update subelement；
Second building subelement, for constructing identification model, identification model includes convolutional neural networks and long short-term memory
Neural network；
Second training subelement, for the convolutional neural networks of training set input identification model and long short-term memory are neural
Network obtains the output result of identification model；
Second updates subelement, the label building loss letter for the output result and limbs signal using identification model
Number, and iteration loss function obtains trained identification model and its network parameter to minimizing.
Specifically, loss function intersects entropy loss (binary cross entropyloss) function, tool using two classification
Body, the probability distribution of two categories is p and q in training set, and wherein p is true distribution, the non-genuine distribution of q.Using cross entropy
Loss function measures the similitude between two probability distribution.For a stochastic variable X, its all possible values
The expectation (E [I (x)]=log (1/p)) of information content is known as entropy.
Two classification cross entropies are as follows:Wherein, p (x) is really to be distributed, and q (x) is logical
Cross the estimated probability of data calculating.It should be noted that the value of cross entropy is bigger, this difference degree is also bigger.
Carry out iteration two using Adam gradient descent method and classify cross entropy loss function to minimizing, obtains trained identification
Model；Obtain the network parameter of the convolutional neural networks and long Memory Neural Networks in short-term in trained identification model.
Optionally, device further includes processing unit, and for pre-processing limbs signal to be identified, pretreatment includes adopting
The human motion signal of low frequency is filtered out with Kalman filter algorithm.
Optionally, device further includes setting unit, computing unit, test cell, acquisition unit.
Setting unit, for the preset sample frequency of sensor to be arranged, so that sensor is carried out according to preset sample frequency
Sampling obtains the actual samples data of default sampling duration；
Computing unit, for calculating the actual samples frequency of sensor according to actual samples data；Judge actual samples frequency
Whether the frequency error value between rate and preset sample frequency exceeds default error range；
Test cell, for improving preset sample frequency step by step when frequency error value is in default error range, until
Frequency error value between the actual samples frequency measured and preset sample frequency exceeds default error range, and current is preset
Maximum sample frequency of the sample frequency as sensor；
Acquisition unit, for acquiring the limb tremor signal to be identified of the first user with maximum sample frequency.It is understood that
Ground is acquired limb tremor signal to be identified by the maximum sample frequency of sensor, can effectively improve recognition efficiency.
The embodiment of the invention provides a kind of computer non-volatile memory medium, storage medium includes the program of storage,
Wherein, when program is run, equipment where control storage medium executes following steps:
Identification model is constructed and trains, whether limbs signal has feature of trembling to identification model for identification；Obtain movement
The limbs signal to be identified of first user of the sensor acquisition in terminal；Limb to be identified is inputted into trained identification model
Body signal；It obtains identification model to identify limbs signal to be identified according to trained network parameter and export recognition result, concurrently
Send recognition result to mobile terminal.
Optionally, when program is run, equipment where control storage medium also executes following steps: obtaining multiple Healthy Peoples
First limbs signal of body and the second limbs signal of multiple limb tremor patient with sympotoms；According to preset format to multiple first limbs
Body signal and multiple second limbs signals carry out sample production respectively, obtain the training set including multiple training samples；Building is known
Other model；By the convolutional neural networks of training set input identification model, the forward direction output of convolutional neural networks is obtained；Using reversed
Propagation Neural Network algorithm updates weight and the biasing of convolutional neural networks, obtains trained identification model, and save training
The network parameter of good identification model.
Optionally, when program is run, equipment where control storage medium also executes following steps: according to preceding to output and instruction
The legitimate reading for practicing sample constructs loss function, and the expression formula of loss function is
Wherein, ElossIndicate that loss function, n indicate training sample sum, xiIndicate the forward direction output of i-th of training sample, yiIndicate with
xiThe legitimate reading of corresponding i-th of training sample；Convolution is updated using the back-propagation algorithm declined based on batch gradient
The weight of neural network and biasing, optimize loss function and minimize it, and obtain trained identification model.
Optionally, when program is run, equipment where control storage medium also executes following steps: limbs to be identified are believed
It number is pre-processed, pretreatment includes the human motion signal that low frequency is filtered out using Kalman filter algorithm.
Optionally, when program is run, equipment where control storage medium also executes following steps: the pre- of sensor is arranged
If sample frequency, so that sensor is sampled according to preset sample frequency, the actual samples data of default sampling duration are obtained；
The actual samples frequency of sensor is calculated according to actual samples data；Judge between actual samples frequency and preset sample frequency
Whether frequency error value exceeds default error range；When frequency error value is in default error range, default adopt is improved step by step
Sample frequency, until the frequency error value between the actual samples frequency measured and preset sample frequency exceeds default error range,
Using current preset sample frequency as the maximum sample frequency of sensor；With maximum sample frequency the first user of acquisition wait know
Other limbs signal.
The embodiment of the invention provides a kind of server 100, including memory 101 and processor 102, memory 101 is used
In the information that storage includes program instruction 103, processor 102 is used to control the execution of program instruction 103, and program instruction is processed
Device is loaded and is performed the steps of when executing
Identification model is constructed and trains, whether limbs signal has feature of trembling to identification model for identification；Obtain movement
The limbs signal to be identified of first user of the sensor acquisition in terminal；Limb to be identified is inputted into trained identification model
Body signal；It obtains identification model to identify limbs signal to be identified according to trained network parameter and export recognition result, concurrently
Send recognition result to mobile terminal.
Optionally, it is also performed the steps of when program instruction is loaded and executed by processor and obtains multiple healthy human bodies
Second limbs signal of the first limbs signal and multiple limb tremor patient with sympotoms；Multiple first limbs are believed according to preset format
Number and multiple second limbs signals carry out sample production respectively, obtain include multiple training samples training set；Building identification mould
Type；By the convolutional neural networks of training set input identification model, the forward direction output of convolutional neural networks is obtained；Utilize backpropagation
Neural network algorithm updates weight and the biasing of convolutional neural networks, obtains trained identification model, and save trained
The network parameter of identification model.
Optionally, it is also performed the steps of when program instruction is loaded and executed by processor according to preceding to output and training sample
This legitimate reading constructs loss function, and the expression formula of loss function is
Wherein, ElossIndicate that loss function, n indicate training sample sum, xiIndicate the forward direction output of i-th of training sample, yiIndicate with
xiThe legitimate reading of corresponding i-th of training sample；Convolution is updated using the back-propagation algorithm declined based on batch gradient
The weight of neural network and biasing, optimize loss function and minimize it, and obtain trained identification model.
Optionally, also performed the steps of when program instruction is loaded and executed by processor by limbs signal to be identified into
Row pretreatment, pretreatment include the human motion signal that low frequency is filtered out using Kalman filter algorithm.
Optionally, setting the default of sensor is also performed the steps of when program instruction is loaded and executed by processor to adopt
Sample frequency obtains the actual samples data of default sampling duration so that sensor is sampled according to preset sample frequency；According to
The actual samples frequency of actual samples data calculating sensor；Judge the frequency between actual samples frequency and preset sample frequency
Whether error amount exceeds default error range；When frequency error value is in default error range, default sampling frequency is improved step by step
Rate will be worked as until the frequency error value between the actual samples frequency measured and preset sample frequency exceeds default error range
Maximum sample frequency of the preceding preset sample frequency as sensor；The limb to be identified of the first user is acquired with maximum sample frequency
Body signal.
It should be noted that terminal involved in the embodiment of the present invention can include but is not limited to personal computer
(Personal Computer, PC), personal digital assistant (Personal Digital Assistant, PDA), wireless handheld
Equipment, tablet computer (Tablet Computer), mobile phone, MP3 player, MP4 player etc..
It is understood that the application can be mounted in the application program (nativeApp) in terminal, or may be used also
To be a web page program (webApp) of browser in terminal, the embodiment of the present invention is to this without limiting.
It is apparent to those skilled in the art that for convenience and simplicity of description, the system of foregoing description,
The specific work process of device and unit, can refer to corresponding processes in the foregoing method embodiment, and details are not described herein.
In several embodiments provided by the present invention, it should be understood that disclosed system, device and method can be with
It realizes by another way.For example, the apparatus embodiments described above are merely exemplary, for example, the unit
It divides, only a kind of logical function partition, there may be another division manner in actual implementation, for example, multiple units or group
Part can be combined or can be integrated into another system, or some features can be ignored or not executed.Another point, it is shown
Or the mutual coupling, direct-coupling or communication connection discussed can be through some interfaces, device or unit it is indirect
Coupling or communication connection can be electrical property, mechanical or other forms.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple
In network unit.It can select some or all of unit therein according to the actual needs to realize the mesh of this embodiment scheme
's.
It, can also be in addition, the functional units in various embodiments of the present invention may be integrated into one processing unit
It is that each unit physically exists alone, can also be integrated in one unit with two or more units.Above-mentioned integrated list
Member both can take the form of hardware realization, can also realize in the form of hardware adds SFU software functional unit.
The above-mentioned integrated unit being realized in the form of SFU software functional unit can store and computer-readable deposit at one
In storage media.Above-mentioned SFU software functional unit is stored in a storage medium, including some instructions are used so that a computer
It is each that device (can be personal computer, server or network equipment etc.) or processor (Processor) execute the present invention
The part steps of embodiment the method.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory (Read-
Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic or disk etc. it is various
It can store the medium of program code.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Within mind and principle, any modification, equivalent substitution, improvement and etc. done be should be included within the scope of the present invention.
Claims (10)
1. a kind of limb tremor detection method, which is characterized in that the described method includes:
Identification model is constructed and trains, whether limbs signal has feature of trembling to the identification model for identification；
Obtain the limbs signal to be identified of the first user of the sensor acquisition in mobile terminal；
The limbs signal to be identified is inputted into trained identification model；
The identification model is obtained to identify the limbs signal to be identified according to trained network parameter and export recognition result,
And the recognition result is sent to the mobile terminal.
2. the method according to claim 1, wherein the building and training identification model, comprising:
Obtain the first limbs signal of multiple healthy human bodies and the second limbs signal of multiple limb tremor patient with sympotoms；
Sample production is carried out respectively to the multiple first limbs signal and the multiple second limbs signal according to preset format,
Obtain the training set including multiple training samples；
Construct the identification model；
The convolutional neural networks that the training set is inputted to the identification model, the forward direction for obtaining the convolutional neural networks are defeated
Out；
Weight and the biasing that the convolutional neural networks are updated using reverse transmittance nerve network algorithm are obtained trained described
Identification model, and save the network parameter of the trained identification model.
3. according to the method described in claim 2, it is characterized in that, described using described in the update of reverse transmittance nerve network algorithm
The weight of convolutional neural networks and biasing, the method for obtaining the trained identification model, comprising:
Loss function, the expression formula of the loss function are constructed according to the legitimate reading of forward direction output and the training sample
ForWherein, ElossIndicate that the loss function, n indicate the training
Total sample number, xiIndicate the forward direction output of i-th of training sample, yiExpression and xiThe true knot of corresponding i-th of training sample
Fruit；
Weight and the biasing that convolutional neural networks are updated using the back-propagation algorithm declined based on batch gradient, to the loss
Function, which optimizes, to be minimized it, and the trained identification model is obtained.
4. described in any item methods according to claim 1~3, which is characterized in that the sensor obtained in mobile terminal
After the limbs signal to be identified of first user of acquisition, and it is described to be identified being inputted into the trained identification model
Before limbs signal, the method also includes:
The limbs signal to be identified is pre-processed, the pretreatment includes filtering out low frequency using Kalman filter algorithm
Human motion signal.
5. described in any item methods according to claim 1~3, which is characterized in that the sensor includes polyaxial acceleration
At least one of sensor, polyaxial gyroscope, polyaxial inclinator；The mobile terminal is mobile phone, iPad, smartwatch
Or any one in wearable smart machine.
6. the method according to claim 1, wherein the first of the sensor acquisition obtained in mobile terminal
Before the limbs signal to be identified of user, the method also includes:
The preset sample frequency of the sensor is set, so that the sensor is sampled according to the preset sample frequency,
Obtain the actual samples data of default sampling duration；
The actual samples frequency of the sensor is calculated according to the actual samples data；
Judge whether the frequency error value between the actual samples frequency and the preset sample frequency exceeds default error model
It encloses；
When the frequency error value is in the default error range, the preset sample frequency is improved step by step, until measuring
The actual samples frequency and the preset sample frequency between frequency error value exceed the default error range, will work as
Maximum sample frequency of the preceding preset sample frequency as the sensor；
The limbs signal to be identified of first user is acquired with the maximum sample frequency.
7. a kind of limb tremor detection device, which is characterized in that described device includes:
Construction unit, for constructing and training identification model, whether limbs signal has and trembles the identification model for identification
Feature；
Acquiring unit, the limbs signal to be identified of the first user for obtaining the acquisition of the sensor in mobile terminal；
Input unit, for inputting the limbs signal to be identified into trained identification model；
Transmission unit identifies the limbs signal to be identified simultaneously according to trained network parameter for obtaining the identification model
Recognition result is exported, and sends the recognition result to the mobile terminal.
8. device according to claim 7, which is characterized in that the construction unit includes:
Obtain subelement, for obtain multiple healthy human bodies the first limbs signal and multiple limb tremor patient with sympotoms second
Limbs signal；
Subelement is made, for dividing according to preset format the multiple first limbs signal and the multiple second limbs signal
Not carry out sample production, obtain include multiple training samples training set；
Subelement is constructed, for constructing the identification model；
Training subelement obtains the convolution mind for the training set to be inputted to the convolutional neural networks of the identification model
Forward direction output through network；
Subelement is updated, for updating weight and the biasing of the convolutional neural networks using reverse transmittance nerve network algorithm,
The trained identification model is obtained, and saves the network parameter of the trained identification model.
9. a kind of computer non-volatile memory medium, the storage medium includes the program of storage, which is characterized in that described
Equipment perform claim program controls the storage medium when running where requires the detection of limb tremor described in 1 to 6 any one
Method.
10. a kind of server, including memory and processor, the memory is for storing the information including program instruction, institute
Processor is stated for controlling the execution of program instruction, it is characterised in that: described program instruction is real when being loaded and executed by processor
The step of limb tremor detection method described in existing claim 1 to 6 any one.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811422696.1A CN109276255B (en) | 2018-11-27 | 2018-11-27 | Method and device for detecting tremor of limbs |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811422696.1A CN109276255B (en) | 2018-11-27 | 2018-11-27 | Method and device for detecting tremor of limbs |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109276255A true CN109276255A (en) | 2019-01-29 |
| CN109276255B CN109276255B (en) | 2023-02-28 |
Family
ID=65172826
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811422696.1A Active CN109276255B (en) | 2018-11-27 | 2018-11-27 | Method and device for detecting tremor of limbs |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109276255B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109949932A (en) * | 2019-03-28 | 2019-06-28 | 南开大学 | A kind of judgment method of the essential tremor extent based on machine learning |
| CN110037684A (en) * | 2019-04-01 | 2019-07-23 | 上海数创医疗科技有限公司 | Device based on the identification rhythm of the heart type for improving convolutional neural networks |
| CN110037683A (en) * | 2019-04-01 | 2019-07-23 | 上海数创医疗科技有限公司 | The improvement convolutional neural networks and its training method of rhythm of the heart type for identification |
| CN110522455A (en) * | 2019-09-26 | 2019-12-03 | 安徽中医药大学 | A kind of WD based on deep learning trembles grade appraisal procedure |
| CN110522456A (en) * | 2019-09-26 | 2019-12-03 | 安徽中医药大学 | A kind of WD based on deep learning trembles conditions of patients self-evaluating system |
| CN111292844A (en) * | 2020-01-21 | 2020-06-16 | 桂林医学院附属医院 | Parkinson disease condition monitoring system |
| CN112075940A (en) * | 2020-09-21 | 2020-12-15 | 哈尔滨工业大学 | Tremor detection system based on bidirectional long-time and short-time memory neural network |
| CN113229798A (en) * | 2021-05-18 | 2021-08-10 | 平安科技（深圳）有限公司 | Model migration training method and device, computer equipment and readable storage medium |
| CN113240018A (en) * | 2021-05-19 | 2021-08-10 | 哈尔滨医科大学 | Hand-drawn graph classification method and system based on error back propagation algorithm |
| CN113282203A (en) * | 2021-04-30 | 2021-08-20 | 深圳市联谛信息无障碍有限责任公司 | Interface switching method and device for user with tremor of limbs and electronic equipment |
| CN113855570A (en) * | 2021-09-30 | 2021-12-31 | 平安科技（深圳）有限公司 | Parkinson disease medicine taking reminding method and system, electronic equipment and storage medium |
| CN116757534A (en) * | 2023-06-15 | 2023-09-15 | 中国标准化研究院 | Intelligent refrigerator reliability analysis method based on neural training network |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050243061A1 (en) * | 2004-04-30 | 2005-11-03 | Hillcrest Communications, Inc. | Methods and devices for identifying users based on tremor |
| WO2017167518A1 (en) * | 2016-03-31 | 2017-10-05 | Fotonation Limited | Biometric recognition system |
| WO2017192028A1 (en) * | 2016-05-03 | 2017-11-09 | Universiti Sains Malaysia | Tremor motion quantifying method and system |
| CN107958257A (en) * | 2017-10-11 | 2018-04-24 | 华南理工大学 | A kind of Chinese traditional medicinal materials recognition method based on deep neural network |
| US20180113498A1 (en) * | 2015-03-25 | 2018-04-26 | Koninklijke Philips N.V. | Health wearable that automatically changes sensor reading timings |
| CN108197648A (en) * | 2017-12-28 | 2018-06-22 | 华中科技大学 | A kind of Fault Diagnosis Method of Hydro-generating Unit and system based on LSTM deep learning models |
| CN108875608A (en) * | 2018-06-05 | 2018-11-23 | 合肥湛达智能科技有限公司 | A kind of automobile traffic signal recognition method based on deep learning |
-
2018
- 2018-11-27 CN CN201811422696.1A patent/CN109276255B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050243061A1 (en) * | 2004-04-30 | 2005-11-03 | Hillcrest Communications, Inc. | Methods and devices for identifying users based on tremor |
| US20180113498A1 (en) * | 2015-03-25 | 2018-04-26 | Koninklijke Philips N.V. | Health wearable that automatically changes sensor reading timings |
| WO2017167518A1 (en) * | 2016-03-31 | 2017-10-05 | Fotonation Limited | Biometric recognition system |
| WO2017192028A1 (en) * | 2016-05-03 | 2017-11-09 | Universiti Sains Malaysia | Tremor motion quantifying method and system |
| CN107958257A (en) * | 2017-10-11 | 2018-04-24 | 华南理工大学 | A kind of Chinese traditional medicinal materials recognition method based on deep neural network |
| CN108197648A (en) * | 2017-12-28 | 2018-06-22 | 华中科技大学 | A kind of Fault Diagnosis Method of Hydro-generating Unit and system based on LSTM deep learning models |
| CN108875608A (en) * | 2018-06-05 | 2018-11-23 | 合肥湛达智能科技有限公司 | A kind of automobile traffic signal recognition method based on deep learning |
Non-Patent Citations (1)
| Title |
|---|
| 朱凤超: "基于大规模医疗数据的病症识别与预警", 《中国优秀博硕士学位论文全文数据库（硕士） 医药卫生科技辑》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109949932A (en) * | 2019-03-28 | 2019-06-28 | 南开大学 | A kind of judgment method of the essential tremor extent based on machine learning |
| CN110037684A (en) * | 2019-04-01 | 2019-07-23 | 上海数创医疗科技有限公司 | Device based on the identification rhythm of the heart type for improving convolutional neural networks |
| CN110037683A (en) * | 2019-04-01 | 2019-07-23 | 上海数创医疗科技有限公司 | The improvement convolutional neural networks and its training method of rhythm of the heart type for identification |
| CN110522455A (en) * | 2019-09-26 | 2019-12-03 | 安徽中医药大学 | A kind of WD based on deep learning trembles grade appraisal procedure |
| CN110522456A (en) * | 2019-09-26 | 2019-12-03 | 安徽中医药大学 | A kind of WD based on deep learning trembles conditions of patients self-evaluating system |
| CN111292844B (en) * | 2020-01-21 | 2023-03-21 | 桂林医学院附属医院 | Parkinson disease condition monitoring system |
| CN111292844A (en) * | 2020-01-21 | 2020-06-16 | 桂林医学院附属医院 | Parkinson disease condition monitoring system |
| CN112075940A (en) * | 2020-09-21 | 2020-12-15 | 哈尔滨工业大学 | Tremor detection system based on bidirectional long-time and short-time memory neural network |
| CN113282203A (en) * | 2021-04-30 | 2021-08-20 | 深圳市联谛信息无障碍有限责任公司 | Interface switching method and device for user with tremor of limbs and electronic equipment |
| CN113229798A (en) * | 2021-05-18 | 2021-08-10 | 平安科技（深圳）有限公司 | Model migration training method and device, computer equipment and readable storage medium |
| CN113229798B (en) * | 2021-05-18 | 2023-08-22 | 平安科技（深圳）有限公司 | Model migration training method, device, computer equipment and readable storage medium |
| CN113240018A (en) * | 2021-05-19 | 2021-08-10 | 哈尔滨医科大学 | Hand-drawn graph classification method and system based on error back propagation algorithm |
| CN113855570A (en) * | 2021-09-30 | 2021-12-31 | 平安科技（深圳）有限公司 | Parkinson disease medicine taking reminding method and system, electronic equipment and storage medium |
| CN116757534A (en) * | 2023-06-15 | 2023-09-15 | 中国标准化研究院 | Intelligent refrigerator reliability analysis method based on neural training network |
| CN116757534B (en) * | 2023-06-15 | 2024-03-15 | 中国标准化研究院 | Intelligent refrigerator reliability analysis method based on neural training network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109276255B (en) | 2023-02-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109276255A (en) | A kind of limb tremor detection method and device | |
| CN111027487B (en) | Behavior recognition system, method, medium and equipment based on multi-convolution kernel residual error network | |
| CN107153871B (en) | Falling detection method based on convolutional neural network and mobile phone sensor data | |
| CN105902257B (en) | Sleep state analysis method and device, intelligent wearable device | |
| KR101593315B1 (en) | Method, apparatus and computer program product for activity recognition | |
| CN107773214A (en) | A kind of method, computer-readable medium and the system of optimal wake-up strategy | |
| Pires et al. | Identification of activities of daily living using sensors available in off-the-shelf mobile devices: Research and hypothesis | |
| CN112057066A (en) | Heart rate detection method, wearable device and computer storage medium | |
| CN116821809B (en) | Vital sign data acquisition system based on artificial intelligence | |
| CN114707562A (en) | Electromyographic signal sampling frequency control method and device and storage medium | |
| CN113870239A (en) | Vision detection method and device, electronic equipment and storage medium | |
| US20220391697A1 (en) | Machine-learning based gesture recognition with framework for adding user-customized gestures | |
| CN114343574A (en) | Sleep quality evaluation method, client, server and storage medium | |
| Stewart et al. | Practical automated activity recognition using standard smartphones | |
| CN109886402A (en) | Deep learning model training method, device, computer equipment and storage medium | |
| Ntalampiras et al. | An incremental learning mechanism for human activity recognition | |
| CN116739037A (en) | Personality model construction method and device with personality characteristics | |
| CN112883939B (en) | Real-time falling detection model training method, real-time falling detection method and device | |
| CN109166628A (en) | The method and device of locomitivity detection, storage medium, processor | |
| CN107909501A (en) | The smell and correlating method of behavior, smell social contact method and device | |
| CN109965847B (en) | Server and signal analysis system | |
| CN113095153A (en) | Mobile terminal human situation recognition method based on depth residual error network | |
| CN108681412B (en) | Emotion recognition device and method based on array type touch sensor | |
| Dao et al. | Smart lifelogging: recognizing human activities using PHASOR | |
| Lentzas et al. | Evaluating state-of-the-art classifiers for human activity recognition using smartphones |
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 |