CN111990977A - Wearable optical fiber sensor is to human biological parameter monitoring devices based on neural network - Google Patents

Wearable optical fiber sensor is to human biological parameter monitoring devices based on neural network Download PDF

Info

Publication number
CN111990977A
CN111990977A CN202010762944.8A CN202010762944A CN111990977A CN 111990977 A CN111990977 A CN 111990977A CN 202010762944 A CN202010762944 A CN 202010762944A CN 111990977 A CN111990977 A CN 111990977A
Authority
CN
China
Prior art keywords
layer
output
neural network
input
optical fiber
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202010762944.8A
Other languages
Chinese (zh)
Inventor
陈勐勐
仇飞
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing Xiaozhuang University
Original Assignee
Nanjing Xiaozhuang University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nanjing Xiaozhuang University filed Critical Nanjing Xiaozhuang University
Priority to CN202010762944.8A priority Critical patent/CN111990977A/en
Publication of CN111990977A publication Critical patent/CN111990977A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/01—Measuring temperature of body parts ; Diagnostic temperature sensing, e.g. for malignant or inflamed tissue
    • 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
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/22—Ergometry; Measuring muscular strength or the force of a muscular blow
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/68—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
    • A61B5/6801—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
    • A61B5/6802—Sensor mounted on worn items
    • A61B5/6804—Garments; Clothes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent

Abstract

The invention provides a physiological parameter monitoring device of a wearable optical fiber sensor based on a neural network, which comprises a monitoring unit and a signal detection unit, wherein the monitoring unit monitors physiological signals and inputs the monitored signals to the signal detection unit to realize the monitoring of physiological parameters; the monitoring unit comprises a phase-sensitive optical time domain reflectometer system and a monitoring body; the phase-sensitive optical time domain reflectometer system comprises a phase-sensitive optical time domain reflectometer, a fan-in and a fan-out, wherein a detection signal of the phase-sensitive optical time domain reflectometer enters the monitoring body through the fan-in from a signal output port, and a scattering signal generated by the monitoring body is received by the phase-sensitive optical time domain reflectometer through the fan-out; the signal detection unit comprises a model establishing subunit, an application model subunit and an evaluation result subunit. The invention provides a physiological parameter monitoring device of a wearable optical fiber sensor based on a neural network, which can be used for intelligently feeding back and guiding the rehabilitation training of human bodies, the skill training of athletes and the like.

Description

Wearable optical fiber sensor is to human biological parameter monitoring devices based on neural network
Technical Field
The invention relates to the technical field of optical fiber sensing, in particular to a physiological parameter monitoring device of a wearable optical fiber sensor based on a neural network.
Background
The working principle of the optical fiber sensor is that the change of the external physical quantity can modulate a certain parameter of the light wave transmitted in the optical fiber, so that the parameter changes along with the change of the external physical quantity, and the modulated light signal is detected, so that the measured physical quantity is demodulated. In the optical fiber sensor, the optical fiber can be used as a propagation medium of optical waves, the optical waves propagated in the optical fiber are changed due to the change of external factors, and the optical fiber can be used as a sensing element to detect physical quantities such as temperature, vibration, strain, shape and the like, so that the detection, intelligent processing and application of the optical fiber sensor on the market to human biological parameters are rare; even consider when measuring the detection to the human body, need lay wire on the human body earlier, this kind of operation detects the trouble, wastes time and energy, and can't carry out autonomic study to the testing data, influences detection effect.
Disclosure of Invention
The invention aims to provide a physiological parameter monitoring device of a wearable optical fiber sensor based on a neural network, aiming at solving the problems, the neural network system for detecting the biological parameters of the human body based on the wearable optical fiber sensor can be used for detecting a large number of parameters such as various torsion, motion range, temperature and the like when the limbs, waist and even head of the human body move, and the range of the detection result can be intelligently applied to corresponding practical applications such as the rehabilitation training of the human body, the skill training of athletes and the like under the intelligent processing of the neural network system.
The adopted technical scheme is as follows: a physiological parameter monitoring device of a wearable optical fiber sensor based on a neural network comprises a monitoring unit and a signal detection unit, wherein the monitoring unit monitors physiological signals and inputs the monitored signals to the signal detection unit to realize the monitoring of physiological parameters;
the monitoring unit comprises a phase-sensitive optical time domain reflectometer system and a monitoring body; the monitoring body comprises an outer fabric layer, a waterproof layer is arranged at the bottom of the outer fabric layer, a carbon fiber layer is arranged at the bottom of the waterproof layer, and a plurality of optical fibers are embedded in the carbon fiber layer; the detection signal of the phase sensitive optical time domain reflectometer enters the optical fiber through a signal output port by fan-in, and the scattering signal generated by the optical fiber is received by the phase sensitive optical time domain reflectometer through fan-out;
the signal detection unit comprises a model establishing subunit, an application model subunit and an evaluation result subunit, wherein the model establishing subunit is used for training a classification prediction model according to the data monitored by the monitoring unit; the application model subunit is used for carrying out classification prediction according to the trained classification prediction model; and the evaluation result subunit is used for carrying out evaluation according to the classification prediction result.
Preferably, the physiological signal monitored by the monitoring body includes, but is not limited to, torque force when the human body moves, a movement range and human body temperature.
Preferably, the optical fiber is a multi-core optical fiber of a seven-core optical fiber, a four-core optical fiber or a three-core optical fiber.
Preferably, the signal detection unit implements monitoring of the physiological parameter, and specifically includes:
acquiring an input vector and an output vector of a neural network;
constructing a neural network model according to the input vector and the output vector;
obtaining a plurality of training samples of the neural network model; the training samples comprise input samples and output samples;
performing cyclic reciprocating training on the neural network model according to the plurality of training samples to generate a trained BP neural network model;
acquiring human body parameters to be predicted as current input vectors;
and inputting the current input vector into the trained neural network model, and predicting to obtain the human body activity index to be predicted, the human body motor skill index and the human body rehabilitation index.
Preferably, the model building subunit is configured to train a classification prediction model according to the data monitored by the monitoring unit, and the training of the classification prediction model according to the data monitored by the monitoring unit includes:
the method for constructing the neural network model according to the input vector and the output vector specifically comprises the following steps:
determining input vectors and output vectors of a neural network, wherein the input vectors include but are not limited to torsion, motion range and human body temperature when a human body moves, and the output vectors include but are not limited to human body activity indexes, human body motion skill indexes and human body rehabilitation indexes;
determining the number of input layer neurons of the neural network according to the dimension of the input vector;
determining the number of neurons in an output layer of the neural network according to the dimension of the output vector;
determining the number of hidden layer neurons of the BP neural network model according to the number of input layer neurons and the number of output layer neurons;
and constructing the neural network model according to the number of the input layer neurons, the number of the output layer neurons and the number of the hidden layer neurons.
Preferably, the model building subunit is configured to train a classification prediction model according to the data monitored by the monitoring unit, and train a classification prediction model according to the data monitored by the monitoring unit further includes:
training the constructed neural network model according to a plurality of training samples, firstly carrying out forward propagation and then carrying out reverse propagation to generate the trained neural network model; the method specifically comprises the following steps:
according to the kth input sample and formula
Figure BDA0002613585250000031
Determining a hidden layer input vector; wherein
Figure BDA0002613585250000032
A hidden layer input vector corresponding to a kth input sample, wherein h is 1, 2. n is the number of input layer neurons;
Figure BDA0002613585250000033
the connection weight from the input layer to the hidden layer; x is the number ofi(k) Is the kth input sample; bnA threshold value for each neuron of the hidden layer;
according to the hidden layer input vector
Figure BDA0002613585250000034
And formula
Figure BDA0002613585250000035
Determining a hidden layer output vector; wherein
Figure BDA0002613585250000036
A hidden layer output vector corresponding to the kth input sample;
outputting a vector according to the hidden layer
Figure BDA0002613585250000037
And formula
Figure BDA0002613585250000038
Determining an output layer input vector; wherein
Figure BDA0002613585250000039
An output layer input vector corresponding to the kth input sample; the connection weight from the hidden layer to the output layer; boIs the threshold value of each neuron of the output layer;
according to the output layer input vector
Figure BDA00026135852500000310
And formula
Figure BDA00026135852500000311
Determining an output layer output vector; wherein
Figure BDA00026135852500000312
Outputting a vector for an output layer corresponding to the kth input sample;
preferably, the connection weight from the hidden layer to the output layer is corrected according to the output layer input vector and the output layer output vector;
correcting the connection weight from the input layer to the hidden layer according to the connection weight of the hidden layer and the output layer;
determining a current neural network model according to the connection weight from the hidden layer to the output layer and the connection weight from the input layer to the hidden layer;
acquiring actual output of the current BP neural network model;
determining a global error according to the actual output;
judging whether the global error is smaller than a preset error value or not, and obtaining a first judgment result;
if the first judgment result is that the global error is smaller than a preset error value, determining that the current neural network model is the trained BP neural network model;
if the first judgment result is that the global error is not less than a preset error value, making k equal to k +1, and returning to the step of inputting the sample and the formula according to the kth
Figure BDA0002613585250000041
A step of determining a hidden layer input vector.
The correcting the connection weight from the hidden layer to the output layer according to the output layer input vector and the output layer output vector specifically includes:
correcting formula according to weight from hidden layer to output layer
Figure BDA0002613585250000042
Modifying the connection weights from hidden layer to output layer
Figure BDA0002613585250000043
Wherein
Figure BDA0002613585250000044
Is a connection weight
Figure BDA0002613585250000045
The nth correction value of (1);
Figure BDA0002613585250000046
is a connection weight
Figure BDA0002613585250000047
The (N + 1) th correction value; eta is the weight change rate;o(k) is the partial derivative of the error function to each neuron in the output layer.
Preferably, the modifying the connection weight from the input layer to the hidden layer according to the connection weight between the hidden layer and the output layer specifically includes:
correcting formula according to weight from input layer to hidden layer
Figure BDA0002613585250000048
Modifying connection weights of input layer to hidden layer
Figure BDA0002613585250000049
Wherein
Figure BDA00026135852500000410
Is a connection weight
Figure BDA00026135852500000411
The nth correction value of (1);
Figure BDA00026135852500000412
is a connection weight
Figure BDA00026135852500000413
The (N + 1) th correction value;h(k) is the partial derivative of the error function to each neuron in the hidden layer.
Preferably, the determining a global error according to the actual output specifically includes:
according to the actual output and the global error formula
Figure BDA00026135852500000414
Determining a global error; wherein E is a global error; m is the variable dimension of each training sample; do(k) An expected output vector corresponding to the kth input sample; y iso(k) The actual output vector corresponding to the kth input sample.
Compared with the prior art, the invention has the beneficial effects that:
according to the physiological parameter monitoring device of the wearable optical fiber sensor based on the neural network, the optical fiber grating is arranged in the clothing body, only clothing needs to be sleeved during detection, repeated wiring is not needed, the self-adaption degree is high, the time of preparation work for detection is short, and the detection efficiency is improved. In the neural network system based on the wearable optical fiber sensor, the detection system adopts a neural network algorithm to perform autonomous learning, and meanwhile, in the autonomous learning process, errors are continuously reduced through forward propagation and backward propagation, and the precision requirement is met. The device is used for detecting various torques, motion ranges and temperatures of a large number of limbs, waists or even brainbags of the human body during motion to obtain a large number of parameters and the like, and monitoring results can be intelligently fed back to guide rehabilitation training of the human body, skill training of athletes and the like through intelligent processing of a neural network system.
Drawings
Fig. 1 is a block diagram of a physiological parameter monitoring device of a wearable optical fiber sensor based on a neural network according to an embodiment of the present invention;
FIG. 2 is a schematic view of a monitoring body applied to a jacket according to an embodiment of the present invention;
FIG. 3 is a schematic structural view of a seven-core optical fiber according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of an optical path connection structure of a monitoring unit according to an embodiment of the present invention;
FIG. 5 is a block diagram of an algorithm flow of a signal detection unit according to an embodiment of the present invention;
FIG. 6 is a flowchart illustrating the steps of a training method for a BP neural network model according to an embodiment of the present invention;
FIG. 7 is a schematic diagram of a transmission process of a BP neural network model;
fig. 8 is a schematic diagram of a BP neural network model constructed according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In the description of the present invention, it is to be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", and the like, indicate orientations and positional relationships based on those shown in the drawings, and are used only for convenience of description and simplicity of description, and do not indicate or imply that the equipment or element being referred to must have a particular orientation, be constructed and operated in a particular orientation, and thus, should not be considered as limiting the present invention.
Fig. 1 is a block diagram of a physiological parameter monitoring device of a wearable optical fiber sensor based on a neural network, which includes a monitoring unit and a signal detection unit, wherein the monitoring unit monitors a physiological signal and inputs the monitored signal to the signal detection unit to realize monitoring of a physiological parameter;
the monitoring unit comprises a phase-sensitive optical time domain reflectometer system and a monitoring body; the monitoring body comprises an outer fabric layer, a waterproof layer is arranged at the bottom of the outer fabric layer, a carbon fiber layer is arranged at the bottom of the waterproof layer, and a plurality of optical fibers are embedded in the carbon fiber layer; the detection signal of the phase sensitive optical time domain reflectometer enters the optical fiber through a signal output port by fan-in, and the scattering signal generated by the optical fiber is received by the phase sensitive optical time domain reflectometer through fan-out;
the signal detection unit comprises a model establishing subunit, an application model subunit and an evaluation result subunit, wherein the model establishing subunit is used for training a classification prediction model according to the data monitored by the monitoring unit; the application model subunit is used for carrying out classification prediction according to the trained classification prediction model; and the evaluation result subunit is used for carrying out evaluation according to the classification prediction result.
Fig. 2 is a monitoring body, the monitoring body of the present embodiment is a jacket provided with optical fibers, and of course, the monitoring body can also be all wearable fabrics of human bodies such as trousers, shoes, hats, scarves, etc., the optical fibers used in the present embodiment are seven-core optical fibers, and the structure is as shown in fig. 3; the optical path connection structure of the monitoring unit is shown in fig. 4, a detection signal of the phase-sensitive optical time domain reflectometer system is fanned into a connection optical fiber, and a scattering signal generated by the optical fiber is received by the phase-sensitive optical time domain reflectometer through fanning out and is connected with a multi-core optical fiber (3); the phase-sensitive optical time domain reflectometer system can be arranged at the fixed positions on the walls or floors of a user training room, a rehabilitation room and a gymnasium, so that the position change of the optical fiber moving along with the human body after the human body is worn can be detected.
Specifically, the rayleigh scattering signal is intercepted by the phase-sensitive optical time domain reflectometer system in a segmented manner, the characteristic value is extracted at one time by using a data analysis system in the phase-sensitive optical time domain reflectometer system, the curvature radius of the shape change area of the seven-core optical fiber can be accurately calculated according to the strain value obtained by the phase-sensitive optical time domain reflectometer and a geometric method for describing a space curve, and particularly, the shape reconstruction characteristic value is obtained by solving a Frenet-Serret ordinary differential equation: and cutting the vector, the normal vector and the secondary normal vector to further obtain a torsion characteristic value of the seven-core optical fiber and obtain the movement torsion and the movement range of the human body during movement. The test system used comprised an 4/8 channel demodulator, a 7 channel distribution box, a 7-core fiber, and shape sensing test software, provided a method of tracking the shape along the length of the fiber and ensured high spatial resolution. The curvature of the optical fiber can be deduced, and the direction of the optical fiber can be deduced, so that the three-dimensional reconstruction of the shape of the optical fiber is realized. And the characteristics of small volume, high sensitivity, MRI compatibility and the like of the optical fiber are combined to perform shape sensing to realize position tracking.
To calculate the curvature radius, a local coordinate system is first defined to calculate θiThe values of (a) are shown in the example of fig. 3, which shows a 7-core fiber cross-sectional structure distribution, a 7-core fiber bending axis and a neutral axis thereof. In the example of fig. 3 in which core 3 is under tension and cores 2 and 7 are under compression (torsion of human body movement, etc.), the relationship between the radius of curvature due to the stress applied to the cross section and the strain value of each core can be given by expressions (1), (2):
R＝r2/2,4＝r3/3,4
R＝r2/2,4＝r7/7,4
wherein R is the bending radius, R2、r3、r7Respectively the distances of core 2, core 3, core 7 to the neutral axis,2,4、3,4、7,4the strain differences of core 2, core 3, core 7 and core 4, respectively. According to r in FIG. 12、r3、r7Trigonometric function relationship with r, and θ2、θ3、θ7The phase relation of 2 pi/3 of phase difference between the included angles can be calculated2、r3、r7The value of (c). Finally, r is added2、r3、r7The bending degree of the seven-core optical fiber can be determined by accurately calculating the curvature radius of the shape change area of the seven-core optical fiber instead of the equations (1) and (2). And (5) solving Frenet-Serret ordinary differential equation to obtain characteristic values. Assume initial conditions (T)0，N0，B0，R0) Wherein T is a unit tangent vector, N is a unit normal vector, B is a unit minor normal vector, R is a curvature radius, kappa is the curvature of the curve, and tau is the flexibility of the curve. According to the key parameters, the torsion characteristic value of the seven-core optical fiber is further obtained, and the motion torsion sum of the human body during motion is obtainedThe range of motion.
The Frenet-Serret ordinary differential equation is as above.
Figure BDA0002613585250000071
In addition, measuring the temperature of the human body using optical fibers is a well-known means in the art and will not be described herein.
As shown in fig. 5, the signal detection unit realizes the monitoring of the physiological parameter, and the specific steps include:
step 101: acquiring an input vector and an output vector of a neural network;
step 102: constructing a neural network model according to the input vector and the output vector;
the method for constructing the neural network model in the embodiment specifically comprises the following steps:
determining the number n of input layer neurons of the BP neural network model according to the dimension of the input vector; in this embodiment, the input vector is x (x1, x2, x3) (the torsion, the range of motion, and the body temperature when the body moves), and thus the number n of neurons in the input layer is determined to be 3;
determining the number of neurons in an output layer of the BP neural network model according to the dimension of the output vector; the output vector of each sample is human activity index/human motor skill index/human rehabilitation index, so the number q of neurons in the output layer is 2. Determining the number p of hidden layer neurons of the BP neural network model according to the number n of input layer neurons and the number q of output layer neurons; generally, the number of hidden layer nodes is greater than the number of nodes of the input layer and the output layer, and excessive hidden layer nodes cause overfitting, so that the number p of hidden layer neurons of the neural network model is determined to be 5 in the embodiment;
constructing the neural network model according to the number of the input layer neurons, the number of the output layer neurons and the number of the hidden layer neurons;
the BP neural network model has n neurons in the input layer, p neurons in the hidden layer and q neurons in the output layer, wherein the input vector is x ═ (x)1，x2，…，xn) The hidden layer input vector is
Figure BDA0002613585250000081
The hidden layer output vector is
Figure BDA0002613585250000082
The output layer input vector is
Figure BDA0002613585250000083
The output layer outputs a vector of
Figure BDA0002613585250000084
The desired output vector is do＝(d1，d2，…，dq) The connection weight of the input layer and the middle layer is wihThe connection weight of the hidden layer and the output layer is wh0The threshold of each neuron in the hidden layer is bhThe threshold value of each neuron of the output layer is b0The number k of sample data is 1, 2. The activation function is an sigmoid function:
Figure BDA0002613585250000085
the error function is:
Figure BDA0002613585250000086
the BP neural network model input vector of the aboveground biomass prediction constructed by the invention is x ═ (x)1,x2,x3)(x1,x2,x3Respectively representing the torsion, the motion range and the human body temperature when the human body moves); is input vector x ═ x1,x2,x3) Is weighted by wihThen entering an input vector of the hidden layer, wherein the dimension is changed from 3 to 5;
Figure BDA0002613585250000087
Figure BDA0002613585250000091
to pass through each neuron threshold bhThe vector of the post output, the dimension is 5;
Figure BDA0002613585250000092
is composed of
Figure BDA0002613585250000093
Figure BDA0002613585250000094
Weight w passing through hidden layer and output layerh0The vector of the output layer obtained after weighting, at this moment, the dimension is changed from 5 to 1; the actual output (human activity index/human motor skill index/human rehabilitation index) of a plurality of training samples is obtained.
The method for establishing the prediction model can also consider using a support vector machine, gradient lifting, Gaussian process regression and other methods, but the prediction of the BP neural network model is most accurate aiming at the prediction of the human physiological parameters, so the method of the embodiment selects to establish the BP neural network model.
Step 103: obtaining a plurality of training samples of the BP neural network model; the training samples include input samples and output samples.
Step 104: performing cyclic reciprocating training on the BP neural network model according to a plurality of training samples to generate a trained BP neural network model; as shown in fig. 6, the training method of the BP neural network model specifically includes the following steps:
step (1): and (5) initializing the network. And respectively assigning random numbers in an interval (-1,1) to each connection weight, setting an error function e, and giving calculation precision and maximum learning times M.
The calculation accuracy is preferably 10e-5The maximum number of learning times is preferably 100.
Step (2): randomly choosing the kth input sample and the corresponding expected output:
x(k)＝(x1(k),x2(k),…,xn(k))
do(k)＝(d1(k),d2(k),…,dq(k))
wherein x1(k)，x2(k)，...，xn(k) Respectively representing 3 input variables (torsion, motion range and human body temperature during human body motion); do(k) Representing the dependent variable (human activity index/human motor skill index/human rehabilitation index).
And (3): calculating the input and output of each neuron of the hidden layer:
Figure BDA0002613585250000095
Figure BDA0002613585250000096
Figure BDA0002613585250000101
Figure BDA0002613585250000102
wherein the hidden layer input vector
Figure BDA0002613585250000103
Is obtained by weighting the input layer and the output vector of the hidden layer
Figure BDA0002613585250000104
Is the hidden layer output value calculated by the activation function after weighting; output layer input vector
Figure BDA0002613585250000105
The hidden layer output value is weighted to obtain an output layer input value; output layer output vector
Figure BDA0002613585250000106
The output layer output value is obtained by substituting the output layer input value into the activation function.
Specifically, according to the kth input sample and formula
Figure BDA0002613585250000107
Determining a hidden layer input vector; wherein
Figure BDA0002613585250000108
A hidden layer input vector corresponding to a kth input sample, wherein h is 1, 2. n is the number of input layer neurons; w is aihThe connection weight from the input layer to the hidden layer; x is the number ofi(k) Is the kth input sample; bhA threshold value for each neuron of the hidden layer;
according to the hidden layer input vector
Figure BDA0002613585250000109
And formula
Figure BDA00026135852500001010
Determining a hidden layer output vector; wherein
Figure BDA00026135852500001011
A hidden layer output vector corresponding to the kth input sample;
outputting a vector according to the hidden layer
Figure BDA00026135852500001012
And formula
Figure BDA00026135852500001013
Determining an output layer input vector; wherein
Figure BDA00026135852500001014
An output layer input vector corresponding to the kth input sample; w is ah0The connection weight from the hidden layer to the output layer; b0Is the threshold value of each neuron of the output layer;
according to the output layer input vector
Figure BDA00026135852500001015
And formula
Figure BDA00026135852500001016
Determining an output layer output vector; wherein
Figure BDA00026135852500001017
Outputting a vector for an output layer corresponding to the kth input sample;
and 4, step 4: and correcting the connection weight from the hidden layer to the output layer according to the input vector of the output layer and the output vector of the output layer.
Calculating partial derivatives of error function to each neuron of output layer by using expected output and actual output of networko(k) The connection weight from the modified hidden layer to the output layer can satisfy the relation:
Figure BDA00026135852500001018
after one-time forward propagation (from an input layer, a hidden layer and an output layer), the value of an error function (a function formed by taking a value calculated by BP and an actually trained value as independent variables) at the moment is generally larger than the preset calculation precision, so that the reverse propagation (weight is corrected in the past to reduce the error) is required, and the weight and the threshold are repeatedly corrected all the time to enable the value of the error function to reach the calculation precision. So here the calculated partial derivatives are the correction weights, here the mathematical reasoning process, and the symbolic representation is the same as before.
Figure BDA0002613585250000111
Figure BDA0002613585250000112
Using neurons of the output layero(k) Correcting connection weight value by combining output of each neuron of hidden layer
Figure BDA0002613585250000113
Figure BDA0002613585250000114
And obtaining a weight value correction formula from the hidden layer to the output layer:
Figure BDA0002613585250000115
wherein
Figure BDA0002613585250000116
Is a connection weight
Figure BDA0002613585250000117
The nth correction value of (1);
Figure BDA0002613585250000118
is a connection weight
Figure BDA0002613585250000119
The (N + 1) th correction value; eta is weight change rate, 0 < eta < 1, and eta is affected by too large value
Figure BDA00026135852500001110
Too small will result in
Figure BDA00026135852500001111
The convergence rate of the solution is too slow;o(k) is the partial derivative of the error function to each neuron in the output layer.
And (5): and correcting the connection weight from the input layer to the hidden layer according to the connection weight of the hidden layer and the output layer.
Using the connection weights from the hidden layer to the output layer, the output layero(k) Calculating partial derivative of error function to each neuron of hidden layer according to output of hidden layerh(k) Can correct the input layer to be hiddenThe connection weight of the layers satisfies the formula:
Figure BDA00026135852500001112
Figure BDA00026135852500001113
Figure BDA00026135852500001114
Figure BDA0002613585250000121
using neurons of the hidden layerh(k) And correcting the connection weight value by the input parameter of the input layer neuron:
Figure BDA0002613585250000122
and obtaining a weight value correction formula from the input layer to the hidden layer:
Figure BDA0002613585250000123
wherein
Figure BDA0002613585250000124
Is a connection weight
Figure BDA0002613585250000125
The nth correction value of (1);
Figure BDA0002613585250000126
is a connection weight
Figure BDA0002613585250000127
The (N + 1) th correction value;h(k) is the partial derivative of the error function to each neuron in the hidden layer.
And (6): and determining the current BP neural network model according to the connection weight from the hidden layer to the output layer and the connection weight from the input layer to the hidden layer.
And (7): and acquiring the actual output of the current BP neural network model.
And (8): and determining a global error according to the actual output.
The global error calculation formula is as follows:
Figure BDA0002613585250000128
wherein E is a global error; m is the variable dimension of each training sample; do(k) An expected output vector corresponding to the kth input sample; y iso(k) The actual output vector corresponding to the kth input sample.
And (9): and judging whether the global error is smaller than a preset error value or not to obtain a first judgment result. If the first judgment result is that the global error is smaller than a preset error value or the learning frequency is larger than a set maximum frequency, finishing the algorithm and determining that the current BP neural network model is the trained BP neural network model; and (4) if the first judgment result is that the global error is not less than a preset error value, making k equal to k +1, selecting a next learning sample and a corresponding expected output, returning to the step (3), and performing a next round of learning.
Step 105: and acquiring the human body parameters to be predicted as the current input vector.
Step 106: and inputting the current input vector into the trained BP neural network model to obtain the human body activity index/human body motor skill index/human body rehabilitation index of the human body to be predicted.
When the relevant indexes such as the human body activity index, the human body motor skill index, the human body rehabilitation index and the like are obtained, the set relevant threshold values can be compared, and when the set relevant threshold values are exceeded, an alarm can be triggered to prompt a user to stop the ongoing activity or movement.
FIG. 7 is a schematic diagram of a transmission process of the BP neural network model; fig. 8 is a schematic diagram of a BP neural network model constructed according to an embodiment of the present invention.
According to the physiological parameter monitoring device of the wearable optical fiber sensor based on the neural network, the optical fiber grating is arranged in the clothing body, only clothing needs to be sleeved during detection, repeated wiring is not needed, the self-adaption degree is high, the time of preparation work for detection is short, and the detection efficiency is improved. In the neural network system based on the wearable optical fiber sensor, the detection system adopts a neural network algorithm to perform autonomous learning, and meanwhile, in the autonomous learning process, errors are continuously reduced through forward propagation and backward propagation, and the precision requirement is met. The device is used for detecting various torques, motion ranges and temperatures of a large number of limbs, waists or even brainbags of the human body during motion to obtain a large number of parameters and the like, and monitoring results can be intelligently fed back to guide rehabilitation training of the human body, skill training of athletes and the like through intelligent processing of a neural network system.
The foregoing shows and describes the general principles, essential features, and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, and the preferred embodiments of the present invention are described in the above embodiments and the description, and are not intended to limit the present invention. The scope of the invention is defined by the appended claims and equivalents thereof.

Claims (9)

1. A physiological parameter monitoring device of a wearable optical fiber sensor based on a neural network is characterized by comprising a monitoring unit and a signal detection unit, wherein the monitoring unit monitors physiological signals and inputs the monitored signals to the signal detection unit to realize the monitoring of physiological parameters;
the monitoring unit comprises a phase-sensitive optical time domain reflectometer system and a monitoring body; the monitoring body comprises an outer fabric layer, a waterproof layer is arranged at the bottom of the outer fabric layer, a carbon fiber layer is arranged at the bottom of the waterproof layer, and a plurality of optical fibers are embedded in the carbon fiber layer; the detection signal of the phase sensitive optical time domain reflectometer enters the optical fiber through a signal output port by fan-in, and the scattering signal generated by the optical fiber is received by the phase sensitive optical time domain reflectometer through fan-out;
the signal detection unit comprises a model establishing subunit, an application model subunit and an evaluation result subunit, wherein the model establishing subunit is used for training a classification prediction model according to the data monitored by the monitoring unit; the application model subunit is used for carrying out classification prediction according to the trained classification prediction model; and the evaluation result subunit is used for carrying out evaluation according to the classification prediction result.
2. The physiological parameter monitoring device of the wearable optical fiber sensor based on the neural network as claimed in claim 1, wherein the physiological signal monitored by the monitoring body includes but is not limited to torsion, range of motion and body temperature when the body moves.
3. The physiological parameter monitoring device of a wearable optical fiber sensor based on neural network as claimed in claim 1, wherein said optical fiber is multi-core optical fiber of seven-core optical fiber, four-core optical fiber or three-core optical fiber.
4. The physiological parameter monitoring device of the wearable optical fiber sensor based on the neural network as claimed in claim 1, wherein the signal detection unit implements monitoring of the physiological parameter, and specifically comprises:
acquiring an input vector and an output vector of a neural network;
constructing a neural network model according to the input vector and the output vector;
obtaining a plurality of training samples of the neural network model; the training samples comprise input samples and output samples;
performing cyclic reciprocating training on the neural network model according to the plurality of training samples to generate a trained BP neural network model;
acquiring human body parameters to be predicted as current input vectors;
and inputting the current input vector into the trained neural network model, and predicting to obtain the human body activity index to be predicted, the human body motor skill index and the human body rehabilitation index.
5. The physiological parameter monitoring device of a wearable optical fiber sensor based on neural network as claimed in claim 1, wherein the model building subunit is configured to train a classification prediction model according to the data monitored by the monitoring unit, and the training of the classification prediction model according to the data monitored by the monitoring unit comprises:
the method for constructing the neural network model according to the input vector and the output vector specifically comprises the following steps:
determining input vectors and output vectors of a neural network, wherein the input vectors include but are not limited to torsion, motion range and human body temperature when a human body moves, and the output vectors include but are not limited to human body activity indexes, human body motion skill indexes and human body rehabilitation indexes;
determining the number of input layer neurons of the neural network according to the dimension of the input vector;
determining the number of neurons in an output layer of the neural network according to the dimension of the output vector;
determining the number of hidden layer neurons of the BP neural network model according to the number of input layer neurons and the number of output layer neurons;
and constructing the neural network model according to the number of the input layer neurons, the number of the output layer neurons and the number of the hidden layer neurons.
6. The physiological parameter monitoring device of a wearable optical fiber sensor based on neural network as claimed in claim 5, wherein the model building subunit is configured to train a classification prediction model according to the data monitored by the monitoring unit, and the training of the classification prediction model according to the data monitored by the monitoring unit further comprises:
training the constructed neural network model according to a plurality of training samples, firstly carrying out forward propagation and then carrying out reverse propagation to generate the trained neural network model; the method specifically comprises the following steps:
according to the kth input sample and formula
Figure FDA0002613585240000021
Determining a hidden layer input vector; wherein
Figure FDA0002613585240000022
A hidden layer input vector corresponding to a kth input sample, wherein h is 1, 2. n is the number of input layer neurons;
Figure FDA0002613585240000023
the connection weight from the input layer to the hidden layer; x is the number ofi(k) Is the kth input sample; bnA threshold value for each neuron of the hidden layer;
according to the hidden layer input vector
Figure FDA0002613585240000024
And formula
Figure FDA0002613585240000025
Determining a hidden layer output vector; wherein
Figure FDA0002613585240000031
A hidden layer output vector corresponding to the kth input sample;
outputting a vector according to the hidden layer
Figure FDA0002613585240000032
And formula
Figure FDA0002613585240000033
Determining an output layer input vector; wherein
Figure FDA0002613585240000034
An output layer input vector corresponding to the kth input sample; the connection weight from the hidden layer to the output layer; boIs the threshold value of each neuron of the output layer;
according to the output layer input vector
Figure FDA0002613585240000035
And formula
Figure FDA0002613585240000036
Determining an output layer output vector; wherein
Figure FDA0002613585240000037
Outputting a vector for an output layer corresponding to the kth input sample;
correcting the connection weight from the hidden layer to the output layer according to the input vector of the output layer and the output vector of the output layer;
correcting the connection weight from the input layer to the hidden layer according to the connection weight of the hidden layer and the output layer;
determining a current neural network model according to the connection weight from the hidden layer to the output layer and the connection weight from the input layer to the hidden layer;
acquiring actual output of the current BP neural network model;
determining a global error according to the actual output;
judging whether the global error is smaller than a preset error value or not, and obtaining a first judgment result;
if the first judgment result is that the global error is smaller than a preset error value, determining that the current neural network model is the trained BP neural network model;
if the first judgment result is that the global error is not less than a preset error value, making k equal to k +1, and returning to the output according to the kthInput sample and formula
Figure FDA0002613585240000038
A step of determining a hidden layer input vector.
7. The physiological parameter monitoring device of a wearable optical fiber sensor based on a neural network as claimed in claim 6, wherein the modifying the connection weight from the hidden layer to the output layer according to the output layer input vector and the output layer output vector specifically comprises:
correcting formula according to weight from hidden layer to output layer
Figure FDA0002613585240000039
Modifying the connection weights from hidden layer to output layer
Figure FDA00026135852400000310
Wherein
Figure FDA00026135852400000311
Is a connection weight
Figure FDA00026135852400000312
The nth correction value of (1);
Figure FDA00026135852400000313
is a connection weight
Figure FDA00026135852400000314
The (N + 1) th correction value; eta is the weight change rate;o(k) is the partial derivative of the error function to each neuron in the output layer.
8. The physiological parameter monitoring device of a wearable optical fiber sensor based on a neural network as claimed in claim 6, wherein the modifying the connection weight from the input layer to the hidden layer according to the connection weight of the hidden layer and the output layer specifically comprises:
correcting formula according to weight from input layer to hidden layer
Figure FDA0002613585240000041
Modifying connection weights of input layer to hidden layer
Figure FDA0002613585240000042
Wherein
Figure FDA0002613585240000043
Is a connection weight
Figure FDA0002613585240000044
The nth correction value of (1);
Figure FDA0002613585240000045
is a connection weight
Figure FDA0002613585240000046
The (N + 1) th correction value;h(k) is the partial derivative of the error function to each neuron in the hidden layer.
9. The physiological parameter monitoring device of a wearable optical fiber sensor based on a neural network as claimed in claim 6, wherein the determining a global error according to the actual output specifically comprises:
according to the actual output and the global error formula
Figure FDA0002613585240000047
Determining a global error; wherein E is a global error; m is the variable dimension of each training sample; do(k) An expected output vector corresponding to the kth input sample; y iso(k) The actual output vector corresponding to the kth input sample.
CN202010762944.8A 2020-07-31 2020-07-31 Wearable optical fiber sensor is to human biological parameter monitoring devices based on neural network Pending CN111990977A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010762944.8A CN111990977A (en) 2020-07-31 2020-07-31 Wearable optical fiber sensor is to human biological parameter monitoring devices based on neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010762944.8A CN111990977A (en) 2020-07-31 2020-07-31 Wearable optical fiber sensor is to human biological parameter monitoring devices based on neural network

Publications (1)

Publication Number Publication Date
CN111990977A true CN111990977A (en) 2020-11-27

Family

ID=73463285

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010762944.8A Pending CN111990977A (en) 2020-07-31 2020-07-31 Wearable optical fiber sensor is to human biological parameter monitoring devices based on neural network

Country Status (1)

Country Link
CN (1) CN111990977A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116540361A (en) * 2023-07-05 2023-08-04 中国电信股份有限公司 Multi-core transmission system and multi-core transmission method

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6498652B1 (en) * 2000-02-08 2002-12-24 Deepak Varshneya Fiber optic monitor using interferometry for detecting vital signs of a patient
CN109001136A (en) * 2018-09-20 2018-12-14 杭州绿洁水务科技股份有限公司 A kind of COD on-line monitoring method based on ultraviolet visible light absorption spectrum
CN109214591A (en) * 2018-10-12 2019-01-15 北京林业大学 A kind of xylophyta ground biomass prediction technique and system
CN110569593A (en) * 2019-09-05 2019-12-13 武汉纺织大学 Method and system for measuring three-dimensional size of dressed human body, storage medium and electronic equipment

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6498652B1 (en) * 2000-02-08 2002-12-24 Deepak Varshneya Fiber optic monitor using interferometry for detecting vital signs of a patient
CN109001136A (en) * 2018-09-20 2018-12-14 杭州绿洁水务科技股份有限公司 A kind of COD on-line monitoring method based on ultraviolet visible light absorption spectrum
CN109214591A (en) * 2018-10-12 2019-01-15 北京林业大学 A kind of xylophyta ground biomass prediction technique and system
CN110569593A (en) * 2019-09-05 2019-12-13 武汉纺织大学 Method and system for measuring three-dimensional size of dressed human body, storage medium and electronic equipment

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116540361A (en) * 2023-07-05 2023-08-04 中国电信股份有限公司 Multi-core transmission system and multi-core transmission method
CN116540361B (en) * 2023-07-05 2023-09-12 中国电信股份有限公司 Multi-core transmission system and multi-core transmission method

Similar Documents

Publication Publication Date Title
Wu et al. A novel DAS signal recognition method based on spatiotemporal information extraction with 1DCNNs-BiLSTM network
RU2013134980A (en) DEVICE AND METHOD FOR GOLF SWING ANALYSIS
CN109940622A (en) It is a kind of based on the robot arm of current of electric without sensing collision checking method
CN110401978B (en) Indoor positioning method based on neural network and particle filter multi-source fusion
CN111461187B (en) Intelligent building settlement detection system
CN110236550B (en) Human gait prediction device based on multi-mode deep learning
US20190008419A1 (en) Rehabilitation evaluation apparatus, rehabilitation evaluation method, and rehabilitation evaluation program
CN113296087B (en) Frequency modulation continuous wave radar human body action recognition method based on data enhancement
CN114896672B (en) Bridge dynamic displacement reconstruction method based on CSO-BilSTM network
CN108846200B (en) Quasi-static bridge influence line identification method based on iteration method
CN108804851A (en) A kind of high-precision propylene polymerization production process optimal soft survey instrument of chaos gunz optimizing
CN107967489A (en) A kind of method for detecting abnormality and system
CN110044525A (en) A kind of flexible resistive dot matrix pressure detecting system, method and apparatus
CN103637800B (en) Based on the method for the bioelectrical impedance analysis of eight sections of impedance models
CN110308432A (en) A kind of radar self-adaption waveform selection Activity recognition method neural network based
CN111990977A (en) Wearable optical fiber sensor is to human biological parameter monitoring devices based on neural network
Lyu et al. Three-fingers FBG tactile sensing system based on squeeze-and-excitation LSTM for object classification
Wang et al. Machine learning-based human motion recognition via wearable plastic fiber sensing system
CN110160492A (en) A kind of the inclination monitoring method and device of power transmission tower
CN109187898B (en) Soft measurement method and device for ammonia nitrogen content of water in aquaculture environment
CN113820062B (en) Temperature compensation method of six-dimensional force sensor
CN107437112B (en) A kind of mixing RVM model prediction methods based on the multiple dimensioned kernel function of improvement
CN113609750B (en) Current calculation method based on magnetic field sensing signal machine learning
Yen et al. Feature Fusion-Based Deep Learning Network to Recognize Table Tennis Actions.
CN107590975A (en) The implementation method of warning system based on optical fiber, smart coat and piezoelectric transducer

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