CN113080907A - Pulse wave signal processing method and device - Google Patents
Pulse wave signal processing method and device Download PDFInfo
- Publication number
- CN113080907A CN113080907A CN202110400929.3A CN202110400929A CN113080907A CN 113080907 A CN113080907 A CN 113080907A CN 202110400929 A CN202110400929 A CN 202110400929A CN 113080907 A CN113080907 A CN 113080907A
- Authority
- CN
- China
- Prior art keywords
- pulse wave
- blood pressure
- wave signal
- wave signals
- standard pulse
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- 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
- A61B5/021—Measuring pressure in heart or blood vessels
- A61B5/02108—Measuring pressure in heart or blood vessels from analysis of pulse wave characteristics
-
- 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
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- Biophysics (AREA)
- Public Health (AREA)
- Pathology (AREA)
- Cardiology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Physiology (AREA)
- Veterinary Medicine (AREA)
- Vascular Medicine (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Psychiatry (AREA)
- Signal Processing (AREA)
- Measuring Pulse, Heart Rate, Blood Pressure Or Blood Flow (AREA)
Abstract
The invention provides a pulse wave signal processing method and a device, wherein the method comprises the following steps: acquiring a pulse wave signal to be processed; preprocessing the pulse wave signals to obtain standard pulse wave signals; respectively extracting key frequency domain characteristics of the standard pulse wave signals through a frequency domain characteristic processing submodel of the blood pressure identification model, and extracting key time domain characteristics of the standard pulse wave signals through a time domain characteristic processing submodel of the blood pressure identification model; fusing the key time domain characteristics and the key frequency domain characteristics by using a characteristic fusion sub-model of the blood pressure identification model to obtain fusion characteristics of the standard pulse wave signals; and obtaining a blood pressure identification result of the standard pulse wave signal based on the fusion characteristics by using an output sub-model of the blood pressure identification model. The key frequency domain characteristics and the key time domain characteristics of the standard pulse wave signals can be extracted, blood pressure identification is carried out according to the key frequency domain characteristics and the key time domain characteristics, and the accuracy of blood pressure identification results can be effectively improved.
Description
Technical Field
The invention relates to the technical field of data processing, in particular to a pulse wave signal processing method and device.
Background
Blood pressure is one of the most important indexes in health monitoring, and monitoring blood pressure can not only monitor the physical condition of people but also prevent life style related diseases.
In the related technology, a feature engineering of the blood pressure is constructed based on prior knowledge, and then the obtained features are used as input of a machine learning model, so that the blood pressure value is predicted.
Disclosure of Invention
The invention aims to provide a pulse wave signal processing method which can improve the accuracy of a blood pressure identification result.
The invention also provides a pulse wave signal processing device which is used for ensuring the realization and the application of the method in practice.
A pulse wave signal processing method comprising:
acquiring a pulse wave signal to be processed;
preprocessing the pulse wave signals to obtain standard pulse wave signals;
extracting key frequency domain characteristics of the standard pulse wave signals through a pre-constructed frequency domain characteristic processing submodel of the blood pressure identification model, and extracting key time domain characteristics of the standard pulse wave signals through a time domain characteristic processing submodel of the blood pressure identification model;
fusing the key time domain characteristics and the key frequency domain characteristics by applying a characteristic fusion sub-model of the blood pressure identification model to obtain fusion characteristics of the standard pulse wave signals;
and obtaining the blood pressure identification result of the standard pulse wave signal based on the fusion characteristic by applying the output sub-model of the blood pressure identification model.
Optionally, the above method, wherein the preprocessing is performed on the pulse wave signal to obtain a standard pulse wave signal, includes:
and normalizing the pulse wave signals by using a pre-stored normalization parameter to obtain standard pulse wave signals, wherein the normalization parameter comprises a standard pulse wave signal average value and a standard pulse wave signal standard deviation.
The above method, optionally, the process of constructing the blood pressure identification model includes:
acquiring a training data set, wherein the training data set comprises a plurality of training pulse wave signals and a blood pressure value label of each training pulse wave signal;
and training the blood pressure recognition model by using the training data set.
The method described above, optionally, the frequency domain feature processing submodel includes a convolutional neural network and a first attention module;
the method for extracting the key frequency domain characteristics of the standard pulse wave signals through the frequency domain characteristic processing module of the pre-constructed blood pressure identification model comprises the following steps:
extracting frequency domain features of the standard pulse wave signals through the convolutional neural network;
and screening the frequency domain characteristics through the first attention module to obtain key frequency domain characteristics of the standard pulse wave signal.
In the above method, optionally, the time domain processing submodel includes a long-short term memory network and a second attention module;
the method for extracting the key time domain characteristics of the standard pulse wave signals through the pre-constructed time domain characteristic processing submodel of the blood pressure identification model comprises the following steps:
extracting time domain features of the standard pulse wave signals through the long-short term memory network;
and screening the time domain characteristics through the second attention module to obtain key time domain characteristics of the standard pulse wave signal.
A pulse wave signal processing apparatus comprising:
the acquisition unit is used for acquiring a pulse wave signal to be processed;
the preprocessing unit is used for preprocessing the pulse wave signals to obtain standard pulse wave signals;
the characteristic extraction unit is used for extracting key frequency domain characteristics of the standard pulse wave signals through a pre-constructed frequency domain characteristic processing sub-model of the blood pressure identification model and extracting key time domain characteristics of the standard pulse wave signals through a time domain characteristic processing sub-model of the blood pressure identification model;
the feature fusion unit is used for fusing the key time domain features and the key frequency domain features by applying a feature fusion sub-model of the blood pressure identification model to obtain fusion features of the standard pulse wave signals;
and the recognition unit is used for applying the output sub-model of the blood pressure recognition model to obtain the blood pressure recognition result of the standard pulse wave signal based on the fusion characteristic.
The above apparatus, optionally, the preprocessing unit includes:
and the normalization processing subunit is used for applying a pre-stored normalization parameter to perform normalization processing on the pulse wave signals to obtain standard pulse wave signals, wherein the normalization parameter comprises a standard pulse wave signal average value and a standard pulse wave signal standard deviation.
The above apparatus, optionally, further comprises:
the training unit is used for acquiring a training data set, wherein the training data set comprises a plurality of training pulse wave signals and a blood pressure value label of each training pulse wave signal; and training the blood pressure recognition model by using the training data set.
The apparatus described above, optionally, the frequency domain feature processing submodel includes a convolutional neural network and a first attention module;
the feature extraction unit includes:
a first feature extraction subunit, configured to extract, through the convolutional neural network, a frequency domain feature of the standard pulse wave signal;
and the first feature screening unit is used for screening the frequency domain features through the first attention module to obtain key frequency domain features of the standard pulse wave signals.
Optionally, the time domain processing submodel includes a long-term and short-term memory network and a second attention module;
the feature extraction unit includes:
the second characteristic extraction subunit is used for extracting the time domain characteristics of the standard pulse wave signals through the long-term and short-term memory network;
and the second feature screening subunit is used for screening the time domain features through the second attention module to obtain key time domain features of the standard pulse wave signals.
A storage medium, comprising stored instructions, wherein when the instructions are executed, the storage medium controls a device on which the storage medium is located to execute the pulse wave signal processing method.
An electronic apparatus comprising a memory, and one or more instructions, wherein the one or more instructions are stored in the memory and configured to be executed by one or more processors to perform the pulse wave signal processing method as described above.
Compared with the prior art, the invention has the following advantages:
the invention provides a pulse wave signal processing method and a device, firstly, obtaining a pulse wave signal to be processed, preprocessing the pulse wave signal and obtaining a standard pulse wave signal; then, extracting key frequency domain characteristics of the standard pulse wave signals through a pre-constructed frequency domain characteristic processing submodel of the blood pressure identification model, and extracting key time domain characteristics of the standard pulse wave signals through a time domain characteristic processing submodel of the blood pressure identification model; fusing the key time domain characteristics and the key frequency domain characteristics by using a characteristic fusion sub-model of the blood pressure identification model to obtain fusion characteristics of the standard pulse wave signals; and obtaining a blood pressure identification result of the standard pulse wave signal based on the fusion characteristic by applying an output sub-model of the blood pressure identification model. The key frequency domain features and the key time domain features of the standard pulse wave signals can be extracted, the features of the pulse wave signals can be comprehensively obtained, blood pressure identification is carried out according to the obtained key frequency domain features and the obtained key time domain features, and the accuracy of blood pressure identification results can be effectively improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
FIG. 1 is a flowchart of a method for processing a pulse wave signal according to the present invention;
FIG. 2 is a diagram illustrating a structure of a blood pressure identification model according to the present invention;
FIG. 3 is a flow chart of a process for constructing a blood pressure identification model according to the present invention;
FIG. 4 is a flowchart of a process for extracting key frequency domain features of a standard pulse wave signal according to the present invention;
FIG. 5 is a flow chart of a process for extracting key temporal features of a standard pulse wave signal according to the present invention;
FIG. 6 is a diagram of another example of a blood pressure identification model according to the present invention;
FIG. 7 is a schematic structural diagram of a pulse wave signal processing apparatus according to the present invention;
fig. 8 is a schematic structural diagram of an electronic device provided in the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In this application, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The embodiment of the invention provides a pulse wave signal processing method, which can be applied to electronic equipment, wherein a method flow chart of the method is shown in fig. 1, and the method specifically comprises the following steps:
s101: and acquiring a pulse wave signal to be processed.
The pulse wave signal can be a radial artery pressure pulse wave signal RA-PPW or a photoplethysmography pulse wave signal PPG.
S102: and preprocessing the pulse wave signals to obtain standard pulse wave signals.
In the embodiment of the present invention, one feasible way to pre-process the pulse wave signal is to apply a pre-stored normalization parameter to perform normalization processing on the pulse wave signal, so as to obtain a standard pulse wave signal of the pulse wave signal.
S103: and respectively extracting key frequency domain characteristics of the standard pulse wave signals through a pre-constructed frequency domain characteristic processing submodel of the blood pressure identification model, and extracting key time domain characteristics of the standard pulse wave signals through a time domain characteristic processing submodel of the blood pressure identification model.
In the embodiment of the present invention, the key frequency domain feature is a frequency domain feature required for identifying a blood pressure value, and the key time domain feature is a time domain feature required for identifying a blood pressure value.
The structural schematic diagram of the blood pressure identification model is shown in fig. 2, and the blood pressure identification model comprises a frequency domain feature processing sub-model, a time domain feature processing sub-model, a feature fusion sub-model and an output sub-model.
S104: and fusing the key time domain characteristics and the key frequency domain characteristics by applying a characteristic fusion sub-model of the blood pressure identification model to obtain fusion characteristics of the standard pulse wave signals.
In the embodiment of the invention, the characteristic with discrimination in the standard pulse wave signal can be obtained by fusing the key time domain characteristic and the key frequency domain characteristic.
S105: and obtaining the blood pressure identification result of the standard pulse wave signal based on the fusion characteristic by applying the output sub-model of the blood pressure identification model.
In the embodiment of the present invention, the blood pressure identification result may be a blood pressure value tag, and the input sub-model may select the blood pressure tag with the highest probability from the blood pressure value tags according to the fusion characteristics for outputting.
The output submodel may be various types of classifiers.
By applying the method provided by the embodiment of the invention, the key frequency domain characteristics and the key time domain characteristics of the standard pulse wave signals can be extracted, the characteristics of the pulse wave signals can be comprehensively obtained, the blood pressure identification is further carried out according to the obtained key frequency domain characteristics and the obtained key time domain characteristics, and the accuracy of the blood pressure identification result can be effectively improved.
In the method provided in the embodiment of the present invention, based on the above scheme, optionally, the preprocessing the pulse wave signal to obtain a standard pulse wave signal includes:
and normalizing the pulse wave signals by using a pre-stored normalization parameter to obtain standard pulse wave signals, wherein the normalization parameter comprises a standard pulse wave signal average value and a standard pulse wave signal standard deviation.
In the embodiment of the present invention, the pulse wave signal may be normalized by a preset normalization formula, which is as follows:
wherein, s'kIs a standard pulse wave signal, skIs the pulse wave signal, mu is the mean value of the standard pulse wave signal, and sigma is the standard pulseStandard deviation of wave signal, n is positive integer.
Optionally, the standard pulse wave signal mean value and the standard pulse wave signal standard deviation may be based on a plurality of historical pulse wave signals s collected in advance1,…,snThus obtaining the product.
In particular, the normalized signal sequence values are subject to a normal distribution.
In the method provided in the embodiment of the present invention, based on the above scheme, optionally, the construction process of the blood pressure identification model, as shown in fig. 3, may include:
s301: a training data set is obtained, wherein the training data set comprises a plurality of training pulse wave signals and blood pressure value labels of each training pulse wave signal.
In the embodiment of the present invention, one feasible way to obtain the training data set is as follows: the method comprises the steps of obtaining at least one original pulse wave sequence, intercepting each original pulse wave sequence to obtain a plurality of original pulse wave signals with preset lengths, conducting standardization processing on each original pulse wave signal to obtain training pulse wave signals, determining the blood pressure value of each training pulse wave signal, determining the blood pressure value label of each training pulse wave signal based on the blood pressure value of each training pulse wave, and forming a training data set by each training pulse wave signal and the blood pressure value label of each training pulse wave signal.
S302: and training the blood pressure recognition model by using the training data set.
In the embodiment of the present invention, each training pulse wave signal in the training data set and the blood pressure value label of each training pulse signal may be applied to train the blood pressure identification model until the blood pressure identification model meets the set training completion condition, where the training completion condition may be that the training number is greater than a preset number threshold, or that the identification accuracy of the blood pressure identification model is greater than a preset accuracy threshold, and the like.
In the method provided in the embodiment of the present invention, based on the above scheme, optionally, the frequency domain feature processing submodel includes a convolutional neural network and a first attention module;
the process of extracting the key frequency domain features of the standard pulse wave signal through the pre-constructed frequency domain feature processing module of the blood pressure identification model, as shown in fig. 4, may include:
s401: and extracting the frequency domain characteristics of the standard pulse wave signal through the convolutional neural network.
The convolutional neural network can be a one-dimensional convolutional neural network 1D-CNN, and translation invariant features in a certain direction on the standard pulse wave signal can be extracted by adopting the 1D-CNN, namely, the frequency domain features of the standard pulse wave signal can be accurately extracted.
S402: and screening the frequency domain characteristics through the first attention module to obtain key frequency domain characteristics of the standard pulse wave signal.
In an embodiment of the present invention, the first attention module may be a soft attention mechanism SAM module, the frequency domain feature may be in the form of a multidimensional vector, an attention weight corresponding to a vector of each dimension in the frequency domain feature is set in the first attention module, and the key frequency domain feature of the standard pulse wave signal may be obtained by screening the frequency domain feature through the attention weight corresponding to the vector of the unique dimension.
In the method provided in the embodiment of the present invention, based on the above scheme, optionally, the time domain processing submodel includes a long-term and short-term memory network and a second attention module;
the extracting of the key time domain feature of the standard pulse wave signal through the pre-constructed time domain feature processing submodel of the blood pressure identification model, as shown in fig. 5, includes:
s501: and extracting the time domain characteristics of the standard pulse wave signals through the long-short term memory network.
In the embodiment of the present invention, the long-short term memory network may be a GRU model, and the standard pulse wave signal is a time sequence, so that the long-short term memory network may be used to extract the time domain feature of the standard pulse wave signal.
S502: and screening the time domain characteristics through the second attention module to obtain key time domain characteristics of the standard pulse wave signal.
In an embodiment of the present invention, the second attention module may be a hard attention mechanism HAM module, a feature weight may be calculated by the second attention module, and the key time-domain feature of the standard pulse wave signal may be obtained by screening the time-domain feature through the feature weight.
In the pulse wave signal processing method provided by the embodiment of the invention, in the practical application process, collected historical data can be labeled with a sample and then a blood pressure recognition model is trained, referring to fig. 6, which is another structural example diagram of the blood pressure recognition model provided by the embodiment of the invention.
Standardizing RA-PPW signals with fixed length acquired in real time by using standardized parameters shared by a training set to serve as input of a blood pressure identification model; and obtaining a blood pressure value label output by the regression prediction of the blood pressure identification model.
The bandwidth of the RA-PPW signal is 0.5-4Hz and the bandwidth of the respiratory cycle signal is 0-0.5Hz, so that the RA-PPW signal has rich characteristic information in a frequency domain, and meanwhile, the RA-PPW signal is also a time sequence and also has rich time domain characteristics. Therefore, the time domain feature and the frequency domain feature in the RA-PPW signal are sufficient for blood pressure identification, and the generalization performance of a blood pressure identification model capable of continuous blood pressure identification can be improved.
Specifically, the first attention module may be a SAM module, and the second attention module may be a HAM module, where the SAM may calculate an attention weight for all dimensions of the input vector, and assign different weights according to importance, and the HAM may calculate a unique determination weight for the input vector, and in the method of continuous blood pressure measurement using the RA-PPW signal, the CNN has a plurality of convolution kernels, and each convolution kernel is represented by one channel, that is, a dimension is added to the input, and is more suitable for using the SAM; and the output of the LSTM and the input are vectors with the same dimension, so that the feature screening is better by using the HAM.
Therefore, the time-frequency characteristics obtained by performing one-dimensional convolution on the one-dimensional RA-PPW signal are subjected to characteristic screening through SAM, the time-domain characteristics of the one-dimensional RA-PPW signal extracted by GRU are subjected to characteristic screening through HAM, then the two characteristics are fused by using the full-link layer with the same unit number, and finally regression is performed with the actually measured blood pressure value.
In the embodiment of the invention, the feature extraction of the 1D-CNN and the GRU is fused, and the blood pressure identification model obtained by analyzing the importance degree of the fused features by adopting an attention mechanism can well solve the problems of the defects of single model in the time-frequency domain feature extraction and poor generalization performance caused by model overfitting possibly caused by redundant features.
By applying the method provided by the embodiment of the invention, frequency domain feature screening under a soft attention mechanism and time domain feature screening under a hard attention mechanism can be provided based on features extracted by a Convolutional Neural Network (CNN) and a long-short term memory network (GRU), then information fusion is carried out through the same full-connection unit to obtain a deep learning model FSA-THA, information on the time-frequency domain of the RA-PPW signal is fully utilized, the advantages of each network are fully utilized by a blood pressure identification model, the CNN is adopted to extract the frequency domain feature of the RA-PPW signal, the GRU is adopted to extract the time sequence feature of the RA-PPW signal, and different attention mechanisms are introduced to complete time-frequency information screening and then fusion key feature extraction is carried out, so that the optimal feature representation of the RA-PPW signal for blood pressure measurement is obtained, and the problem of insufficient expression of discrimination features existing in artificial features is avoided by the model, meanwhile, the generalization performance of the model is improved, so that the model provided by the invention is more suitable for continuous blood pressure monitoring based on the RA-PPW signal.
Corresponding to the method shown in fig. 1, an embodiment of the present invention further provides a pulse wave signal processing apparatus for implementing the method shown in fig. 1, where the pulse wave signal processing apparatus provided in the embodiment of the present invention may be applied to an electronic device, and a schematic structural diagram of the pulse wave signal processing apparatus is shown in fig. 7, and specifically includes:
an acquisition unit 701 configured to acquire a pulse wave signal to be processed;
a preprocessing unit 702, configured to preprocess the pulse wave signal to obtain a standard pulse wave signal;
a feature extraction unit 703, configured to extract a key frequency domain feature of the standard pulse wave signal through a frequency domain feature processing sub-model of a pre-constructed blood pressure identification model, and extract a key time domain feature of the standard pulse wave signal through a time domain feature processing sub-model of the blood pressure identification model, respectively;
a feature fusion unit 704, configured to fuse the key time domain feature and the key frequency domain feature by using a feature fusion submodel of the blood pressure identification model, so as to obtain a fusion feature of the standard pulse wave signal;
the identification unit 705 is configured to apply the output sub-model of the blood pressure identification model to obtain a blood pressure identification result of the standard pulse wave signal based on the fusion feature.
By applying the device provided by the embodiment of the invention, the key frequency domain characteristics and the key time domain characteristics of the standard pulse wave signals can be extracted, the characteristics of the pulse wave signals can be comprehensively obtained, the blood pressure identification is further carried out according to the obtained key frequency domain characteristics and the obtained key time domain characteristics, and the accuracy of the blood pressure identification result can be effectively improved.
In an embodiment provided by the present invention, based on the above scheme, specifically, the preprocessing unit 702 includes:
and the normalization processing subunit is used for applying a pre-stored normalization parameter to perform normalization processing on the pulse wave signals to obtain standard pulse wave signals, wherein the normalization parameter comprises a standard pulse wave signal average value and a standard pulse wave signal standard deviation.
In an embodiment provided by the present invention, based on the above scheme, specifically, the method further includes:
the training unit is used for acquiring a training data set, wherein the training data set comprises a plurality of training pulse wave signals and a blood pressure value label of each training pulse wave signal; and training the blood pressure recognition model by using the training data set.
In an embodiment provided by the present invention, based on the above scheme, specifically, the frequency domain feature processing submodel includes a convolutional neural network and a first attention module;
the feature extraction unit 704 includes:
a first feature extraction subunit, configured to extract, through the convolutional neural network, a frequency domain feature of the standard pulse wave signal;
and the first feature screening unit is used for screening the frequency domain features through the first attention module to obtain key frequency domain features of the standard pulse wave signals.
In an embodiment provided by the present invention, based on the above scheme, specifically, the time domain processing submodel includes a long-term and short-term memory network and a second attention module;
the feature extraction unit 704 includes:
the second characteristic extraction subunit is used for extracting the time domain characteristics of the standard pulse wave signals through the long-term and short-term memory network;
and the second feature screening subunit is used for screening the time domain features through the second attention module to obtain key time domain features of the standard pulse wave signals.
The specific principle and the implementation process of each unit and module in the pulse wave signal processing apparatus disclosed in the above embodiment of the present invention are the same as the pulse wave signal processing method disclosed in the above embodiment of the present invention, and reference may be made to the corresponding parts in the pulse wave signal processing method provided in the above embodiment of the present invention, which are not described herein again.
The embodiment of the invention also provides a storage medium, which comprises stored instructions, wherein when the instructions are executed, the equipment where the storage medium is located is controlled to execute the pulse wave signal processing method.
An electronic device is provided in an embodiment of the present invention, and the structural diagram of the electronic device is shown in fig. 8, which specifically includes a memory 801 and one or more instructions 802, where the one or more instructions 802 are stored in the memory 801 and configured to be executed by the one or more processors 803 to perform the following operations:
acquiring a pulse wave signal to be processed;
preprocessing the pulse wave signals to obtain standard pulse wave signals;
extracting key frequency domain characteristics of the standard pulse wave signals through a pre-constructed frequency domain characteristic processing submodel of the blood pressure identification model, and extracting key time domain characteristics of the standard pulse wave signals through a time domain characteristic processing submodel of the blood pressure identification model;
fusing the key time domain characteristics and the key frequency domain characteristics by applying a characteristic fusion sub-model of the blood pressure identification model to obtain fusion characteristics of the standard pulse wave signals;
and obtaining the blood pressure identification result of the standard pulse wave signal based on the fusion characteristic by applying the output sub-model of the blood pressure identification model.
It should be noted that, in the present specification, the embodiments are all described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments may be referred to each other. For the device-like embodiment, since it is basically similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.
Finally, it should also be noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
For convenience of description, the above devices are described as being divided into various units by function, and are described separately. Of course, the functions of the units may be implemented in the same software and/or hardware or in a plurality of software and/or hardware when implementing the invention.
From the above description of the embodiments, it is clear to those skilled in the art that the present invention can be implemented by software plus necessary general hardware platform. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which may be stored in a storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method according to the embodiments or some parts of the embodiments.
The pulse wave signal processing method provided by the invention is described in detail above, and the principle and the implementation of the invention are explained in the present document by applying specific examples, and the description of the above examples is only used to help understanding the method and the core idea of the invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (10)
1. A pulse wave signal processing method, comprising:
acquiring a pulse wave signal to be processed;
preprocessing the pulse wave signals to obtain standard pulse wave signals;
extracting key frequency domain characteristics of the standard pulse wave signals through a pre-constructed frequency domain characteristic processing submodel of the blood pressure identification model, and extracting key time domain characteristics of the standard pulse wave signals through a time domain characteristic processing submodel of the blood pressure identification model;
fusing the key time domain characteristics and the key frequency domain characteristics by applying a characteristic fusion sub-model of the blood pressure identification model to obtain fusion characteristics of the standard pulse wave signals;
and obtaining the blood pressure identification result of the standard pulse wave signal based on the fusion characteristic by applying the output sub-model of the blood pressure identification model.
2. The method according to claim 1, wherein the preprocessing the pulse wave signal to obtain a standard pulse wave signal comprises:
and normalizing the pulse wave signals by using a pre-stored normalization parameter to obtain standard pulse wave signals, wherein the normalization parameter comprises a standard pulse wave signal average value and a standard pulse wave signal standard deviation.
3. The method of claim 1, wherein the construction of the blood pressure identification model comprises:
acquiring a training data set, wherein the training data set comprises a plurality of training pulse wave signals and a blood pressure value label of each training pulse wave signal;
and training the blood pressure recognition model by using the training data set.
4. The method of claim 1, wherein the frequency domain feature processing submodel comprises a convolutional neural network and a first attention module;
the method for extracting the key frequency domain characteristics of the standard pulse wave signals through the frequency domain characteristic processing module of the pre-constructed blood pressure identification model comprises the following steps:
extracting frequency domain features of the standard pulse wave signals through the convolutional neural network;
and screening the frequency domain characteristics through the first attention module to obtain key frequency domain characteristics of the standard pulse wave signal.
5. The method of claim 1, wherein the temporal processing submodel comprises a long short term memory network and a second attention module;
the method for extracting the key time domain characteristics of the standard pulse wave signals through the pre-constructed time domain characteristic processing submodel of the blood pressure identification model comprises the following steps:
extracting time domain features of the standard pulse wave signals through the long-short term memory network;
and screening the time domain characteristics through the second attention module to obtain key time domain characteristics of the standard pulse wave signal.
6. A pulse wave signal processing apparatus characterized by comprising:
the acquisition unit is used for acquiring a pulse wave signal to be processed;
the preprocessing unit is used for preprocessing the pulse wave signals to obtain standard pulse wave signals;
the characteristic extraction unit is used for extracting key frequency domain characteristics of the standard pulse wave signals through a pre-constructed frequency domain characteristic processing sub-model of the blood pressure identification model and extracting key time domain characteristics of the standard pulse wave signals through a time domain characteristic processing sub-model of the blood pressure identification model;
the feature fusion unit is used for fusing the key time domain features and the key frequency domain features by applying a feature fusion sub-model of the blood pressure identification model to obtain fusion features of the standard pulse wave signals;
and the recognition unit is used for applying the output sub-model of the blood pressure recognition model to obtain the blood pressure recognition result of the standard pulse wave signal based on the fusion characteristic.
7. The apparatus of claim 6, wherein the pre-processing unit comprises:
and the normalization processing subunit is used for applying a pre-stored normalization parameter to perform normalization processing on the pulse wave signals to obtain standard pulse wave signals, wherein the normalization parameter comprises a standard pulse wave signal average value and a standard pulse wave signal standard deviation.
8. The apparatus of claim 6, further comprising:
the training unit is used for acquiring a training data set, wherein the training data set comprises a plurality of training pulse wave signals and a blood pressure value label of each training pulse wave signal; and training the blood pressure recognition model by using the training data set.
9. The apparatus of claim 6, wherein the frequency domain feature processing submodel comprises a convolutional neural network and a first attention module;
the feature extraction unit includes:
a first feature extraction subunit, configured to extract, through the convolutional neural network, a frequency domain feature of the standard pulse wave signal;
and the first feature screening unit is used for screening the frequency domain features through the first attention module to obtain key frequency domain features of the standard pulse wave signals.
10. The apparatus of claim 6, wherein the temporal processing submodel comprises a long short term memory network and a second attention module;
the feature extraction unit includes:
the second characteristic extraction subunit is used for extracting the time domain characteristics of the standard pulse wave signals through the long-term and short-term memory network;
and the second feature screening subunit is used for screening the time domain features through the second attention module to obtain key time domain features of the standard pulse wave signals.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110400929.3A CN113080907B (en) | 2021-04-14 | 2021-04-14 | Pulse wave signal processing method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110400929.3A CN113080907B (en) | 2021-04-14 | 2021-04-14 | Pulse wave signal processing method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113080907A true CN113080907A (en) | 2021-07-09 |
| CN113080907B CN113080907B (en) | 2022-10-25 |
Family
ID=76677357
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110400929.3A Active CN113080907B (en) | 2021-04-14 | 2021-04-14 | Pulse wave signal processing method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113080907B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114098691A (en) * | 2022-01-26 | 2022-03-01 | 之江实验室 | Pulse wave identity authentication method, device and medium based on Gaussian mixture model |
| CN114259255A (en) * | 2021-12-06 | 2022-04-01 | 深圳信息职业技术学院 | Modal fusion fetal heart rate classification method based on frequency domain signals and time domain signals |
| CN114947782A (en) * | 2022-06-09 | 2022-08-30 | 重庆理工大学 | System and method for central arterial pressure waveform reconstruction based on PPG signal |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180228444A1 (en) * | 2017-02-16 | 2018-08-16 | Tata Consultancy Services Limited | Method and system for detection of coronary artery disease in a person using a fusion approach |
| CN108764084A (en) * | 2018-05-17 | 2018-11-06 | 西安电子科技大学 | Video classification methods based on spatial domain sorter network and the time domain network integration |
| CN109758160A (en) * | 2019-01-11 | 2019-05-17 | 南京邮电大学 | A kind of Woundless blood sugar prediction technique based on LSTM-RNN model |
| CN111248879A (en) * | 2020-02-20 | 2020-06-09 | 电子科技大学 | Hypertension old people activity analysis method based on multi-mode attention fusion |
| CN111493855A (en) * | 2020-04-21 | 2020-08-07 | 重庆理工大学 | System and method for non-invasive measurement of individualized cardiac output |
| CN111950444A (en) * | 2020-08-10 | 2020-11-17 | 北京师范大学珠海分校 | Video behavior identification method based on temporal-spatial feature fusion deep learning network |
| CN112016473A (en) * | 2020-08-31 | 2020-12-01 | 国网安徽省电力有限公司电力科学研究院 | Power distribution network high-resistance grounding fault diagnosis method based on semi-supervised learning and attention mechanism |
| CN112274127A (en) * | 2020-10-28 | 2021-01-29 | 河北工业大学 | Noninvasive continuous blood pressure detection method and device based on one-way pulse wave |
| CN112381011A (en) * | 2020-11-18 | 2021-02-19 | 中国科学院自动化研究所 | Non-contact heart rate measurement method, system and device based on face image |
| CN112587153A (en) * | 2020-12-08 | 2021-04-02 | 合肥工业大学 | End-to-end non-contact atrial fibrillation automatic detection system and method based on vPPG signal |
-
2021
- 2021-04-14 CN CN202110400929.3A patent/CN113080907B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180228444A1 (en) * | 2017-02-16 | 2018-08-16 | Tata Consultancy Services Limited | Method and system for detection of coronary artery disease in a person using a fusion approach |
| CN108764084A (en) * | 2018-05-17 | 2018-11-06 | 西安电子科技大学 | Video classification methods based on spatial domain sorter network and the time domain network integration |
| CN109758160A (en) * | 2019-01-11 | 2019-05-17 | 南京邮电大学 | A kind of Woundless blood sugar prediction technique based on LSTM-RNN model |
| CN111248879A (en) * | 2020-02-20 | 2020-06-09 | 电子科技大学 | Hypertension old people activity analysis method based on multi-mode attention fusion |
| CN111493855A (en) * | 2020-04-21 | 2020-08-07 | 重庆理工大学 | System and method for non-invasive measurement of individualized cardiac output |
| CN111950444A (en) * | 2020-08-10 | 2020-11-17 | 北京师范大学珠海分校 | Video behavior identification method based on temporal-spatial feature fusion deep learning network |
| CN112016473A (en) * | 2020-08-31 | 2020-12-01 | 国网安徽省电力有限公司电力科学研究院 | Power distribution network high-resistance grounding fault diagnosis method based on semi-supervised learning and attention mechanism |
| CN112274127A (en) * | 2020-10-28 | 2021-01-29 | 河北工业大学 | Noninvasive continuous blood pressure detection method and device based on one-way pulse wave |
| CN112381011A (en) * | 2020-11-18 | 2021-02-19 | 中国科学院自动化研究所 | Non-contact heart rate measurement method, system and device based on face image |
| CN112587153A (en) * | 2020-12-08 | 2021-04-02 | 合肥工业大学 | End-to-end non-contact atrial fibrillation automatic detection system and method based on vPPG signal |
Non-Patent Citations (2)
| Title |
|---|
| 朱平芳: "《统计学理论前沿》", 31 July 2016 * |
| 王骞: "脉搏波时频域的连续血压估计研究", 《中国优秀硕士学位论文全文数据库（电子期刊）》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114259255A (en) * | 2021-12-06 | 2022-04-01 | 深圳信息职业技术学院 | Modal fusion fetal heart rate classification method based on frequency domain signals and time domain signals |
| CN114259255B (en) * | 2021-12-06 | 2023-12-08 | 深圳信息职业技术学院 | Modal fusion fetal heart rate classification method based on frequency domain signals and time domain signals |
| CN114098691A (en) * | 2022-01-26 | 2022-03-01 | 之江实验室 | Pulse wave identity authentication method, device and medium based on Gaussian mixture model |
| CN114947782A (en) * | 2022-06-09 | 2022-08-30 | 重庆理工大学 | System and method for central arterial pressure waveform reconstruction based on PPG signal |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113080907B (en) | 2022-10-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110021439B (en) | Medical data classification method and device based on machine learning and computer equipment | |
| CN110377911B (en) | Method and device for identifying intention under dialog framework | |
| CN113080907B (en) | Pulse wave signal processing method and device | |
| CN113724695B (en) | Electronic medical record generation method, device, equipment and medium based on artificial intelligence | |
| EP3136297A1 (en) | System and method for determining information and outliers from sensor data | |
| CN113204969A (en) | Medical named entity recognition model generation method and device and computer equipment | |
| KR102198273B1 (en) | Machine learning based voice data analysis method, device and program | |
| CN117115581A (en) | Intelligent misoperation early warning method and system based on multi-mode deep learning | |
| CN111881722B (en) | Cross-age face recognition method, system, device and storage medium | |
| CN111145903B (en) | Method and device for acquiring vertigo inquiry text, electronic equipment and inquiry system | |
| CN113724819B (en) | Training method, device, equipment and medium for medical named entity recognition model | |
| CN111292851A (en) | Data classification method and device, computer equipment and storage medium | |
| CN117198468B (en) | Intervention scheme intelligent management system based on behavior recognition and data analysis | |
| CN113420122A (en) | Method, device and equipment for analyzing text and storage medium | |
| CN111180025A (en) | Method and device for representing medical record text vector and inquiry system | |
| CN114218427A (en) | Voice quality inspection analysis method, device, equipment and medium | |
| CN111797194B (en) | Text risk detection method and device, electronic equipment and storage medium | |
| CN116663536B (en) | Matching method and device for clinical diagnosis standard words | |
| CN117768022A (en) | Training method of optical fiber detection model, optical fiber detection method and related equipment | |
| CN111259698A (en) | Method and device for acquiring image | |
| CN114495224A (en) | Credit risk determination method and device | |
| CN115687910A (en) | Data processing method and device, computer equipment and readable storage medium | |
| Shen et al. | Multi-modal Human Behaviour Graph Representation Learning for Automatic Depression Assessment | |
| KR102693059B1 (en) | Method for disease management using voice data and apparatus for performing the method | |
| Jegan et al. | Optimized early fusion of handcrafted and deep learning descriptors for voice pathology detection and classification |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |