A blood pressure detection device

Abstract

The present application provides a kind of blood pressure detection device, the device includes: sequentially connected data acquisition module, signal processing module and blood pressure estimation module;Data acquisition module is used to obtain the photoelectric plethysmogram signal of the person to be measured, and photoelectric plethysmogram signal is transmitted to signal processing module;Signal processing module is used to carry out data processing to photoelectric plethysmogram signal, obtains multi-order layer pulse wave signal, and multi-order layer pulse wave signal is transmitted to blood pressure estimation module;Blood pressure estimation module is used to estimate the blood pressure of the person to be measured based on the correlation between the pulse wave signal of each order in multi-order layer pulse wave signal, realizes the rapid and accurate blood pressure detection, while guaranteeing the continuity of blood pressure detection;In addition, blood pressure detection device can be deployed on small mobile device, guarantee the comfort and convenience of blood pressure detection, improve the frequency of blood pressure detection and the applicability of mobile device.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical equipment, in particular to a blood pressure detection device. BACKGROUND

[0002] Blood pressure, as an important parameter for measuring cardiovascular health, the monitoring of the parameter and the design of the detection device are of great significance for the prevention of cardiovascular diseases.

[0003] Currently, there are two main detection methods for continuous blood pressure detection. One is the pressure sensor-based method, which is accurate but has poor comfort and sustainability. In addition, the device using this method is usually large in size, difficult to be used daily, and inconvenient to carry. The second is the photoelectric sensor-based method, which has certain advantages over the former, but still has a series of problems, such as dependence on photoelectric sensor data, numerous data requirements, complex calculation process, and certain errors in detection results.

[0004] Therefore, the development of a wearable blood pressure detection device that is small in size and can quickly and accurately detect continuous blood pressure has become an urgent problem to be solved. SUMMARY

[0005] The present application provides a blood pressure detection device to solve the defects of tedious blood pressure detection, poor comfort, and difficulty in continuous detection in the prior art, achieving rapid and accurate blood pressure detection, improving the comfort and convenience of blood pressure detection, and ensuring the continuity of blood pressure detection.

[0006] The present application provides a blood pressure detection device, characterized in that it comprises a data acquisition module, a signal processing module, and a blood pressure estimation module connected in sequence.

[0007] The data acquisition module is used to acquire the photoelectric plethysmogram signal of the person to be tested and transmit the photoelectric plethysmogram signal to the signal processing module.

[0008] The signal processing module is used to process the photoelectric plethysmogram signal to obtain a multi-layer pulse wave signal and transmit the multi-layer pulse wave signal to the blood pressure estimation module.

[0009] The blood pressure estimation module is used to estimate the blood pressure of the person to be tested based on the correlation between the pulse wave signals of each layer in the multi-layer pulse wave signal.

[0010] The blood pressure detection device provided by the application comprises a blood pressure estimation module, a data collection module, a signal processing module and a blood pressure estimation model.

[0011] The initial blood pressure estimation model is a large convolution kernel neural network obtained by replacing a convolution kernel of a second size in a convolution neural network with a convolution kernel of a first size, and is constructed by a regressor.

[0012] The blood pressure detection device provided by the application further comprises a signal evaluation module, and the data collection module and the signal processing module are connected through the signal evaluation module.

[0013] The data collection module is configured to transmit the photoelectric plethysmogram to the signal evaluation module.

[0014] The signal evaluation module is configured to evaluate the quality of the photoelectric plethysmogram, determine effective plethysmogram based on a quality evaluation result obtained by quality evaluation, and transmit the effective plethysmogram to the signal processing module.

[0015] The signal processing module is configured to perform data processing on the effective plethysmogram to obtain a multi-order plethysmogram.

[0016] The blood pressure detection device provided by the application further comprises a signal correction module, and the signal correction module is connected with the signal evaluation module and the signal processing module.

[0017] The signal evaluation module is further configured to determine ineffective plethysmogram based on the quality evaluation result, and transmit the ineffective plethysmogram to the signal correction module.

[0018] The signal correction module is configured to perform quality correction on the ineffective plethysmogram to obtain recovered plethysmogram, and transmit the recovered plethysmogram to the signal processing module.

[0019] The signal processing module is specifically configured to perform data processing on the effective plethysmogram and the recovered plethysmogram to obtain a multi-order plethysmogram.

[0020] According to the blood pressure detection device provided by the application, the signal evaluation module is specifically used for performing window processing on the photoelectric plethysmogram signal, performing quality evaluation on each windowed plethysmogram signal obtained through the window processing, obtaining a signal autocorrelation coefficient corresponding to each windowed plethysmogram signal, and determining a quality evaluation result based on the signal autocorrelation coefficient and a signal coefficient threshold.

[0021] According to the blood pressure detection device provided by the application, the signal evaluation module is specifically used for determining the any windowed plethysmogram signal as an effective plethysmogram signal when the signal autocorrelation coefficient corresponding to the any windowed plethysmogram signal is greater than or equal to the signal coefficient threshold.

[0022] The signal evaluation module is specifically used for determining the any windowed plethysmogram signal as an invalid plethysmogram signal when the signal autocorrelation coefficient corresponding to the any windowed plethysmogram signal is less than the signal coefficient threshold.

[0023] The signal coefficient threshold is determined based on periodicity of a normal human heart rate range.

[0024] According to the blood pressure detection device provided by the application, the signal autocorrelation coefficient corresponding to the any windowed plethysmogram signal is determined based on the following formula:

[0025]

[0026]

[0027]

[0028]

[0029] In the formula, X i represents an i-th signal value of the any windowed plethysmogram signal X, n is the length of the any windowed plethysmogram signal X, represents the mean value of the any windowed plethysmogram signal X, A i represents an i-th signal value of a lag front segment of the any windowed plethysmogram signal X, B i represents an i-th signal value of a lag rear segment of the any windowed plethysmogram signal X, and ACF represents the signal autocorrelation coefficient of the any windowed plethysmogram signal X.

[0030] According to the blood pressure detection device provided by the application, the signal correction module is specifically used for adding white noise to the invalid plethysmogram signal, performing quality correction on the plethysmogram signal after the white noise addition by using a sequence-to-sequence model, and obtaining a recovered plethysmogram signal.

[0031] The signal processing module is specifically used for high-frequency filtering on the photoelectric plethysmogram signal to obtain a first pulse wave signal, and performing data processing on the first pulse wave signal to obtain a multi-layer pulse wave signal.

[0032] The signal processing module is specifically used for wavelet filtering on the first pulse wave signal to obtain a second pulse wave signal, and performing data processing on the second pulse wave signal to obtain a multi-layer pulse wave signal, wherein the multi-layer pulse wave signal comprises a first-layer conductance velocity pulse wave signal, a second-layer conductance acceleration pulse wave signal, and the photoelectric plethysmogram signal.

[0033] The blood pressure detection device provided by the application can obtain a multi-layer pulse wave signal through data processing on the basis of a photoelectric plethysmogram signal, and estimate blood pressure according to the correlation between pulse wave signals of different layers, so that rapid and accurate blood pressure detection is realized, and the continuity of blood pressure detection is ensured. In addition, the blood pressure detection device can be arranged on a small mobile device, such as an electronic watch, a smart bracelet or the like, so that the comfort and convenience of blood pressure detection are ensured, and the frequency of blood pressure detection and the applicability of the mobile device are improved. BRIEF DESCRIPTION OF DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0035] Figure 1 Fig. 1 is a structural schematic diagram of the blood pressure detection device provided by the application;

[0036] Figure 2 Fig. 2 is a structural schematic diagram of the blood pressure estimation model provided by the application;

[0037] Figure 3 Fig. 3 is a general structural diagram of the blood pressure detection device provided by the application.

[0038] Reference signs:

[0039] 110: data acquisition module; 120: signal processing module; 130: blood pressure estimation module; 140: signal evaluation module; 150: signal correction module; 160: storage module. DETAILED DESCRIPTION

[0040] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be clearly and completely described below in combination with the drawings in the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the protection scope of the present application.

[0041] At present, there are mainly two types of continuous blood pressure detection methods, which are a pressure sensor-based method and a photoelectric sensor-based method. Among them, the continuous blood pressure detection method based on the pressure sensor has high accuracy, but the comfort is poor because the pressure sensor needs to be implanted or pressed, and it may cause trauma or congestion to the patient. In addition, due to the poor wearing comfort, the user needs to keep relatively still, so the sustainability of the detection is poor, and the continuous detection can only be completed in a short time. In addition, the device using this method is often large in size, difficult to be used daily, and inconvenient to carry.

[0042] The continuous blood pressure detection method based on the photoelectric sensor has the advantages of miniaturization of related devices and good wearing comfort, but it still has a series of problems that cannot be ignored, for example, in the method based on pulse transit time (PTT), the assumption is that the elasticity of the blood vessel is constant, but the elasticity of the blood vessel will change under different physiological conditions of the human body, so the estimated blood pressure value will have certain errors. For example, in the pulse wave feature extraction method, it has dependence on a single photoelectric plethysmography sensor data, and the deep learning model applied in the method requires a large amount of data and a complex calculation process. All of these make the research and development of wearable blood pressure detection devices that are small in size and can quickly and accurately detect continuous blood pressure face great challenges.

[0043] To this end, the present application provides a blood pressure detection device, which is based on the photoelectric plethysmography pulse wave signal, obtains multi-layer pulse wave signals through data processing, and estimates blood pressure according to the correlation between the pulse wave signals of each layer, realizes fast and accurate blood pressure detection, and ensures the continuity of blood pressure detection. In addition, the blood pressure detection device in the present application can be deployed on a small mobile device, such as an electronic watch, a smart bracelet, etc., so as to ensure the convenience of blood pressure detection and improve the frequency of blood pressure detection and the applicability of the mobile device.

[0044] Figure 1 The present application provides a blood pressure detection device, which is based on the photoelectric plethysmography pulse wave signal, obtains multi-layer pulse wave signals through data processing, and estimates blood pressure according to the correlation between the pulse wave signals of each layer, realizes fast and accurate blood pressure detection, and ensures the continuity of blood pressure detection. In addition, the blood pressure detection device in the present application can be deployed on a small mobile device, such as an electronic watch, a smart bracelet, etc., so as to ensure the convenience of blood pressure detection and improve the frequency of blood pressure detection and the applicability of the mobile device. Figure 1 As shown in FIG. 1, the device comprises a data acquisition module 110, a signal processing module 120, and a blood pressure estimation module 130 connected in sequence.

[0045] The data acquisition module 110 is configured to acquire the photoplethysmogram signal of the person to be measured and transmit the photoplethysmogram signal to the signal processing module 120.

[0046] The signal processing module 120 is configured to perform data processing on the photoplethysmogram signal to obtain a multi-order layer pulse wave signal and transmit the multi-order layer pulse wave signal to the blood pressure estimation module 130.

[0047] The blood pressure estimation module 130 is configured to estimate the blood pressure of the person to be measured based on the correlation between the pulse wave signals of each order layer in the multi-order layer pulse wave signal.

[0048] Specifically, in the embodiment of the present application, the blood pressure detection device comprises the data acquisition module 110, the signal processing module 120, and the blood pressure estimation module 130, and the three are sequentially connected, that is, the data acquisition module 110 is connected with the signal processing module 120, and the signal processing module 120 is connected with the blood pressure estimation module 130.

[0049] The data acquisition module 110 is constructed on the basis of the photoelectric module, which can acquire the photoplethysmogram signal collected by the photoelectric module. The photoelectric module is composed of a photoelectric sensing device and a light device, wherein the photoelectric sensing device can be a photoelectric sensor, and the light device is a device capable of emitting green light, such as an LED lamp, a green laser lamp, etc.

[0050] Since the photoelectric module uses the photoplethysmographic (PPG) method when collecting the photoplethysmogram signal of the person to be measured, the photoplethysmogram signal collected can also be expressed as the PPG signal of the person to be measured.

[0051] After the data acquisition module 110 acquires the PPG signal of the person to be measured, it can transmit the PPG signal to the connected signal processing module 120. The signal processing module 120 can receive the PPG signal of the person to be measured transmitted by the data acquisition module 110 and perform data processing on the PPG signal to meet the requirements of the input data in the subsequent blood pressure estimation process, thereby obtaining a multi-order layer pulse wave signal.

[0052] Specifically, considering that there is a certain noise component in the PPG signal, and due to device / heating, arm shaking, breathing, etc. during the acquisition process, it may cause the signal to produce baseline drift, in addition, in view of the less direct information contained in the obtained one-dimensional single-channel PPG signal, in order to enrich the information and ensure the effectiveness of the signal, the signal processing module 120 can perform data processing on the PPG signal after receiving the PPG signal of the to-be-measured person. The data processing here can be filtering and denoising, signal augmentation, signal value normalization, etc. Finally, a multi-order pulse wave signal can be obtained.

[0053] Here, the multi-order pulse wave signal includes a plurality of order pulse wave signals obtained by signal augmentation, for example, a first-order pulse wave signal of the derivative velocity, a second-order pulse wave signal of the derivative acceleration, a photoelectric volume pulse wave signal, etc.

[0054] After that, the signal processing module 120 can transmit the multi-order pulse wave signal to the blood pressure estimation module 130, and the blood pressure estimation module 130 can receive the multi-order pulse wave signal of the to-be-measured person, and can use the correlation between the pulse wave signals of each order in the multi-order pulse wave signal to estimate the blood pressure, thereby obtaining the blood pressure detection result of the to-be-measured person. The blood pressure detection result here can be the systolic pressure and diastolic pressure of the to-be-measured person.

[0055] Specifically, the process of blood pressure estimation module 130 estimating the blood pressure of the to-be-measured person can be realized by a blood pressure estimation model. Here, the blood pressure estimation module 130 can use the pulse wave signals of different orders in the multi-order pulse wave signal as the inputs of different channels in the blood pressure estimation model after receiving the multi-order pulse wave signal, and use the channel attention mechanism to estimate the blood pressure by using the correlation between the pulse wave signals of different channels, thereby obtaining the systolic pressure and diastolic pressure of the to-be-measured person.

[0056] In the embodiment of the present application, the pulse wave signals of each order in the multi-order pulse wave signal are used as the input of the blood pressure estimation model in the form of inter-channel combination, and the blood pressure is evaluated by the blood pressure estimation model, so that the systolic pressure and diastolic pressure can be output by the model. In this process, the use of inter-channel attention can greatly improve the detection accuracy of the model for blood pressure estimation, so that the accuracy and reliability of the blood pressure detection result output by the model are higher, and the effectiveness is stronger.

[0057] The blood pressure detection device provided by the application is based on the photoelectric volume pulse wave signal, obtains multi-layer pulse wave signals through data processing, and estimates blood pressure according to the correlation between the pulse wave signals of each layer, realizes rapid and accurate blood pressure detection, and guarantees the continuity of blood pressure detection. In addition, the blood pressure detection device in the application can be deployed on a small mobile device, such as an electronic watch, a smart bracelet, etc., so as to guarantee the comfort and convenience of blood pressure detection, and improve the frequency of blood pressure detection and the applicability of the mobile device.

[0058] Based on the above embodiment, the blood pressure estimation module 130 is specifically configured to apply a blood pressure estimation model to estimate the blood pressure of the to-be-measured person based on the correlation between the pulse wave signals of each layer in the multi-layer pulse wave signals; the blood pressure estimation model is obtained by applying the sample multi-layer pulse wave signals of the sample person to the initial blood pressure estimation model for model fine-tuning.

[0059] The initial blood pressure estimation model is constructed based on a large convolution kernel neural network obtained by replacing a convolution kernel of a second size in a convolutional neural network with a convolution kernel of a first size, and a regressor, and the first size is larger than the second size.

[0060] Specifically, considering that in the process of blood pressure detection in the traditional scheme, the numerous requirements of the deep learning model for the amount of data and the complexity of the calculation process result in that the blood pressure detection is very cumbersome, and the detection result also has deviation, in the embodiment of the application, the blood pressure estimation module 130 designs a convolutional neural network based on light-weight convolution when performing blood pressure estimation, and builds a blood pressure estimation model based on the convolutional neural network, and performs blood pressure estimation based on the model, so as to reduce the amount of model parameters while maintaining the accuracy, realize the light-weight of the model and the simplification of the blood pressure detection process, guarantee the detection accuracy, and optimize the detection efficiency.

[0061] Here, in the convolutional neural network based on light-weight convolution, a larger size convolution kernel is introduced to replace the small size convolution kernel in the traditional convolutional neural network, so as to obtain a more effective receptive field while reducing the number of network layers, thereby guaranteeing the detection accuracy of the network. In short, it is obtained by replacing a convolution kernel of a second size in a convolutional neural network with a convolution kernel of a first size, and the first size must be larger than the second size, and the convolution kernels of the convolution layers in the convolutional neural network obtained in this way are all large size, so it can also be called a large convolution kernel neural network. The large convolution kernel neural network is used as a pre-training model, and a regressor is connected based on the pre-training model to construct a blood pressure estimation model.

[0062] Specifically, after receiving the multi-layered pulse wave signal transmitted by the signal processing module 120, the blood pressure estimation module 130 can apply a blood pressure estimation model to perform blood pressure estimation to obtain the systolic pressure and diastolic pressure of the person to be measured, that is, the pulse wave signals of each layer in the multi-layered pulse wave signal can be combined between channels as the input of the blood pressure estimation model, in which a channel attention mechanism is adopted to utilize the correlation between the input signals of different channels to estimate the blood pressure of the person to be measured, thereby obtaining the systolic pressure and diastolic pressure.

[0063] Before inputting the multi-layered pulse wave signal into the blood pressure estimation model, the blood pressure estimation model needs to be pre-trained, so the blood pressure estimation model constructed by the large convolution kernel neural network and the regressor described above can be referred to as an initial blood pressure estimation model, and the model fine-tuning can be performed on the model by using the sample multi-layered pulse wave signal of the sample person, thereby obtaining the trained blood pressure estimation model.

[0064] In the embodiment of the present application, when performing model training, the method based on pre-training model and model fine-tuning, and personalized BN (BatchNorm, batch normalization) layer statistical parameters is adopted. After pre-training using data to obtain a pre-training model, the BN layer parameters are replaced with the statistical parameters of the sample person modeled on the basis of the sample multi-layered pulse wave signal of the sample person, to accelerate the convergence of the model. At the same time, the parameters of the convolution layer in the initial blood pressure estimation model are frozen, and only the regressor is fine-tuned, so that the initial blood pressure estimation model can be generalized by a small amount of sample data and extended to a large population domain (such as the population domain not participating in model pre-training), thereby realizing blood pressure estimation in a larger range and with higher accuracy, and finally obtaining the trained blood pressure estimation model.

[0065] It can be understood that, when performing model training, the SGD (Stochastic Gradient Descent) is used as the optimizer of the model training, the MSE Loss (Mean Squared Loss) is used as the loss function, the learning rate is set to 0.001, and the Mini-batch method is used to optimize the model, and the size of each batch is 4096.

[0066] In the embodiment of the present application, when constructing the model, the larger size convolution kernel is used to replace the traditional small size convolution kernel, which can reduce the number of model layers while obtaining a more effective receptive field matching the cardiac cycle, thereby enabling higher estimation accuracy in the specified frequency and heartbeat cycle. In addition, the structure design based on the residual network can reduce the complexity of the model while preventing gradient disappearance.

[0067] The method provided by the embodiment of the application can perform blood pressure estimation by using the lightweight blood pressure estimation model determined based on the large convolution kernel neural network, can reduce the model parameter quantity for convenient deployment, does not lose accuracy, and guarantees the accuracy and reliability of blood pressure detection. In addition, the model training is personalized modeling based on a small amount of sample data, can greatly improve the generalization ability of the model across populations, and provides a key help for blood pressure detection with high accuracy in a larger range.

[0068] Based on the above embodiment, Figure 2 is a structural schematic diagram of the blood pressure estimation model provided by the application, as Figure 2 shown, the two-channel PPG signals in the blood pressure estimation model pass through a convolution layer with a convolution kernel of 3x1 and a step of 1, and can become 128-channel data, and then can pass through a BN layer and a ReLU function (activation layer) in sequence.

[0069] The 128-channel data can pass through an expanded convolution block with a convolution kernel of 15x1 and an expansion rate of 2, wherein the data is divided into four groups by using grouped convolution (the number of groups represents the grouped convolution), and then combined by using concatenate, so that 128-channel data is obtained, and then a BN layer and a ReLU function can be used in sequence.

[0070] The 128-channel data can pass through an expanded convolution block with a convolution kernel of 15x1 and an expansion rate of 2, wherein the data is divided into four groups by using grouped convolution, and then combined by using concatenate, so that 128-channel data is obtained, and then a BN layer and a ReLU function can be used.

[0071] The 128-channel data can pass through an expanded convolution block with a convolution kernel of 15x1 and an expansion rate of 2, wherein the data is divided into four groups by using grouped convolution, and then combined by using concatenate, so that 128-channel data is obtained, and then a BN layer and a ReLU function can be used.

[0072] The 128-channel data can pass through a convolution with a convolution kernel of 3x1 and a step of 1, wherein the data is divided into four groups by using grouped convolution, and then combined by using concatenate, so that 128-channel data is obtained, and then a BN layer and a ReLU function can be used.

[0073] The 128-channel data can pass through a convolution with a convolution kernel of 3x1 and a step of 1, wherein the data is divided into four groups by using grouped convolution, and then combined by using concatenate, so that 128-channel data is obtained, and then a BN layer and a ReLU function can be used.

[0074] The 128-channel data can be converted into 128-channel data by using a convolution kernel of 3x1 and a step length of 1, using grouped convolution, being divided into four groups, and using the concatenate mode to combine, and then a BN layer and a ReLU function can be used.

[0075] The 128-channel data can be converted into 128-channel data by using a convolution kernel of 3x1 and a step length of 1, using grouped convolution, being divided into four groups, and using the concatenate mode to combine, and then a BN layer and a ReLU function can be used.

[0076] After that, the channel attention mechanism can be used to optimize the n x 1 x 128 data into 1 x 1 x 128 data by using average pooling, convolution, and a sigmoid function.

[0077] Finally, a regressor composed of two fully connected layers is used to output the regression, and the fully connected layers are 128 to 64 to 2, and the final result is the systolic pressure and diastolic pressure.

[0078] Based on the above embodiment, the parameters of the BN layer can be calculated by using the Mini-batch method, and the specific calculation formula can be represented as:

[0079]

[0080]

[0081] The calculation formula of the output y of the BN layer is: i

[0082]

[0083] Each batch has m samples, and x i represents the i-th sample, and the size of each sample is (C, T, 1), where C is the channel number, T is the time length, mu is the mean, sigma is the variance, gamma and beta are the learning parameters of the model, and epsilon is a very small value to prevent the denominator from being zero. In the test stage, the BN statistical parameters are determined by the BN parameters of multiple batches in the training stage.

[0084] Based on the above embodiment, the device further comprises a signal evaluation module 140, and the data acquisition module 110 and the signal processing module 120 are connected through the signal evaluation module 140;

[0085] The data acquisition module 110 is used to transmit the photoplethysmogram signal to the signal evaluation module 140.

[0086] ​The signal evaluation module 140 is configured to evaluate the quality of the photoplethysmogram signal, and determine an effective pulse wave signal based on the quality evaluation result, and transmit the effective pulse wave signal to the signal processing module 120.

[0087] The signal processing module 120 is configured to perform data processing on the effective pulse wave signal to obtain a multi-layer pulse wave signal.

[0088] Specifically, in the blood pressure detection device, in addition to the data acquisition module 110, the signal processing module 120, and the blood pressure estimation module 130, the signal evaluation module 140 for quality evaluation is further included, which is connected between the data acquisition module 110 and the signal processing module 120, and connected with both of them, in other words, the data acquisition module 110 and the signal processing module 120 are connected through the signal evaluation module 140. The signal evaluation module 140 can evaluate the quality of the PPG signal acquired by the data acquisition module 110, and can transmit the evaluated effective pulse wave to the signal processing module 120 for corresponding data processing.

[0089] Specifically, considering that there may be some invalid components in the acquired photoplethysmogram signal, that is, some photoplethysmogram signals are unusable due to movement, jitter, etc. during the signal acquisition process, therefore, the data acquisition module 110 can further transmit the acquired photoplethysmogram signal to the signal evaluation module 140 for quality evaluation to determine whether it is usable.

[0090] The signal evaluation module 140 can receive the PPG signal transmitted by the data acquisition module 110, and can evaluate the quality of the PPG signal to obtain a quality evaluation result, and then can screen the effective pulse wave signal from the PPG signal according to the quality evaluation result, that is, can perform quality evaluation and screening on the PPG signal to screen out usable signals, that is, effective pulse wave signals, and eliminate unusable signals, that is, invalid pulse wave signals. Then, the signal quality evaluation module 140 can transmit the effective pulse wave signal to the signal processing module 120 for data processing by the signal processing module 120 to obtain a multi-layer pulse wave signal.

[0091] Based on the above embodiment, the device further includes a signal correction module 150, which is connected with the signal evaluation module 140 and the signal processing module 120, respectively.

[0092] The signal evaluation module 140 is further configured to determine an invalid pulse wave signal based on the quality evaluation result, and transmit the invalid pulse wave signal to the signal correction module 150.

[0093] The signal correction module 150 is configured to correct the quality of the invalid pulse wave signal, and obtain a recovered pulse wave signal, and transmit the recovered pulse wave signal to the signal processing module 120.

[0094] The signal processing module 120 is specifically configured to perform data processing on the valid pulse wave signal and the recovered pulse wave signal, and obtain a multi-order pulse wave signal.

[0095] Specifically, in the above process, after the signal quality evaluation by the signal evaluation module 140, the valid pulse wave signal and the invalid pulse wave signal can be determined from the photoplethysmogram signal, wherein the valid pulse wave signal can be directly transmitted to the signal processing module 120 for data processing, and the invalid pulse wave signal cannot be directly processed due to the failure to meet the signal quality standard. Therefore, for the invalid pulse wave signal, the signal correction module 150 for quality correction is further provided in the embodiment of the present application.

[0096] Here, the signal correction module 150 is arranged between and connected to the signal evaluation module 140 and the signal processing module 120, that is, the two ends of the signal correction module 150 are connected to the signal evaluation module 140 and the signal processing module 120, respectively. It can also be understood that, in addition to being directly connected, the signal evaluation module 140 and the signal processing module 120 can also be connected through the signal correction module 150. The signal correction module 150 can correct the quality of the invalid pulse wave signal transmitted by the signal evaluation module 140, and transmit the corrected pulse wave signal to the signal processing module 120 for corresponding data processing.

[0097] Specifically, considering that the invalid pulse wave signal removed after the quality evaluation by the signal evaluation module 140 cannot be applied in subsequent data processing and blood pressure estimation due to the signal quality problem, after determining the invalid pulse wave signal according to the quality evaluation result, the signal evaluation module 140 can transmit the invalid pulse wave signal to the signal correction module 150 for quality correction by the signal correction module 150 to restore the signal quality.

[0098] The signal correction module 150 can receive the invalid pulse wave signal transmitted by the signal evaluation module 140, and correct the quality of the invalid pulse wave signal, thereby obtaining a corrected pulse wave signal, i.e., a recovered pulse wave signal. That is, the quality of the invalid pulse wave signal can be recovered and corrected to restore the signal quality of the invalid pulse wave signal, thereby obtaining the recovered pulse wave signal. Then, the signal correction module 150 can transmit the recovered pulse wave signal to the signal processing module 120 for data processing by the signal processing module 120.

[0099] Based on the above embodiment, the signal evaluation module 140 is specifically configured to perform windowing processing on the photoplethysmogram signal, and perform quality evaluation on each windowed pulse wave signal obtained by the windowing processing, obtain a signal autocorrelation coefficient corresponding to each windowed pulse wave signal, and determine the quality evaluation result based on the signal autocorrelation coefficient and a signal coefficient threshold.

[0100] Specifically, in the above process, when evaluating the quality of the PPG signal, the signal evaluation module 140 can first perform windowing processing on the PPG signal to obtain a plurality of windowed pulse wave signals, i.e., the PPG signal can be windowed according to a preset window duration, so as to be divided into a plurality of windowed pulse wave signals each having a duration of the preset window duration. The preset window duration is pre-set and can be set according to actual conditions and actual requirements, for example, 4 seconds, 5 seconds, 6 seconds, etc. Preferably, the preset window duration is set to 5 seconds in the embodiment of the present application.

[0101] Then, the quality of each windowed pulse wave signal can be evaluated to obtain a signal autocorrelation coefficient corresponding to each windowed pulse wave signal. Specifically, since the signal quality of each windowed pulse wave signal can be evaluated by its signal autocorrelation coefficient, the signal autocorrelation coefficient of each windowed pulse wave signal can be calculated when the signal quality is evaluated in the embodiment of the present application. The signal autocorrelation coefficient can be obtained by dividing the corresponding windowed pulse wave signal into two segments and measuring the correlation between the two segments.

[0102] Here, the value of the signal autocorrelation coefficient is between 0 and 1. The greater the value, the stronger the correlation between the two segments in the corresponding windowed pulse wave signal, and the better the signal quality. Correspondingly, the smaller the value of the signal autocorrelation coefficient, the closer it is to 0, the weaker the correlation between the two segments in the corresponding windowed pulse wave signal, and the worse the signal quality.

[0103] Then, the quality evaluation result can be determined according to the signal autocorrelation coefficient and a signal coefficient threshold, i.e., the signal quality of each windowed pulse wave signal can be determined by the signal coefficient threshold whether it meets the signal quality requirement. Here, the signal quality requirement can be determined by the signal coefficient threshold. The signal coefficient threshold is pre-set and used to determine whether each windowed pulse wave signal has a periodicity within the normal heart rate range. The signal coefficient threshold can also be set according to actual conditions and actual requirements.

[0104] Further, in the case that any one of the windowed pulse wave signals meets the signal quality requirement, i.e., the windowed pulse wave signal has the periodicity in the normal human heart rate range, it can be determined that the signal quality of the windowed pulse wave signal is excellent, i.e., the quality evaluation result is good, for example, can be very good, good, excellent, etc. Correspondingly, in the case that any one of the windowed pulse wave signals fails to meet the signal quality requirement, i.e., the windowed pulse wave signal does not have the periodicity in the normal human heart rate range, it can be determined that the signal quality of the windowed pulse wave signal is poor, i.e., the quality evaluation result is bad, for example, can be very poor, poor, relatively poor, etc.

[0105] Based on the above embodiment, the signal evaluation module 140 is specifically configured to determine the windowed pulse wave signal as an effective pulse wave signal in the case that the signal autocorrelation coefficient corresponding to any one of the windowed pulse wave signals is greater than or equal to the signal coefficient threshold value;

[0106] The signal evaluation module 140 is specifically configured to determine the windowed pulse wave signal as an invalid pulse wave signal in the case that the signal autocorrelation coefficient corresponding to the windowed pulse wave signal is less than the signal coefficient threshold value;

[0107] The signal coefficient threshold value is determined based on the periodicity in the normal human heart rate range.

[0108] Specifically, in the above process, the signal evaluation module 140 determines the quality evaluation result by comparing the signal autocorrelation coefficient with the signal coefficient threshold value based on the signal autocorrelation coefficient and the signal coefficient threshold value, and determines the effective pulse wave signal and the invalid pulse wave signal according to the quality evaluation result.

[0109] It can be understood that, since the signal coefficient threshold value is determined based on the periodicity in the normal human heart rate range, it is used to determine whether the windowed pulse wave signal has the periodicity in the normal human heart rate range, and thus, in the case that the signal autocorrelation coefficient corresponding to any one of the windowed pulse wave signals is greater than or equal to the signal coefficient threshold value, i.e., the windowed pulse wave signal has the periodicity in the normal human heart rate range, it can be determined that the signal quality of the windowed pulse wave signal is excellent, i.e., the quality evaluation result is good, and thus the windowed pulse wave signal can be determined as an effective pulse wave signal.

[0110] Correspondingly, in the case that the signal autocorrelation coefficient corresponding to any windowed pulse wave signal is less than the signal coefficient threshold, i.e. the windowed pulse wave signal does not have periodicity in the normal human heart rate range, it can be determined that the signal quality of the windowed pulse wave signal is poor, i.e. the quality evaluation result is poor, and thus the windowed pulse wave signal can be determined as an invalid pulse wave signal. It is worth noting that in the embodiment of the present application, various situations that may occur during signal acquisition are considered, and the signal coefficient threshold is set to 0.7 as a signal normalization condition in combination with actual needs.

[0111] Based on the above embodiment, the signal autocorrelation coefficient corresponding to any windowed pulse wave signal is determined based on the following formula:

[0112]

[0113]

[0114]

[0115]

[0116] In the formula, X i represents the i-th signal value of the windowed pulse wave signal X, n is the length of the windowed pulse wave signal X, represents the mean value of the windowed pulse wave signal X, A i is the i-th signal value of the lag front segment of the windowed pulse wave signal X, B i is the i-th signal value of the lag rear segment of the windowed pulse wave signal X, and ACF represents the signal autocorrelation coefficient of the windowed pulse wave signal X.

[0117] Based on the above embodiment, the signal correction module 150 is specifically configured to add white noise to the invalid pulse wave signal, and apply a sequence-to-sequence model to correct the quality of the pulse wave signal after white noise addition to obtain a recovered pulse wave signal.

[0118] Specifically, in the above process, when the signal correction module 150 corrects the quality of the invalid pulse wave signal, it can first add white noise to the received invalid pulse wave signal, and then correct the quality through the model on this basis to obtain the recovered pulse wave signal.

[0119] Specifically, the signal correction module 150 can first add white noise to the invalid pulse wave signal, that is, a self-supervised method can be used to add white noise to the original signal (invalid pulse wave signal) to obtain a pulse wave signal after white noise addition; then, a sequence-to-sequence model, for example, a sequence-to-sequence Transformer model, can be used to recover the poor quality signal to obtain a recovered pulse wave signal, that is, the sequence-to-sequence Transformer model can be applied to correct the quality of the pulse wave signal after white noise addition to recover the signal quality, thereby obtaining the recovered pulse wave signal.

[0120] Based on the above embodiment, the signal processing module 120 is specifically configured to perform high-frequency filtering on the photoplethysmogram pulse wave signal to obtain a first pulse wave signal, and perform data processing on the first pulse wave signal to obtain a multi-layer pulse wave signal.

[0121] Specifically, in the process of data processing the PPG signal to convert it into data that can be received by the blood pressure estimation model in the subsequent blood pressure estimation module 130, thereby obtaining the multi-layer pulse wave signal, the signal processing module 120 can first perform high-frequency filtering on the PPG signal, and then perform data processing on the first pulse wave signal obtained by high-frequency filtering to obtain the multi-layer pulse wave signal.

[0122] Specifically, considering that there is a certain noise component in the PPG signal, therefore, in the embodiment of the application, to maximize the preservation of the original information of the pulse wave signal of the person to be tested, the PPG signal can be high-frequency filtered to filter out high-frequency noise, thereby obtaining a first pulse wave signal. Here, a low-pass filter with a cutoff frequency of 8 Hz can be used to filter out high-frequency noise from the PPG signal to obtain the first pulse wave signal.

[0123] Subsequently, the first pulse wave signal can be processed to obtain a multi-layer pulse wave signal. Here, since the first pulse wave signal is a pulse wave signal after high-frequency noise filtering, when data processing is performed on it, the baseline drift of the signal and the fact that the signal contains less direct information, etc. The data processing here can be wavelet filtering, signal augmentation, signal value normalization, etc. After this processing, the multi-layer pulse wave signal can be obtained.

[0124] Based on the above embodiment, the signal processing module 120 is specifically configured to perform wavelet filtering on the first pulse wave signal to obtain a second pulse wave signal, and perform data processing on the second pulse wave signal to obtain a multi-layer pulse wave signal, the multi-layer pulse wave signal including a first layer of pulse wave signal of the derivative velocity, a second layer of pulse wave signal of the derivative acceleration, and a photoplethysmogram pulse wave signal.

[0125] Specifically, in the above process, when the signal processing module 120 performs data processing on the first pulse wave signal to obtain the multi-order pulse wave signal, the first pulse wave signal can be first wavelet filtered, and then the second pulse wave signal obtained by wavelet filtering is processed to obtain the multi-order pulse wave signal.

[0126] Specifically, considering that the signal may produce baseline drift due to device / device heating, arm shaking, breathing, and other reasons during signal acquisition, therefore, in the embodiment of the application, after obtaining the first pulse wave signal, the baseline drift can be removed. Since the baseline drift is usually nonlinear, a db8 wavelet filter can be used to wavelet filter the first pulse wave signal to remove the baseline drift, thereby obtaining the second pulse wave signal.

[0127] After that, considering that the one-dimensional single-channel PPG signal contains less direct information, in order to enrich the information, the embodiment of the application can generate high-order signals, that is, signal augmentation, to obtain a plurality of order pulse wave signals, including a first-order derivative velocity pulse wave signal (first-order difference), that is, a VPG (velocity plethysmography, first-order derivative velocity photoplethysmography) signal, a second-order derivative acceleration pulse wave signal (second-order difference), that is, an APG (acceleration plethysmography, second-order derivative acceleration photoplethysmography) signal, and a photoplethysmography pulse wave signal, that is, a PPG (photoplethysmographic, photoplethysmography) signal.

[0128] Further, considering the differences of different acquisition devices, mobile devices, population domains, etc., in the embodiment of the application, each order pulse wave signal can be normalized, that is, the Min-Max normalization method can be used to normalize each order pulse wave signal, and finally a multi-order pulse wave signal can be obtained.

[0129] In addition, it is worth noting that in the embodiment of the application, in addition to obtaining the above-mentioned various order pulse wave signals, time domain pulse wave signals can also be obtained through time domain transformation, such as time sequence feature extraction, normalization, etc., and frequency domain pulse wave signals can also be obtained through frequency domain transformation, such as KKT (fast Fourier transform); then blood pressure estimation can be performed according to the above-mentioned various pulse wave signals, which ensures the accuracy and comprehensiveness of blood pressure estimation and improves the accuracy and reliability of the results.

[0130] Based on the above embodiment, the calculation formula of the first-order derivative velocity pulse wave signal is:

[0131]

[0132] wherein x i is the PPG signal value at time t i , x i-1 is the PPG signal value at time t i-1 .

[0133] The calculation formula of the pulse wave signal of the second-order layer derivative acceleration is:

[0134]

[0135] wherein Signal′ i is the first-order difference of signal x at time t i , Signal′ i-1 is the first-order difference of signal x at time t i-1 .

[0136] Further, the process of normalizing the maximum and minimum values of the pulse wave signal of each order by using the Min-Max normalization method can be represented by the following formula:

[0137]

[0138] wherein x is the signal value of the corresponding pulse wave signal, x Min is the minimum value of the corresponding pulse wave signal, and x Max is the maximum value of the corresponding pulse wave signal.

[0139] Based on the above embodiment, the device further comprises a storage module 160, which is connected with the data acquisition module 110 respectively; the data acquisition module 110 is further used for transmitting the photoplethysmogram signal to the storage module 160; and the storage module 160 is used for storing the photoplethysmogram signal.

[0140] Figure 3 is the overall structure diagram of the blood pressure detection device provided by the application, as shown in Figure 3 , the device comprises a data acquisition module 110, a signal processing module 120, a blood pressure estimation module 130, a signal evaluation module 140, a signal correction module 150, and a storage module 160.

[0141] The data acquisition module 110 is used for acquiring the photoplethysmogram signal of the person to be measured, and can transmit the photoplethysmogram signal to the signal evaluation module 140;

[0142] The data acquisition module 110 is further used for transmitting the photoplethysmogram signal to the storage module 160; and the storage module 160 is used for storing the photoplethysmogram signal;

[0143] The signal evaluation module 140 is configured to evaluate the quality of the photoplethysmogram signal, and determine an effective pulse wave signal based on the quality evaluation result, and transmit the effective pulse wave signal to the signal processing module 120.

[0144] The signal evaluation module 140 is further configured to determine an ineffective pulse wave signal based on the quality evaluation result, and transmit the ineffective pulse wave signal to the signal correction module 150.

[0145] The signal correction module 150 is configured to correct the quality of the ineffective pulse wave signal to obtain a recovered pulse wave signal, and transmit the recovered pulse wave signal to the signal processing module 120.

[0146] The signal processing module 120 is configured to process the effective pulse wave signal and the recovered pulse wave signal to obtain a multi-layer pulse wave signal, and transmit the multi-layer pulse wave signal to the blood pressure estimation module 130.

[0147] The blood pressure estimation module 130 is configured to estimate the blood pressure of the subject based on the correlation between the pulse wave signals of the layers in the multi-layer pulse wave signal.

[0148] Further, the signal evaluation module 140 is specifically configured to perform windowing processing on the photoplethysmogram signal, and evaluate the quality of each windowed pulse wave signal to obtain a signal autocorrelation coefficient corresponding to each windowed pulse wave signal, and determine the windowed pulse wave signal as an effective pulse wave signal if the signal autocorrelation coefficient corresponding to the windowed pulse wave signal is greater than or equal to a signal coefficient threshold, or determine the windowed pulse wave signal as an ineffective pulse wave signal if the signal autocorrelation coefficient corresponding to the windowed pulse wave signal is less than the signal coefficient threshold, wherein the signal coefficient threshold is determined based on the periodicity of the normal heart rate range.

[0149] The signal correction module 150 is specifically configured to add white noise to the ineffective pulse wave signal, and apply a sequence-to-sequence model to correct the quality of the pulse wave signal after the white noise addition to obtain a recovered pulse wave signal.

[0150] The signal processing module 120 is specifically configured to perform high-frequency filtering on the photoplethysmogram signal to obtain a first pulse wave signal, perform wavelet filtering on the first pulse wave signal to obtain a second pulse wave signal, and process the second pulse wave signal to obtain a multi-layer pulse wave signal, wherein the multi-layer pulse wave signal includes a first-layer conductance velocity pulse wave signal, a second-layer conductance acceleration pulse wave signal, and the photoplethysmogram signal.

[0151] The blood pressure estimation module 130 is specifically configured to apply a blood pressure estimation model to estimate blood pressure of the to-be-measured person based on the correlation between the pulse wave signals of each layer in the multi-layer pulse wave signal; the blood pressure estimation model is obtained by fine-tuning the initial blood pressure estimation model by using the sample multi-layer pulse wave signal of the sample person; and the initial blood pressure estimation model is constructed based on a large convolution kernel neural network obtained by replacing a convolution kernel of a second size in a convolutional neural network with a convolution kernel of a first size, and a regressor, and the first size is larger than the second size.

[0152] The device provided by the embodiment of the present application realizes rapid and accurate blood pressure detection while ensuring the continuity of blood pressure detection, by obtaining the multi-layer pulse wave signal through data processing based on the photoplethysmogram signal and estimating blood pressure according to the correlation between the pulse wave signals of each layer. In addition, the blood pressure detection device in the present application can be deployed on a small mobile device, such as an electronic watch, a smart bracelet, etc., so as to ensure the comfort and convenience of blood pressure detection and improve the frequency of blood pressure detection and the applicability of the mobile device.

[0153] The device embodiments described above are only schematic, and the units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one place or distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment. Those skilled in the art can understand and implement without creative labor.

[0154] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and necessary general hardware platforms, and of course, can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0155] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A blood pressure detecting apparatus characterized by comprising: The blood pressure detection device comprises a data acquisition module, a signal processing module, and a blood pressure estimation module connected in sequence. The data acquisition module is configured to acquire a photoplethysmogram (PPG) signal of a to-be-tested person and transmit the PPG signal to the signal processing module. The signal processing module is configured to perform data processing on the PPG signal to obtain a multi-order PPG signal and transmit the multi-order PPG signal to the blood pressure estimation module. The blood pressure estimation module is configured to take the PPG signals of different orders in the multi-order PPG signal as inputs of different channels in a blood pressure estimation model, adopt a channel attention mechanism in the blood pressure estimation model, and estimate blood pressure of the to-be-tested person based on the correlation between the PPG signals of different channels to obtain systolic pressure and diastolic pressure of the to-be-tested person.

2. The blood pressure detection device according to claim 1, wherein the blood pressure estimation model is obtained by fine-tuning an initial blood pressure estimation model based on sample multi-order PPG signals of sample persons; and the initial blood pressure estimation model is constructed based on a large-kernel neural network obtained by replacing a second-size convolution kernel in a convolutional neural network with a first-size convolution kernel, and a regressor, wherein the first size is larger than the second size. The blood pressure detection device further comprises a signal evaluation module, and the data acquisition module and the signal processing module are connected through the signal evaluation module. The data acquisition module is configured to transmit the PPG signal to the signal evaluation module.

3. The blood pressure detection apparatus according to claim 1, characterized by, The signal evaluation module is configured to perform quality evaluation on the PPG signal and determine an effective PPG signal based on a quality evaluation result obtained by the quality evaluation, and transmit the effective PPG signal to the signal processing module. The signal processing module is configured to perform data processing on the effective PPG signal to obtain a multi-order PPG signal. The blood pressure detection device further comprises a signal correction module, and the signal evaluation module and the signal processing module are connected to the signal correction module. The signal evaluation module is further configured to determine an ineffective PPG signal based on the quality evaluation result and transmit the ineffective PPG signal to the signal correction module.

4. The blood pressure detection apparatus according to claim 3, characterized by The signal correction module is configured to perform quality correction on the ineffective PPG signal to obtain a recovered PPG signal and transmit the recovered PPG signal to the signal processing module. The signal processing module is specifically configured to perform data processing on the effective PPG signal and the recovered PPG signal to obtain a multi-order PPG signal.

5. The blood pressure detection device according to claim 4, wherein the signal evaluation module is specifically configured to perform windowing processing on the PPG signal, perform quality evaluation on each windowed PPG signal to obtain a signal autocorrelation coefficient corresponding to each windowed PPG signal, and determine a quality evaluation result based on the signal autocorrelation coefficient and a signal autocorrelation threshold.

6. The blood pressure detection device according to claim 5, wherein ​ ​ ​ The signal evaluation module is specifically configured to determine the any windowed pulse wave signal as a valid pulse wave signal if a signal autocorrelation coefficient corresponding to the any windowed pulse wave signal is greater than or equal to the signal coefficient threshold value. The signal evaluation module is specifically configured to determine the any windowed pulse wave signal as an invalid pulse wave signal if the signal autocorrelation coefficient corresponding to the any windowed pulse wave signal is less than the signal coefficient threshold value. The signal coefficient threshold value is determined based on periodicity of a normal human heart rate range.

7. The blood pressure detection apparatus according to claim 5, characterized by, The signal autocorrelation coefficient corresponding to the any windowed pulse wave signal is determined based on the following formula: ； ； ； ； In the formula, This represents the pulse wave signal of any given window. The Each signal value For any given window of pulse wave signal Length, This represents the pulse wave signal of any given window. The mean, To delay the pulse wave signal of any given window of The first part Each signal value To delay the pulse wave signal of any given window of The second part One signal value, This represents the pulse wave signal of any given window. The signal autocorrelation coefficient.

8. The blood pressure detection apparatus according to any one of claims 4 to 7, characterized by, The signal correction module is specifically configured to add white noise to the invalid pulse wave signal, and apply a sequence-to-sequence model to correct the quality of the pulse wave signal after the white noise addition, to obtain a recovered pulse wave signal.

9. The blood pressure measuring apparatus according to any one of claims 1 to 7, wherein The signal processing module is specifically configured to perform high-frequency filtering on the photoplethysmogram signal to obtain a first pulse wave signal, and perform data processing on the first pulse wave signal to obtain a multi-layer pulse wave signal.

10. The blood pressure detection apparatus according to claim 9, characterized by, The signal processing module is specifically configured to perform wavelet filtering on the first pulse wave signal to obtain a second pulse wave signal, and perform data processing on the second pulse wave signal to obtain a multi-layer pulse wave signal, wherein the multi-layer pulse wave signal includes a first-layer derivative velocity pulse wave signal, a second-layer derivative acceleration pulse wave signal, and the photoplethysmogram signal.

Images