Training method and device for vital sign signal extraction model

Abstract

The present disclosure proposes a training method, device, electronic device and storage medium for a vital sign signal extraction model, which relates to the field of life signal detection technology. The method includes: obtaining a vital sign signal extraction model to be trained; obtaining a training sample set, wherein the training sample set includes multiple groups of vital sign data and environmental data; obtaining sample training data based on the vital sign data and environmental data, and determining the sample label value of the vital sign data at the same time; training the vital sign signal extraction model based on the sample label value and the sample training data until the training is completed to obtain the target vital sign signal extraction model. The target vital sign signal extraction model is generated through training, and the collected data is processed based on the model to determine whether the collected object has vital signs. This can greatly improve the time for rescue data processing, improve data processing efficiency, reduce processing costs, and at the same time buy more time for rescue.

Description

Technical Field

[0001] The present disclosure relates to the technical field of vital sign detection, and in particular to a training method, device, electronic device and storage medium for a vital sign signal extraction model. Background Art

[0002] Non-contact life detection technology is a new technology that has emerged in recent years with the needs of medical engineering, military and social development. It breaks through the current detection methods and technologies and uses human life parameters such as breathing, heartbeat and human movement signals to explore living things without contacting the target body.

[0003] In current technology, the signal space of heartbeat and breathing in the weak echo signal of life is orthogonal to the noise space, so the multiple signal classification (MUSIC) algorithm can be used for spectrum estimation. However, the above method takes a lot of time to process the data, which greatly increases the difficulty of rescue and even delays the best rescue time.

[0004] Public Content

[0005] The present disclosure aims to solve one of the technical problems in the related art at least to some extent.

[0006] To this end, an object of the present disclosure is to propose a training method for a vital sign signal extraction model.

[0007] The second objective of the present disclosure is to provide a training device for a vital sign signal extraction model.

[0008] A third objective of the present disclosure is to provide an electronic device.

[0009] A fourth object of the present disclosure is to provide a non-transitory computer-readable storage medium.

[0010] A fifth object of the present disclosure is to provide a computer program product.

[0011] To achieve the above-mentioned purpose, the first aspect of the present disclosure proposes a training method for a vital sign signal extraction model, including: obtaining a vital sign signal extraction model to be trained; obtaining a training sample set, the training sample set including multiple groups of vital sign data and environmental data; obtaining sample training data based on the vital sign data and the environmental data, and determining the sample label value of the vital sign data; training the vital sign signal extraction model based on the sample label value and the sample training data until the training is completed and the target vital sign signal extraction model is obtained.

[0012] According to one embodiment of the present disclosure, a vital sign signal extraction model is trained based on sample signature values ​​and sample training data, including: 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; adjusting model parameters of the vital sign signal extraction model according to the loss value, and returning to train the adjusted vital sign signal extraction model using the next sample training data.

[0013] According to one embodiment of the present disclosure, the vital signs data and the environmental data are respectively data matrices composed of multiple element data, and the element data of the vital signs data correspond to the element data of the environmental data one-to-one. Sample training data is obtained based on the vital signs data and the environmental data, including: for any data matrix in the vital signs data and the environmental data, any data matrix is ​​processed to extract singular entropy, wavelet packet scale entropy and sample entropy as sample training data.

[0014] According to one embodiment of the present disclosure, the singular entropy extraction process includes: determining a singular value matrix based on a 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: Among them, σ i represents the i-th singular value in the singular value matrix, σ1+σ2+…σ k Represents the sum of all singular values ​​in the singular value matrix.

[0015] According to an embodiment of the present disclosure, extracting wavelet packet scale entropy includes: 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 using the following formula: Among them, w i Represents the wavelet coefficient of the wavelet packet decomposition coefficient vector, j is the wavelet packet node, and k is the number of wavelet packet decomposition layers.

[0016] According to one embodiment of the present disclosure, extracting sample entropy of a data matrix includes: expanding the data matrix into a one-dimensional vector sequence; determining a first sample number and a second sample number based on the vector sequence; determining sample entropy based on the first sample number and the second sample number; wherein the sample entropy is determined using the following formula: Among them, A m (r) represents the number of first samples with sample entropy reconstruction dimension m, B m (r) represents the number of second samples whose sample entropy reconstruction dimension is m+1.

[0017] According to one embodiment of the present disclosure, extracting sample entropy of a data matrix includes: determining whether element data in vital signs data meets vital signs conditions; in response to element data in the vital signs data meeting the vital signs conditions, determining that the label value of the element data is 1; in response to site data in the vital signs data not meeting preset conditions, determining that the label value of the element data is 0; and generating sample signature values ​​of the vital signs data based on the label values ​​of all element data.

[0018] To achieve the above-mentioned purpose, the second aspect embodiment of the present disclosure proposes a training device for a vital sign signal extraction model, including: a first acquisition module, used to acquire the vital sign signal extraction model to be trained; a second acquisition module, used to acquire a training sample set, the training sample set including multiple groups of vital sign data and environmental data; a determination module, used to acquire sample training data based on the vital sign data and environmental data, and determine the sample label value of the vital sign data; a training module, used to train the vital sign signal extraction model based on the sample label value and the sample training data, until the training is completed and the target vital sign signal extraction model is obtained.

[0019] To achieve the above-mentioned purpose, the third aspect embodiment of the present disclosure proposes an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to implement the training method of the vital sign signal extraction model as described in the first aspect embodiment of the present disclosure.

[0020] To achieve the above-mentioned purpose, the fourth aspect embodiment of the present disclosure proposes a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to implement the training method of the vital sign signal extraction model as described in the first aspect embodiment of the present disclosure.

[0021] To achieve the above-mentioned purpose, the fifth aspect embodiment of the present disclosure proposes a computer program product, including a computer program, which, when executed by a processor, is used to implement the training method of the vital sign signal extraction model as described in the first aspect embodiment of the present disclosure.

[0022] Through training, a target vital sign signal extraction model is generated, and the collected data is processed based on the model to determine whether the collected object has vital signs. This can greatly improve the time for rescue data processing, improve data processing efficiency, reduce processing costs, and at the same time buy more time for rescue. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a schematic diagram of a training method for a vital sign signal extraction model according to an embodiment of the present disclosure;

[0024] Figure 2 is a schematic diagram of another training method for a vital sign signal extraction model according to an embodiment of the present disclosure;

[0025] Figure 3 is a schematic diagram of another training method for a vital sign signal extraction model according to an embodiment of the present disclosure;

[0026] Figure 4 is a schematic diagram of a training device for a vital sign signal extraction model according to an embodiment of the present disclosure;

[0027] Figure 5 is a schematic diagram of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0028] Embodiments of the present disclosure are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, and should not be construed as limiting the present disclosure.

[0029] Weak signal detection has always been a hot and difficult problem in the field of signal processing. A large amount of clutter and noise will cause strong interference and even cover up useful weak life signals. At the same time, in the process of using ultra-wideband radar life detectors to detect life signals, a large amount of radar echo data needs to be processed. How to ensure accurate extraction of life signals while quickly performing detection work, ensuring the timeliness of the algorithm, and gaining enough time for emergency rescue work are also major problems that need to be overcome in this field.

[0030] The present invention proposes a training method for a vital sign signal extraction model, which solves the problem of long data analysis and processing time and low accuracy. It can accurately locate living organisms and extract heartbeat and respiratory signals, and has high accuracy, timeliness, robustness and signal-to-noise ratio.

[0031] Figure 1 A schematic diagram of an exemplary implementation of a training method for a vital sign signal extraction model proposed in the present disclosure is shown as follows: Figure 1 As shown, the training method of the vital sign signal extraction model includes the following steps:

[0032] S101, obtaining a vital sign signal extraction model to be trained.

[0033] In the embodiment of the present disclosure, the vital sign signal extraction model to be trained is set in advance. The vital sign signal extraction model to be trained can be of various types, for example, it can be a neural network model, or a support vector machine model (SVM), etc., and no limitation is made here.

[0034] Among them, the SVM model is the best off-the-shelf classifier, which means it can be used directly without modification. It can also achieve a lower error rate, and SVM can make good classification decisions for data points outside the training set.

[0035] The neural network model has strong robustness and fault tolerance, self-learning, self-organization, and adaptability, which enables the network to handle uncertain or unknown systems and can fully approximate arbitrarily complex nonlinear relationships. It also has strong information integration capabilities, can process quantitative and qualitative information at the same time, and can coordinate multiple input information relationships well, making it suitable for multi-information fusion and multimedia technology.

[0036] Different models may have different focuses. Therefore, when selecting a vital sign signal extraction model, the corresponding reference model can be selected according to the focus actually required to achieve the best data processing effect.

[0037] S102, obtaining a training sample set, where the training sample set includes multiple groups of vital sign data and environmental data.

[0038] In the embodiments of the present disclosure, the vital signs data may include a variety of data, for example, may include body temperature data, heart rate data and breathing data, etc., which are not limited here and can be set according to actual needs.

[0039] Environmental data may include a variety of data, for example, may include location data, humidity data and waveform data, etc., which are not limited here and can be set according to actual needs.

[0040] The vital sign data and the environmental data can represent the attribute information of the test point. In the embodiment of the present disclosure, the vital sign data and the environmental data are respectively data matrices composed of multiple element data, and the element data of the vital sign data corresponds to the element data of the environmental data one by one. For example, as shown in the following two data matrices:

[0041] The extracted vital signs data is in the following form:

[0042] (15≥M≥5, 20≥N≥10)

[0043] The extracted environmental data is in the following form:

[0044] (15≥M≥5, 20≥N≥10).

[0045] In the disclosed embodiment, the training sample set may be real-time data actually collected by a life detector device, and the life detector device may be an ultra-wideband radar life detector, an infrared life detector, or an audio life detector, etc., without any limitation here.

[0046] Data is collected by life detectors in different collection scenarios. For example, the collection scenarios may be life forms in a static state, passing through obstacles such as wood, concrete, and water, multiple overlapping life forms, life forms at a distance of more than 10 meters, and life forms in a non-static state. It is understandable that the data collected in different scenarios may be different. When there is a lot of clutter and noise, passing through walls and obstacles, and detecting multiple life forms at the same time, the collected vital signs signals are further obscured and difficult to be identified. When in a relatively open environment or an environment with less clutter and noise, the collected data is clearer and it is easier to analyze the vital signs signals in the current data.

[0047] Optionally, the training sample set may also be pre-set data. For example, the training sample set may be obtained by connecting to a training sample library. The training sample library is a pre-set database. The training sample library may be stored in the cloud or in the storage space of an electronic device.

[0048] S103, acquiring sample training data based on the vital sign data and the environmental data, and determining sample label values ​​of the vital sign data.

[0049] 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 sample training data. It should be noted that the training data can be of various types. In the embodiment of the present disclosure, the training data can be singular entropy, wavelet packet scale entropy of the first-order right singular vector, and sample entropy, etc., which are not limited here.

[0050] In the embodiment of the present disclosure, when the subject has vital signs, the sample label value may be a vital sign signal, and when the subject has no vital signs, the sample label value may be a no vital sign signal. The specific data may be determined according to actual settings and is not limited here.

[0051] There are many methods for determining vital signs. For example, when the body temperature data in the vital signs data is greater than the preset body temperature value, the vital signs data can be considered as a vital sign; when the heartbeat data in the vital signs data has a heartbeat value, the vital signs data can be considered as a vital sign; when the breathing data in the vital signs data has breathing, the vital signs data can be considered as a vital sign, etc. There is no limitation here, and the specific setting can be made according to the actual situation.

[0052] Optionally, the sample label value of the vital sign data may also be obtained simultaneously when the vital sign data is obtained.

[0053] S104, training the vital sign signal extraction model based on the sample label value and the sample training data until the training is completed to obtain the target vital sign signal extraction model.

[0054] The training of the vital sign signal extraction model is an iterative process. The training is performed by continuously adjusting the network parameters of the model until the overall loss function value of the model is less than the preset value, or the overall loss function value of the model no longer changes or changes slowly, the model converges, and a trained target vital sign signal extraction model is obtained.

[0055] In the disclosed embodiment, the vital sign signal extraction model to be trained is first obtained, and then a training sample set is obtained, the training sample set including multiple groups of vital sign data and environmental data, and then sample training data is obtained based on the vital sign data and environmental data, and the 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 target vital sign signal extraction model is obtained after the training is completed. The target vital sign signal extraction model is generated through training, and in actual operation, the collected data can be processed by the model to determine whether the collected object has vital signs, thereby greatly improving the time for rescue data processing, improving data processing efficiency, reducing processing costs, and at the same time buying more time for rescue.

[0056] In the embodiment of the present disclosure, the sample signature value of the vital signs data is determined to determine whether the element data in the vital signs data meets the vital signs condition; in response to the element data in the vital signs data meeting the vital signs condition, the label value of the element data is determined to be 1; in response to the site data in the vital signs data not meeting the preset condition, the label value of the element data is determined to be 0; the sample signature value of the vital signs data is generated based on the label values ​​of all the element data. For example, when the vital signs data is in the following form,

[0057]

[0058] The tag value can be:

[0059]

[0060] In the above embodiment, the vital sign signal extraction model is trained based on the sample signature value and the sample training data, and the Figure 2 Explaining further, the method includes:

[0061] S201, inputting sample training data into a vital sign signal extraction model to output a predicted label value.

[0062] S202: Determine the loss value of the vital sign signal extraction model based on the predicted label value and the sample label value.

[0063] The loss function in this embodiment is set in advance and can be set according to actual needs. For example, the loss function can be a hinge loss function, a cross entropy loss function, an exponential loss function, etc., which can be selected according to actual needs and is not limited here.

[0064] 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 the cross entropy loss function or the exponential loss function, and the loss function corresponding to the SVM model may be the hinge loss function, etc.

[0065] After obtaining the predicted label value and the sample label value, the predicted label value and the sample label value may be input into a loss function to determine the loss value of the vital sign signal extraction model.

[0066] S203, adjusting the model parameters of the vital sign signal extraction model according to the loss value, and returning to train the adjusted vital sign signal extraction model using the next sample training data.

[0067] In the disclosed embodiment, after the loss value is obtained, the calculated loss value is compared with the preset value. If the loss value is greater than the preset value, the network parameters of the network to be trained are updated, and the vital sign signal extraction model to be trained is obtained based on the updated network parameters. The loss value of the model to be trained is calculated again according to the training sample set, and this iteration is repeated until the loss value is less than the preset value to obtain a trained target vital sign signal extraction model.

[0068] In the disclosed embodiment, the sample training data is first input into the vital sign signal extraction model to output the predicted label value, and then the loss value of the vital sign signal extraction model is determined based on the predicted label value and the sample label value, and finally the model parameters of the vital sign signal extraction model are adjusted according to the loss value, and the next sample training data is returned to train the adjusted vital sign signal extraction model. Thus, by continuously adjusting the model through the loss value, a more accurate target vital sign signal extraction model can be obtained, and the accuracy of data processing can be improved.

[0069] In the above embodiment, sample training data is obtained based on vital sign data and environmental data, and Figure 3 Explaining further, the method includes:

[0070] S301, for any data matrix in the vital sign data and the environmental data, data processing is performed on any data matrix to extract singular entropy, wavelet packet scale entropy and sample entropy as sample training data.

[0071] In the embodiment of the present disclosure, the process of extracting the singular entropy includes: 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.

[0072] The process of performing singular value decomposition on the vital sign data S is as follows:

[0073]

[0074]

[0075]

[0076]

[0077] The singular value matrix Σkxk has singular values ​​only on the main diagonal, and the other elements are all 0. The main diagonal elements σ1, σ2, σ3, ..., σk are k singular values ​​and satisfy σ1≥σ2≥σ3≥...≥σk. ui represents the i-th column vector of the matrix UMxk, which is called the i-th order left singular vector. vi represents the i-th column vector of the matrix VNxk, which is called the i-th order right singular vector. σi represents the i-th element of the singular value spectrum. For the present invention, the first k (min(M,N)≥k≥3) order singular values ​​σ1, σ2, σ3, ..., σk, left singular vectors u1, u2, u3, ..., uk and right singular vectors v1, v2, v3, ..., vk can be retained as decomposition results.

[0078] The formula for calculating singular entropy of vital sign data S is as follows:

[0079]

[0080] In the embodiment of the present disclosure, extracting wavelet packet scale entropy includes:

[0081] Calculate the wavelet packet scale entropy for the first-order right singular vectors of all data in the vital sign data samples and the environmental data samples;

[0082] The process of performing wavelet packet decomposition on the first-order right singular vector v1 of the vital sign data S is as follows:

[0083] Select the 'db6' wavelet and set the wavelet packet decomposition layer number to 2 for wavelet packet decomposition. The wavelet packet decomposition method is as follows:

[0084]

[0085] Among them, W00 is the original signal v1, W is the 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 the high-pass filter coefficient, and g is the low-pass filter coefficient.

[0086] The calculation formula of wavelet packet scale entropy of each node is:

[0087]

[0088] Wherein, k is the number of wavelet packet decomposition layers, j is the wavelet packet node (j=0, 1, ..., 2k-1), (in the present invention, k=2 can be taken.) is the wavelet packet decomposition coefficient vector, wi(i=1,2,…,x) is the wavelet coefficient in Wkj.

[0089] In an embodiment of the present disclosure, extracting sample entropy of a data matrix includes:

[0090] Calculate sample entropy for all data in the vital sign data sample and the environmental data sample;

[0091] The process of calculating sample entropy of vital sign data S is as follows:

[0092] The vital sign data S is expanded into a one-dimensional sequence {x1, x2, ..., xn}, the sample entropy reconstruction dimension is m = 2, the threshold is r, the standard deviation of the calculated sequence is δ, and the above one-dimensional sequence is organized into a set of vector sequences of dimension m {X m (1),X m (2),…X m (n-m+1)}.

[0093] Among them, X m (i)=[x(i),x(i+1),…x(i+m-1)], 1≤i≤n-m+1 represents the m consecutive values ​​of x starting from the i-th point.

[0094] Define vector X m (i) With X m The distance between (j) is the absolute value of the maximum difference between the corresponding elements of the two, that is:

[0095] d[X m (i),X m (j)]=max k＝0,…,m-1 (|x(i+k)-x(j+k)|)

[0096] For a given X m (i), statistics X m (i) With X mThe number of j (1≤j≤nm,j≠i) whose distance between (j) is less than r·δ, and is denoted as Bi. For 1≤i≤nm, define:

[0097]

[0098]

[0099] Increase the dimension to m+1 and calculate X m+1 (i) With X m+1 The number of j (1≤j≤nm-1,j≠i) whose distance between (j) is less than r·δ is denoted as Ai. Then define:

[0100]

[0101]

[0102] So, B m (r) is the probability that two sequences match m points under similarity tolerance r, and A m (r) is the probability that two sequences match m+1 points under similarity tolerance r. The sample entropy calculation formula is:

[0103]

[0104] Corresponding to the training methods for the vital sign signal extraction models provided in the above-mentioned embodiments, an embodiment of the present disclosure further provides a training device for a vital sign signal extraction model. Since the training device for the vital sign signal extraction model provided in the embodiment of the present disclosure corresponds to the training methods for the vital sign signal extraction models provided in the above-mentioned embodiments, the implementation methods of the above-mentioned training methods for the vital sign signal extraction models are also applicable to the training device for 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.

[0105] Figure 4 A schematic diagram of a training device for a vital sign signal extraction model proposed in the present disclosure, such as Figure 4 As shown, the training device 400 for the 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.

[0106] The first acquisition module 410 is used to acquire the vital sign signal extraction model to be trained.

[0107] The second acquisition module 420 is used to acquire a training sample set, where the training sample set includes multiple groups of vital sign data and environmental data.

[0108] The determination module 430 is used to obtain sample training data based on the vital sign data and the environmental data, and determine the sample label value of the vital sign data.

[0109] The training module 440 is used to train the vital sign signal extraction model based on the sample label value and the sample training data until the training is completed to obtain a target vital sign signal extraction model.

[0110] In one embodiment of the present disclosure, the training module 440 is further used to: input the sample training data into the vital sign signal extraction model to output a predicted label value; determine a loss value of the vital sign signal extraction model based on the predicted label value and the sample label value; adjust the model parameters of the vital sign signal extraction model according to the loss value, and return to train the adjusted vital sign signal extraction model using the next sample training data.

[0111] In one embodiment of the present disclosure, the determination module 430 is further used to: perform data processing on any data matrix in the vital sign data and the environmental data to extract singular entropy, wavelet packet scale entropy and sample entropy as the sample training data.

[0112] In one embodiment of the present disclosure, the determination module 430 is further used to: determine a singular value matrix based on the data matrix; determine a singular entropy of the data matrix based on the singular value matrix; wherein the singular entropy is determined using the following formula: Among them, σ i represents the i-th singular value in the singular value matrix, σ1+σ2+…σ k Represents the sum of all singular values ​​in the singular value matrix.

[0113] In one embodiment of the present disclosure, the determination module 430 is further used to: decompose the first-order right singular vector in the data matrix to obtain a wavelet packet decomposition coefficient vector; determine the wavelet packet scale entropy based on the wavelet packet decomposition coefficient vector; wherein the wavelet packet scale entropy is determined using the following formula: Among them, w i Represents the wavelet coefficient of the wavelet packet decomposition coefficient vector, j is the wavelet packet node, and k is the number of wavelet packet decomposition layers.

[0114] In one embodiment of the present disclosure, the determination module 430 is further used to: expand the data matrix into a one-dimensional vector sequence; determine the first sample number and the second sample number based on the vector sequence; determine the sample entropy based on the first sample number and the second sample number; wherein the sample entropy is determined using the following formula: Among them, A m(r) represents the number of first samples with sample entropy reconstruction dimension m, B m (r) represents the number of second samples whose sample entropy reconstruction dimension is m+1.

[0115] In one embodiment of the present disclosure, the determination module 430 is also used to: determine whether the element data in the vital signs data meets the vital signs conditions; in response to the element data in the vital signs data meeting the vital signs conditions, determine that the label value of the element data is 1; in response to the site data in the vital signs data not meeting the preset conditions, determine that the label value of the element data is 0; and generate a sample signature value of the vital signs data based on the label values ​​of all the element data.

[0116] In order to implement the above embodiment, the present disclosure also provides an electronic device 500, such as Figure 5 As shown, the electronic device 500 includes: a processor 501 and a memory 502 communicatively connected to the processor, the memory 502 stores instructions executable by at least one processor, and the instructions are executed by at least one processor 501 to implement a training method for a vital sign signal extraction model as in the first aspect of the embodiment of the present disclosure.

[0117] In order to implement the above embodiments, the embodiments of the present disclosure also propose a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to implement a training method for a vital sign signal extraction model as in the first aspect of the embodiments of the present disclosure.

[0118] In order to implement the above embodiments, the embodiments of the present disclosure also propose a computer program product, including a computer program, which, when executed by a processor, implements the training method of the vital sign signal extraction model as in the first aspect of the embodiments of the present disclosure.

[0119] In the description of the present disclosure, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present disclosure and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present disclosure.

[0120] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present disclosure, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0121] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0122] Although the embodiments of the present disclosure have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present disclosure. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present disclosure.

Claims

1. A training method for a vital sign signal extraction model, characterized in that: include: Obtain a vital sign signal extraction model to be trained; Acquire a training sample set, wherein the training sample set includes multiple groups of vital sign data and environmental data, wherein the vital sign data and the environmental data are data matrices composed of multiple element data, respectively, and the element data of the vital sign data corresponds to the element data of the environmental data one by one; Acquire sample training data based on the vital sign data and the environmental data, and simultaneously determine sample label values ​​of the vital sign data; Training the vital sign signal extraction model based on the sample label value and the sample training data until the training is completed to obtain a target vital sign signal extraction model; The sample training data includes wavelet packet scale entropy, and extracting the wavelet packet scale entropy includes: 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; The wavelet packet scale entropy is determined by the following formula: Among them, the w i represents the wavelet coefficient of the wavelet packet decomposition coefficient vector, j is the wavelet packet node, k is the wavelet packet decomposition layer number, is the wavelet packet decomposition coefficient vector, w i for The wavelet coefficients in , i = 1, 2, ...x.

2. The method according to claim 1, characterized in that The training of the vital sign signal extraction model based on the sample label value and the sample training data includes: 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; The model parameters of the vital sign signal extraction model are adjusted according to the loss value, and the adjusted vital sign signal extraction model is trained by returning to the next sample training data.

3. The method according to claim 1 or 2, characterized in that: The acquiring sample training data based on the vital sign data and the environmental data comprises: For any data matrix in the vital sign data and the environmental data, data processing is performed on the any data matrix to extract singular entropy, wavelet packet scale entropy and sample entropy as the sample training data.

4. The method according to claim 3, characterized in that The singular entropy extraction process includes: determining a singular value matrix based on the data matrix; determining the singular entropy of the data matrix based on the singular value matrix; The singular entropy is determined by the following formula: Among them, the σ i represents the i-th singular value in the singular value matrix, the σ1+σ2+…σ k represents the sum of all singular values ​​in the singular value matrix.

5. The method according to claim 3, characterized in that: Extracting sample entropy of the data matrix includes: 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 vector sequence; determining the sample entropy based on the first sample size and the second sample size; The sample entropy is determined using the following formula: Among them, the A m (r) represents the number of the first samples whose sample entropy reconstruction dimension is m, and the B m (r) represents the number of the second samples whose sample entropy reconstruction dimension is m+1.

6. The method according to claim 1, characterized in that The determining of the sample label value of the vital sign data comprises: Determining whether the element data in the vital sign data meets the vital sign condition; In response to element data in the vital sign data satisfying the vital sign condition, determining that a label value of the element data is 1; In response to the site data in the vital sign data not satisfying a preset condition, determining that the label value of the element data is 0; A sample label value of the vital sign data is generated based on the label values ​​of all the element data.

7. A training device for a vital sign signal extraction model, characterized in that: include: A first acquisition module is used to acquire a vital sign signal extraction model to be trained; A second acquisition module is used to acquire a training sample set, wherein the training sample set includes multiple groups of vital sign data and environmental data, wherein the vital sign data and the environmental data are data matrices composed of multiple element data, respectively, and the element data of the vital sign data corresponds to the element data of the environmental data one by one; A determination module, configured to obtain sample training data based on the vital sign data and the environmental data, and determine a sample label value of the vital sign data; A training module, used for training the vital sign signal extraction model based on the sample label value and the sample training data, until the training is completed to obtain a target vital sign signal extraction model; The sample training data includes wavelet packet scale entropy, and extracting the wavelet packet scale entropy includes: 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; The wavelet packet scale entropy is determined by the following formula: Among them, the w i represents the wavelet coefficient of the wavelet packet decomposition coefficient vector, j is the wavelet packet node, k is the wavelet packet decomposition layer number, is the wavelet packet decomposition coefficient vector, w i for The wavelet coefficients in , i = 1, 2, ...x.

8. An electronic device, characterized in that: Including memory and processor; The processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, so as to implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images