CN114970616A - Training method and device for vital sign signal extraction model - Google Patents
Training method and device for vital sign signal extraction model Download PDFInfo
- Publication number
- CN114970616A CN114970616A CN202210529418.6A CN202210529418A CN114970616A CN 114970616 A CN114970616 A CN 114970616A CN 202210529418 A CN202210529418 A CN 202210529418A CN 114970616 A CN114970616 A CN 114970616A
- Authority
- CN
- China
- Prior art keywords
- data
- vital sign
- training
- sample
- signal extraction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012549 training Methods 0.000 title claims abstract description 121
- 238000000605 extraction Methods 0.000 title claims abstract description 90
- 238000000034 method Methods 0.000 title claims abstract description 45
- 230000007613 environmental effect Effects 0.000 claims abstract description 27
- 238000012545 processing Methods 0.000 claims abstract description 18
- 239000011159 matrix material Substances 0.000 claims description 43
- 239000013598 vector Substances 0.000 claims description 30
- 238000000354 decomposition reaction Methods 0.000 claims description 21
- 230000008569 process Effects 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 7
- 230000004044 response Effects 0.000 claims description 2
- 238000001514 detection method Methods 0.000 abstract description 11
- 238000013480 data collection Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 15
- 238000010586 diagram Methods 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 5
- 230000029058 respiratory gaseous exchange Effects 0.000 description 5
- 238000012706 support-vector machine Methods 0.000 description 5
- 102100037651 AP-2 complex subunit sigma Human genes 0.000 description 3
- 101000806914 Homo sapiens AP-2 complex subunit sigma Proteins 0.000 description 3
- 230000036760 body temperature Effects 0.000 description 3
- 238000003062 neural network model Methods 0.000 description 3
- 241001334146 Rugopharynx delta Species 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000000149 penetrating effect Effects 0.000 description 1
- 230000000241 respiratory effect Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000007480 spreading Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
- 239000002023 wood Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
- A61B5/0205—Simultaneously evaluating both cardiovascular conditions and different types of body conditions, e.g. heart and respiratory condition
-
- 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
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Data Mining & Analysis (AREA)
- Heart & Thoracic Surgery (AREA)
- Veterinary Medicine (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Cardiology (AREA)
- Signal Processing (AREA)
- Animal Behavior & Ethology (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- General Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Pulmonology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Psychiatry (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The present disclosure provides a training method, an apparatus, an electronic device and a storage medium for a vital sign signal extraction model, which relate to the technical field of vital signal detection, and the method includes: acquiring a vital sign signal extraction model to be trained; acquiring a training sample set, wherein the training sample set comprises a plurality of groups of vital sign data and environmental data; acquiring sample training data based on the vital sign data and the environmental data, and simultaneously determining a sample label value of the vital sign data; and training the vital sign signal extraction model based on the sample label value and the sample training data until the training is finished to obtain the target vital sign signal extraction model. Target vital sign signal extraction model is generated through training to data collection is handled based on the model, and whether there is vital sign in order to judge the collection object, from this, can promote rescue data processing's time greatly, promote data processing efficiency, reduce the processing cost, can strive for more time for the rescue simultaneously.
Description
Technical Field
The present disclosure relates to the field of vital signal detection technologies, and in particular, to a training method and apparatus for a vital sign signal extraction model, an electronic device, and a storage medium.
Background
The non-contact life detection technology is a new technology which is newly developed along with the development needs of medical engineering, military and society in recent years, breaks through the existing detection method and technology, and searches for a life body by utilizing life parameters of the human body, such as respiration, heartbeat and human body movement signals, under the condition of not contacting a target body.
In the prior art, the signal space of heartbeat and respiration in the weak living body echo signal is orthogonal to the noise space, so that a multiple signal classification (MUSIC) algorithm can be adopted for spectrum estimation. However, the above method needs a lot of time for processing data, which greatly increases the rescue difficulty and even delays the optimal rescue time.
BRIEF SUMMARY OF THE PRESENT DISCLOSURE
The present disclosure is directed to solving, at least to some extent, one of the technical problems in the related art.
To this end, it is an object of the present disclosure to propose a training method of a vital sign signal extraction model.
A second objective of the present disclosure is to provide a training apparatus for a vital sign signal extraction model.
A third object of the present disclosure is to provide an electronic device.
A fourth object of the present disclosure is to propose a non-transitory computer-readable storage medium.
A fifth object of the present disclosure is to propose a computer program product.
To achieve the above object, an embodiment of a first aspect of the present disclosure provides a training method of a vital sign signal extraction model, including: acquiring a vital sign signal extraction model to be trained; acquiring a training sample set, wherein the training sample set comprises a plurality of groups of vital sign data and environmental data; acquiring sample training data based on the vital sign data and the environmental data, and simultaneously determining a sample label value of the vital sign data; and training the vital sign signal extraction model based on the sample label value and the sample training data until the training is finished to obtain the target vital sign signal extraction model.
According to one embodiment of the present disclosure, training a vital sign signal extraction model based on sample signatures and sample training data includes: inputting sample training data into a vital sign signal extraction model to output a predicted label value; determining a loss value of the vital sign signal extraction model based on the prediction tag value and the sample tag value; and adjusting the model parameters of the vital sign signal extraction model according to the loss values, and returning to train the adjusted vital sign signal extraction model by adopting the next sample training data.
According to an embodiment of the present disclosure, the vital sign data and the environmental data are data matrices formed by a plurality of element data, the element data of the vital sign data and the element data of the environmental data are in one-to-one correspondence, and sample training data is obtained based on the vital sign data and the environmental data, including: and (3) aiming at any data matrix in the vital sign data and the environmental data, performing data processing on any data matrix to extract singular entropy, wavelet packet scale entropy and sample entropy as sample training data.
According to one embodiment of the disclosure, the singular entropy extraction process includes: determining a singular value matrix based on the data matrix; determining singular entropy of the data matrix based on the singular value matrix; the following formula is adopted to determine the singular entropy:
wherein σ i Representing the ith singular value, σ, in a matrix of singular values 1 +σ 2 +…σ k Representing the summation of all singular values in a matrix of singular values.
According to one embodiment of the present disclosure, extracting wavelet packet scale entropy comprises: decomposing a first-order right singular vector in the data matrix to obtain a wavelet packet decomposition coefficient vector; determining a wavelet packet scale entropy based on the wavelet packet decomposition coefficient vector; wherein, the wavelet packet scale entropy is determined by adopting the following formula:
wherein, w i And j is a wavelet packet node, and k is the number of wavelet packet decomposition layers.
According to one embodiment of the present disclosure, extracting sample entropy of a data matrix comprises: expanding the data matrix into a one-dimensional vector sequence; determining a first number of samples and a second number of samples based on the sequence of vectors; determining a sample entropy based on the first number of samples and the second number of samples; wherein, the sample entropy is determined by the following formula:
wherein A is m (r) first number of samples of dimension m of entropy reconstruction of samples, B m (r) represents a second number of samples of dimension m +1 of the entropy reconstruction of the samples.
According to one embodiment of the present disclosure, extracting sample entropy of a data matrix comprises: judging whether the element data in the vital sign data meet the vital sign conditions or not; responding to the condition that the element data in the vital sign data meet the vital sign condition, and determining that the label value of the element data is 1; responding to the situation that the bit data in the vital sign data does not meet the preset condition, and determining that the tag value of the element data is 0; a sample label value for the vital sign data is generated based on the label values of all the element data.
To achieve the above object, an embodiment of a second aspect of the present disclosure provides a training apparatus for a vital sign signal extraction model, including: the first acquisition module is used for acquiring a vital sign signal extraction model to be trained; the second acquisition module is used for acquiring a training sample set, and the training sample set comprises a plurality of groups of vital sign data and environmental data; the determining module is used for acquiring sample training data based on the vital sign data and the environment data and determining a sample label value of the vital sign data; and the training module is used for training the vital sign signal extraction model based on the sample label value and the sample training data until the training is finished to obtain the target vital sign signal extraction model.
To achieve the above object, an embodiment of a third aspect of the present disclosure provides an electronic device, including: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executable by the at least one processor to implement a method for training a vital sign signal extraction model as defined in an embodiment of the first aspect of the disclosure.
To achieve the above object, a fourth aspect of the present disclosure provides a non-transitory computer-readable storage medium storing computer instructions, where the computer instructions are used to implement a training method for a vital sign signal extraction model according to the first aspect of the present disclosure.
To achieve the above object, a fifth aspect of the present disclosure provides a computer program product, which includes a computer program, when being executed by a processor, is configured to implement the training method for a vital sign signal extraction model according to the first aspect of the present disclosure.
Target vital sign signal extraction model is generated through training to data collection is handled based on the model, and whether there is vital sign in order to judge the collection object, from this, can promote rescue data processing's time greatly, promote data processing efficiency, reduce the processing cost, can strive for more time for the rescue simultaneously.
Drawings
Fig. 1 is a schematic diagram of a training method of a vital sign signal extraction model according to an embodiment of the present disclosure;
fig. 2 is a schematic diagram of another training method of a vital sign signal extraction model according to an embodiment of the present disclosure;
fig. 3 is a schematic diagram of another training method of a vital sign signal extraction model according to an embodiment of the present disclosure;
fig. 4 is a schematic diagram of a training apparatus for a vital sign signal extraction model according to an embodiment of the present disclosure;
fig. 5 is a schematic diagram of an electronic device according to an embodiment of the present disclosure.
Detailed Description
Reference will now be made in detail to the embodiments of the present disclosure, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary and intended to be illustrative of the present disclosure, and should not be construed as limiting the present disclosure.
Weak signal detection is always a hotspot and difficult problem in the field of signal processing, and a large amount of clutter, noise and the like can cause strong interference and even cover useful weak vital signals. Meanwhile, in the process of detecting the life signal by using the ultra-wideband radar life detector, a large amount of radar echo data needs to be processed, so that the detection work is quickly carried out while the life signal is accurately extracted, the timeliness of the algorithm is ensured, enough time is strived for emergency rescue work, and the important problem which needs to be overcome in the field is solved.
The invention provides a training method of a vital sign signal extraction model, which solves the problems of long time and low accuracy in analyzing and processing data, can accurately position a vital body and extract heartbeat and respiratory signals, and has high accuracy, timeliness, robustness and signal-to-noise ratio.
Fig. 1 is a schematic diagram of an exemplary embodiment of a training method of a vital sign signal extraction model proposed in the present disclosure, as shown in fig. 1, the training method of the vital sign signal extraction model includes the following steps:
and S101, acquiring a vital sign signal extraction model to be trained.
In the embodiment of the present disclosure, the vital sign signal extraction model to be trained is set in advance, and the vital sign signal extraction model to be trained may be various, for example, a neural network model, a Support Vector Machine (SVM), or the like, and is not limited herein.
The SVM model is the best existing classifier, and the existing classifier can be directly used without modification. And a lower error rate can be obtained, and the SVM can make a good classification decision on data points outside the training set.
The neural network model has strong robustness and fault tolerance, self-learning, self-organization and self-adaptability, so that the network can process uncertain or unknown systems, can fully approximate any complex nonlinear relation, has strong information synthesis capability, can process quantitative and qualitative information simultaneously, can well coordinate various input information relations, and is suitable for multi-information fusion and multimedia technology.
Different models and different emphasis directions can be adopted, so that when the vital sign signal extraction model is selected, the corresponding reference model can be selected according to the emphasis of actual needs, and the optimal data processing effect is achieved.
S102, a training sample set is obtained, and the training sample set comprises a plurality of groups of vital sign data and environmental data.
In the embodiment of the present disclosure, the vital sign data may include various data, for example, body temperature data, heartbeat data, respiration data, etc., which are not limited herein and may be set according to actual needs.
The environmental data may include various data, for example, position data, humidity data, waveform data, etc., which are not limited herein and may be set according to actual needs.
The vital sign data and the environmental data may characterize attribute information of the test site. In the embodiment of the present disclosure, the vital sign data and the environmental data are data matrices formed by a plurality of element data, and the element data of the vital sign data corresponds to the element data of the environmental data one to one. For example, as shown in the following two data matrices:
the extracted vital sign data is in the form of:
the extracted environmental data is in the form of:
in the embodiment of the present disclosure, the training sample set may be real-time data actually collected by a life detection device, which may be an ultra-wideband radar life detection device, an infrared life detection device, or an audio life detection device, and the like, without any limitation.
The data acquisition is carried out by the life detection instrument equipment under different acquisition scenes, for example, the acquisition scenes can be static state, life bodies penetrating barriers such as wood boards, concrete, water and the like, a plurality of mutually overlapped life bodies, life bodies with the distance of more than 10m, non-static state life bodies and the like. It can be understood that the data collected under different scenes can be different, when a large amount of clutter and noise exists, the collected data pass through walls and obstacles, a plurality of vital bodies are detected simultaneously, and the like, the collected vital sign signals are further covered and difficult to identify, and when the collected data are in a wider environment or an environment with less clutter and noise, the collected data are clearer and the vital sign signals in the current data are easier to distinguish.
Optionally, the training sample set may also be preset data, for example, the training sample set may be obtained by connecting a training sample library, where the training sample library is a preset database, and the training sample library may be stored in the cloud or in a storage space of the electronic device
S103, acquiring sample training data based on the vital sign data and the environment data, and determining a sample label value of the vital sign data.
In the embodiment of the present disclosure, after the vital sign data and the environmental data are obtained, the vital sign data and the environmental data need to be processed to obtain the sample training data. It should be noted that the training data may be various, and in the embodiment of the present disclosure, the training data may be singular entropy, wavelet packet scale entropy of first-order right singular vectors, sample entropy, and the like, which is not limited herein.
In the embodiment of the present disclosure, when the collecting subject has a vital sign, the sample label value may be a sign signal, and when the collecting subject does not have a vital sign, the sample label value may be an uncomfortableness signal. The specific data may be determined according to actual settings, and is not limited herein.
The method for determining the vital sign can be various, for example, when the body temperature data in the vital sign data is greater than a preset body temperature value, the vital sign data can be considered as the vital sign; when the heartbeat data in the vital sign data is a heartbeat value, the vital sign data can be considered as a vital sign; when the respiration data in the vital sign data is respiration, the vital sign data may be considered as a vital sign, and the like, and the vital sign data is not limited herein and may be specifically set according to an actual situation.
Optionally, the sample tag value of the vital sign data may also be obtained at the same time as the vital sign data is obtained.
And S104, training the vital sign signal extraction model based on the sample label value and the sample training data until the training is finished to obtain the target vital sign signal extraction model.
The training of the vital sign signal extraction model is a repeated iteration process, and the training is carried out by continuously adjusting the network parameters of the model until the overall loss function value of the model is smaller than a preset value, or the overall loss function value of the model is not changed or the change amplitude is slow, the model converges, and the trained target vital sign signal extraction model is obtained.
In the embodiment of the disclosure, a vital sign signal extraction model to be trained is firstly obtained, then a training sample set is obtained, the training sample set comprises a plurality of groups of vital sign data and environmental data, then sample training data is obtained based on the vital sign data and the environmental data, a sample label value of the vital sign data is determined, and finally the vital sign signal extraction model is trained based on the sample label value and the sample training data until the training is finished to obtain a target vital sign signal extraction model. Generate target vital sign signal through the training and draw the model, in actual operation, can handle data collection through the model to whether the judgement is the collection object and has vital sign, from this, can promote the time of rescue data processing greatly, promote data processing efficiency, reduce the processing cost, can strive for more time for the rescue simultaneously.
In the embodiment of the disclosure, a sample signature value of the vital sign data is determined, and whether element data in the vital sign data meets a vital sign condition is judged; responding to the condition that the element data in the vital sign data meet the vital sign condition, and determining that the label value of the element data is 1; responding to the situation that the bit data in the vital sign data does not meet the preset condition, and determining that the tag value of the element data is 0; a sample label value for the vital sign data is generated based on the label values of all the element data. By way of example, when the vital sign data is in the form of,
the tag value may be:
in the above embodiment, the training of the vital sign signal extraction model is performed based on the sample signature and the sample training data, which can be further explained by fig. 2, and the method includes:
s201, inputting the sample training data into the vital sign signal extraction model to output a prediction label value.
S202, determining a loss value of the vital sign signal extraction model based on the prediction label value and the sample label value.
The loss function in this embodiment is set in advance, and can be set according to actual needs. For example, the loss function may be a hinge loss function, a cross entropy loss function, an exponential loss function, and the like, and may be specifically selected according to actual needs, which is not limited herein.
It should be noted that the loss functions corresponding to different types of vital sign signal extraction models may be different, for example, the loss function corresponding to the neural network model may be a cross entropy loss function or an exponential loss function, and the loss function corresponding to the SVM model may be a hinge loss function, etc.
After the predicted label value and the sample label value are obtained, the predicted label value and the sample label value can be input into a loss function to determine a loss value of the vital sign signal extraction model.
And S203, adjusting the model parameters of the vital sign signal extraction model according to the loss values, and returning to train the adjusted vital sign signal extraction model by adopting the next sample training data.
In the embodiment of the disclosure, after the loss value is obtained, the calculated loss value is compared with a preset value, if the loss value is greater than the preset value, the network parameter of the to-be-trained network is updated, the model is extracted based on the to-be-trained vital sign signal after the network parameter is updated, the loss value of the to-be-trained model is calculated again according to the training sample set, and iteration is performed until the loss value is smaller than the preset value, so that the trained target vital sign signal extraction model is obtained.
In the embodiment of the disclosure, sample training data is first input into the vital sign signal extraction model to output a prediction tag value, then a loss value of the vital sign signal extraction model is determined based on the prediction tag value and the sample tag value, finally, a model parameter of the vital sign signal extraction model is adjusted according to the loss value, and the adjusted vital sign signal extraction model is trained by returning to the next sample training data. Therefore, the model is continuously adjusted through the loss value, a more accurate target vital sign signal extraction model can be obtained, and the accuracy of data processing is improved.
In the above embodiment, the obtaining of the sample training data based on the vital sign data and the environmental data can be further explained by using fig. 3, where the method includes:
s301, aiming at any data matrix in the vital sign data and the environmental data, data processing is carried out on any data matrix, so that singular entropy, wavelet packet scale entropy and sample entropy are extracted and serve as sample training data.
In an embodiment of the present disclosure, the singular entropy extracting process includes: and determining a singular value matrix based on the data matrix, and determining the singular entropy of the data matrix based on the singular value matrix.
The singular value decomposition process of the vital sign data S is as follows:
the singular value matrix sigma kxk has singular values only on the main diagonal, all other elements are 0, main diagonal elements sigma 1, sigma 2, sigma 3, … and sigma k are k singular values, and the condition that sigma 1 is larger than or equal to sigma 2, sigma 3 is larger than or equal to … and larger than or equal to sigma k is met. ui denotes the ith column vector of the matrix UMxk, called the ith left singular vector. vi denotes the ith column vector of the matrix VNxk, referred to as the ith-order right singular vector. σ i denotes the i-th element of the singular value spectrum. For the present invention, the top k (min (M, N) ≧ k ≧ 3) order singular values σ 1, σ 2, σ 3, …, σ k, left singular vectors u1, u2, u3, …, uk, and right singular vectors v1, v2, v3, …, vk may be retained as decomposition results.
The vital sign data S is calculated by the following singular entropy formula:
in an embodiment of the present disclosure, extracting wavelet packet scale entropy includes:
calculating wavelet packet scale entropy of first-order right singular vectors of all data in the vital sign data sample and the environmental data sample;
the wavelet packet decomposition process of the first-order right singular vector v1 of the vital sign data S is as follows:
selecting 'db 6' wavelets, setting the number of wavelet packet decomposition layers to be 2, and performing wavelet packet decomposition, wherein the wavelet packet decomposition method comprises the following steps:
wherein, W00 is an original signal v1, W is a wavelet packet decomposition coefficient vector of each layer, n and j are wavelet packet nodes, k is the number of wavelet packet decomposition layers, h is a high-pass filter coefficient, and g is a low-pass filter coefficient.
The wavelet packet scale entropy calculation formula of each node is as follows:
where k is the number of wavelet packet decomposition layers, j is a wavelet packet node (j is 0,1, …,2k-1), (in the present invention, k is 2.)
Is a wavelet packet decomposition coefficient vector, wi (i ═ 1,2, …, x) is Wkj for each wavelet coefficient.
In an embodiment of the present disclosure, extracting sample entropy of a data matrix includes:
calculating sample entropy for all data in the vital sign data samples and the environmental data samples;
the process of calculating the sample entropy of the vital sign data S is as follows:
spreading the vital sign data S into a one-dimensional sequence { X1, X2, …, xn }, sampling the entropy reconstruction dimension m of 2, the threshold value r, calculating the standard deviation of the sequence as delta, and forming a group of vector sequences { X, with dimension m, of the one-dimensional sequences according to the sequence numbers m (1),X m (2),…X m (n-m+1)}。
Wherein X m (i)＝[x(i),x(i+1),…x(i+m-1)]1. ltoreq. i.ltoreq.n-m +1 represents the values of m consecutive x from the point i.
Definition vector X m (i) And X m (j) The distance between the two elements is the absolute value of the maximum difference between the two corresponding elements, namely:
d[X m (i),X m (j)]＝max k＝0,…,m-1 (|x(i+k)-x(j+k)|)
for a given X m (i) Statistics of X m (i) And X m (j) The number of j (1. ltoreq. j. ltoreq.n-m, j. noteq. i) whose distance between them is smaller than r.delta is denoted as Bi. For 1. ltoreq. i.ltoreq.n-m, define:
add dimension to m +1, calculate X m+1 (i) And X m+1 (j) The number of j (1. ltoreq. j. ltoreq.n-m-1, j. noteq. i) whose distance between them is smaller than r.delta, and is denoted as Ai. Then define:
thus, B m (r) is the probability that two sequences match m points with a similarity tolerance of r, and A m (r) is the probability that two sequences match m +1 points with a similarity tolerance of r. The sample entropy calculation formula is:
corresponding to the training methods of the vital sign signal extraction models provided in the above-mentioned several embodiments, an embodiment of the present disclosure further provides a training device of the vital sign signal extraction model, and since the training device of the vital sign signal extraction model provided in the embodiment of the present disclosure corresponds to the training methods of the vital sign signal extraction models provided in the above-mentioned several embodiments, the implementation manner of the training method of the vital sign signal extraction model is also applicable to the training device of the vital sign signal extraction model provided in the embodiment of the present disclosure, and will not be described in detail in the following embodiments.
Fig. 4 is a schematic diagram of a training apparatus for a vital sign signal extraction model proposed in the present disclosure, and as shown in fig. 4, the training apparatus 400 for a vital sign signal extraction model includes: a first acquisition module 410, a second acquisition module 420, a determination module 430, and a training module 440.
The first obtaining module 410 is configured to obtain a vital sign signal extraction model to be trained.
A second obtaining module 420, configured to obtain a training sample set, where the training sample set includes multiple sets of vital sign data and environmental data.
A determining module 430, configured to obtain sample training data based on the vital sign data and the environment data, and determine a sample tag value of the vital sign data.
The training module 440 is configured to train the vital sign signal extraction model based on the sample label value and the sample training data until the training is finished to obtain a target vital sign signal extraction model.
In an embodiment of the disclosure, the training module 440 is further configured to: inputting the sample training data into the vital sign signal extraction model to output a predicted label value; determining a loss value of the vital sign signal extraction model based on the predicted label value and the sample label value; and adjusting the model parameters of the vital sign signal extraction model according to the loss values, and returning to train the adjusted vital sign signal extraction model by adopting the next sample training data.
In an embodiment of the disclosure, the determining module 430 is further configured to: and aiming at any data matrix in the vital sign data and the environment data, performing data processing on the data matrix to extract singular entropy, wavelet packet scale entropy and sample entropy as the sample training data.
In an embodiment of the disclosure, the determining module 430 is further configured to: determining a singular value matrix based on the data matrix; determining singular entropy of the data matrix based on the singular value matrix; the singular entropy is determined by adopting the following formula:
wherein σ i Representing the ith singular value, σ, in a matrix of singular values 1 +σ 2 +…σ k Representing the summation of all singular values in a matrix of singular values.
In an embodiment of the disclosure, the determining module 430 is further configured to: decomposing a first-order right singular vector in the data matrix to obtain a wavelet packet decomposition coefficient vector; determining a wavelet packet scale entropy based on the wavelet packet decomposition coefficient vector; wherein, the wavelet packet scale entropy is determined by adopting the following formula:
wherein w i Wavelet coefficients representing a vector of wavelet packet decomposition coefficients, j beingAnd k is the number of wavelet packet decomposition layers.
In an embodiment of the disclosure, the determining module 430 is further configured to: expanding the data matrix into a one-dimensional vector sequence; determining a first number of samples and a second number of samples based on the sequence of vectors; determining a sample entropy based on the first number of samples and the second number of samples; wherein, the sample entropy is determined by the following formula:
wherein A is m (r) first number of samples with dimension m of sample entropy reconstruction, B m (r) represents the second number of samples with the sample entropy reconstruction dimension m + 1.
In an embodiment of the disclosure, the determining module 430 is further configured to: judging whether the element data in the vital sign data meet the vital sign conditions; in response to element data in the vital sign data meeting the vital sign condition, determining that a tag value of the element data is 1; responding to the vital sign data that the bit data does not meet the preset condition, and determining that the tag value of the element data is 0; and generating a sample label value of the vital sign data based on the label values of all the element data.
In order to implement the foregoing embodiment, an embodiment of the present disclosure further provides an electronic device 500, as shown in fig. 5, where the electronic device 500 includes: the processor 501 and the memory 502 communicatively connected to the processor, the memory 502 storing instructions executable by the at least one processor, the instructions being executable by the at least one processor 501 to implement the training method of the vital sign signal extraction model as embodied in the first aspect of the present disclosure.
In order to achieve the above embodiments, the embodiments of the present disclosure further propose a non-transitory computer-readable storage medium storing computer instructions for causing a computer to implement the training method of the vital sign signal extraction model as embodied in the first aspect of the present disclosure.
In order to implement the above embodiments, the present disclosure also proposes a computer program product, which includes a computer program, and when executed by a processor, the computer program implements the training method of the vital sign signal extraction model as in the first aspect of the present disclosure.
In the description of the present disclosure, 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," "axial," "radial," "circumferential," and the like are used in the orientations and positional relationships indicated in the drawings for convenience in describing the present disclosure and to simplify the description, but are not intended to indicate or imply that the referenced devices or elements must have a particular orientation, be constructed and operated in a particular orientation, and are therefore not to be considered limiting of the present disclosure.
Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present disclosure, "a plurality" means two or more unless specifically limited otherwise.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Although embodiments of the present disclosure have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present disclosure, and that changes, modifications, substitutions and alterations may be made to the above embodiments by those of ordinary skill in the art within the scope of the present disclosure.
Claims (10)
1. A training method of a vital sign signal extraction model is characterized by comprising the following steps:
acquiring a vital sign signal extraction model to be trained;
acquiring a training sample set, wherein the training sample set comprises a plurality of groups of vital sign data and environmental data;
obtaining sample training data based on the vital sign data and the environmental data, and simultaneously determining a sample label value of the vital sign data;
and training the vital sign signal extraction model based on the sample label value and the sample training data until the training is finished to obtain a target vital sign signal extraction model.
2. The method of claim 1, wherein training the vital sign signal extraction model based on the sample signatures and the sample training data comprises:
inputting the sample training data into the vital sign signal extraction model to output a predicted label value;
determining a loss value of the vital sign signal extraction model based on the predicted label value and the sample label value;
and adjusting the model parameters of the vital sign signal extraction model according to the loss values, and returning to train the adjusted vital sign signal extraction model by adopting the next sample training data.
3. The method according to claim 1 or 2, wherein the vital sign data and the environment data are data matrices composed of a plurality of element data, the element data of the vital sign data and the element data of the environment data are in one-to-one correspondence, and the obtaining of the sample training data based on the vital sign data and the environment data includes:
and aiming at any data matrix in the vital sign data and the environment data, performing data processing on the data matrix to extract singular entropy, wavelet packet scale entropy and sample entropy as the sample training data.
4. The method of claim 3, wherein the singular entropy extracting process comprises:
determining a matrix of singular values based on the data matrix;
determining the singular entropy of the data matrix based on the singular value matrix;
wherein the singular entropy is determined using the following formula:
wherein, the sigma i Representing the ith singular value, σ, in the matrix of singular values 1 +σ 2 +…σ k Representing the summation of all singular values in the matrix of singular values.
5. The method of claim 3, wherein extracting the wavelet packet scale entropy comprises:
decomposing the first-order right singular vector in the data matrix to obtain a wavelet packet decomposition coefficient vector;
determining the wavelet packet scale entropy based on the wavelet packet decomposition coefficient vector;
wherein the wavelet packet scale entropy is determined by the following formula:
wherein, the w i And the wavelet coefficients represent the wavelet packet decomposition coefficient vectors, j is a wavelet packet node, and k is the number of wavelet packet decomposition layers.
6. The method of claim 3, wherein extracting sample entropies of the data matrix comprises:
expanding the data matrix into a one-dimensional vector sequence;
determining a first number of samples and a second number of samples based on the sequence of vectors;
determining the sample entropy based on the first number of samples and the second number of samples;
wherein the sample entropy is determined using the following formula:
wherein, A is m (r) represents the first number of samples of the sample entropy reconstruction dimension m, said B m (r) represents the second number of samples of the sample entropy reconstruction dimension m + 1.
7. The method of claim 1, wherein the determining a sample signature value of the vital sign data comprises:
judging whether the element data in the vital sign data meet the vital sign conditions;
in response to element data in the vital sign data meeting the vital sign condition, determining that a tag value of the element data is 1;
responding to the vital sign data that the bit data does not meet the preset condition, and determining that the tag value of the element data is 0;
generating a sample label value of the vital sign data based on the label values of all the element data.
8. A training device for a vital sign signal extraction model is characterized by comprising:
the first acquisition module is used for acquiring a vital sign signal extraction model to be trained;
the second acquisition module is used for acquiring a training sample set, and the training sample set comprises a plurality of groups of vital sign data and environmental data;
the determining module is used for acquiring sample training data based on the vital sign data and the environment data and determining a sample label value of the vital sign data;
and the training module is used for training the vital sign signal extraction model based on the sample label value and the sample training data until the training is finished to obtain a target vital sign signal extraction model.
9. An electronic device comprising a memory, a processor;
wherein the processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory for implementing the method according to any one of claims 1 to 7.
10. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210529418.6A CN114970616A (en) | 2022-05-16 | 2022-05-16 | Training method and device for vital sign signal extraction model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210529418.6A CN114970616A (en) | 2022-05-16 | 2022-05-16 | Training method and device for vital sign signal extraction model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114970616A true CN114970616A (en) | 2022-08-30 |
Family
ID=82982600
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210529418.6A Pending CN114970616A (en) | 2022-05-16 | 2022-05-16 | Training method and device for vital sign signal extraction model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114970616A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140275832A1 (en) * | 2013-03-14 | 2014-09-18 | Koninklijke Philips N.V. | Device and method for obtaining vital sign information of a subject |
| CN111680651A (en) * | 2020-06-12 | 2020-09-18 | 武汉星巡智能科技有限公司 | Non-contact vital sign detection method, device, storage medium and system |
| US20200320286A1 (en) * | 2019-04-02 | 2020-10-08 | Wistron Corporation | Living body detection method and living body detection system |
| CN112754431A (en) * | 2020-12-31 | 2021-05-07 | 杭州电子科技大学 | Respiration and heartbeat monitoring system based on millimeter wave radar and lightweight neural network |
| CN113887397A (en) * | 2021-09-29 | 2022-01-04 | 中山大学中山眼科中心 | Classification method and classification system of electrophysiological signals based on ocean predator algorithm |
| CN113935359A (en) * | 2020-06-29 | 2022-01-14 | 深圳辉煌耀强科技有限公司 | Vital sign detection method and vital sign detection equipment |
| CN113935231A (en) * | 2021-09-14 | 2022-01-14 | 南京邮电大学 | Large-scale structural damage monitoring and evaluating method based on sparse array |
-
2022
- 2022-05-16 CN CN202210529418.6A patent/CN114970616A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140275832A1 (en) * | 2013-03-14 | 2014-09-18 | Koninklijke Philips N.V. | Device and method for obtaining vital sign information of a subject |
| US20200320286A1 (en) * | 2019-04-02 | 2020-10-08 | Wistron Corporation | Living body detection method and living body detection system |
| CN111680651A (en) * | 2020-06-12 | 2020-09-18 | 武汉星巡智能科技有限公司 | Non-contact vital sign detection method, device, storage medium and system |
| CN113935359A (en) * | 2020-06-29 | 2022-01-14 | 深圳辉煌耀强科技有限公司 | Vital sign detection method and vital sign detection equipment |
| CN112754431A (en) * | 2020-12-31 | 2021-05-07 | 杭州电子科技大学 | Respiration and heartbeat monitoring system based on millimeter wave radar and lightweight neural network |
| CN113935231A (en) * | 2021-09-14 | 2022-01-14 | 南京邮电大学 | Large-scale structural damage monitoring and evaluating method based on sparse array |
| CN113887397A (en) * | 2021-09-29 | 2022-01-04 | 中山大学中山眼科中心 | Classification method and classification system of electrophysiological signals based on ocean predator algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111767930A (en) | Method for detecting abnormal time series data of Internet of things and related equipment thereof | |
| CN108447057B (en) | SAR image change detection method based on significance and depth convolution network | |
| KR20190072652A (en) | Information processing apparatus and information processing method | |
| CN108171119B (en) | SAR image change detection method based on residual error network | |
| CN112861066B (en) | Machine learning and FFT (fast Fourier transform) -based blind source separation information source number parallel estimation method | |
| CN113239022B (en) | Method and device for complementing missing data in medical diagnosis, electronic device and medium | |
| CN114399066A (en) | Mechanical equipment predictability maintenance system and maintenance method based on weak supervision learning | |
| CN112580780A (en) | Model training processing method, device, equipment and storage medium | |
| Bing et al. | Short-term traffic flow local prediction based on combined kernel function relevance vector machine model | |
| CN116796275A (en) | Multi-mode time sequence anomaly detection method for industrial equipment | |
| CN115795401A (en) | Ocean pasture full-factor monitoring sensor multi-data fusion system | |
| CN112711032B (en) | Radar target detection method and system based on graph data and GCN | |
| CN113361194B (en) | Sensor drift calibration method based on deep learning, electronic equipment and storage medium | |
| CN111291481B (en) | Bayesian model-based structure early warning analysis method | |
| CN114970616A (en) | Training method and device for vital sign signal extraction model | |
| CN113740671A (en) | Fault arc identification method based on VMD and ELM | |
| CN110990383A (en) | Similarity calculation method based on industrial big data set | |
| CN114636890B (en) | Case-based reasoning fault positioning method and system, storage medium and electronic equipment | |
| CN112380949B (en) | Method and system for detecting arrival time of micro-seismic waves | |
| WO2023025451A1 (en) | Method and device for generating synthetic training data for an ultrasonic sensor model | |
| CN110738191A (en) | object classification method, device, equipment and medium based on sonar signals | |
| Zhao et al. | Machine learning-enabled calibration of river routing model parameters | |
| CN113052219B (en) | Abnormal track detection method and device and electronic equipment | |
| JP5191443B2 (en) | Multidimensional signal pair compression apparatus, multidimensional signal pair compression method, and program | |
| CN117349770B (en) | Structural health monitoring multi-strain sensor data anomaly detection and repair method |
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 |