Stress index calculation method based on pulse signals

Abstract

The invention relates to the technical field of biomedical signal processing, in particular to a pulse signal-based stress index calculation method, which comprises the following steps of: acquiring a pulse signal of an individual in real time by utilizing pulse signal acquisition equipment, preprocessing the acquired pulse signal, removing useless information and noise in the pulse signal, and calculating the stress index of the individual according to the useless information and noise. Extracting characteristic parameters related to the stress state from the preprocessed pulse signals, constructing a stress index calculation model based on the extracted characteristic parameters, and accurately mapping the characteristic parameters into stress index values through the stress index calculation model; the pulse signal is collected and processed in real time, the stress index value can be rapidly calculated, and the method has the advantage of being high in real-time performance.

Description

Technical Field

[0001] The invention relates to the technical field of biomedical signal processing, and in particular to a method for calculating a stress index based on a pulse signal. Background Art

[0002] The pulse signal is a physiological signal that reflects the periodic pulsation of the blood vessel wall caused by the periodic contraction and relaxation of the heart. When the heart contracts, the left ventricle ejects blood into the aorta, the pressure in the aorta increases, and the blood propagates to the periphery along the arterial wall in the form of waves. This pressure wave causes the arterial wall to expand and contract periodically, forming a pulse. From a physiological point of view, the rhythmic beating of the heart is the power source for the generation of the pulse signal, and the vascular system is the carrier of the pulse signal. The waveform characteristics and frequency changes of the pulse signal contain rich stress information. However, most of the stress assessment methods based on pulse signals in the existing technology have problems such as complex calculation, poor real-time performance or insufficient accuracy. Summary of the invention

[0003] The purpose of the present invention is to provide a method for calculating a stress index based on a pulse signal to solve the problems raised in the above background technology.

[0004] To achieve the above object, the present invention provides the following technical solution: a method for calculating a stress index based on a pulse signal, the method comprising the following steps: Step 1: signal acquisition, using pulse signal acquisition equipment to collect individual pulse signals in real time; Step 2: Preprocessing: preprocessing the collected pulse signal to remove useless information and noise in the pulse signal; Step 3: Feature extraction, extracting characteristic parameters related to the stress state from the preprocessed pulse signal; Step 4: Stress index calculation: Based on the extracted characteristic parameters, a stress index calculation model is constructed, and the characteristic parameters are accurately mapped to stress index values ​​through the stress index calculation model; Step 5: Result output and evaluation: output the stress index value calculated in step 4 to the user or monitoring system, and evaluate the stress state according to the preset threshold; the user or monitoring system takes intervention measures according to the evaluation results to reduce the individual's stress level.

[0005] Preferably, the preprocessing in step 2 includes filtering, denoising and signal enhancement.

[0006] Preferably, the characteristic parameters in step three include pulse rate variability PRV, waveform morphology changes and pulse wave transmission time PTT.

[0007] Preferably, in step 4, the stress index calculation model is constructed as follows: The extracted feature parameters are normalized according to the formula, where is the original data, and is the minimum and maximum value of the original data value range, and is the normalized data; The normalized characteristic parameter data and the corresponding stress index label data are divided into a training set and a validation set, and the stress index label data is obtained by measuring after the stress state is induced experimentally; and the above data are then sorted into a format suitable for neural network input; A multilayer perceptron is designed, including an input layer, a hidden layer and an output layer. The number of neurons in the input layer is the same as the dimension of the input feature, that is, 3 neurons correspond to the three characteristic parameters of pulse rate variability PRV, waveform morphology change and pulse wave conduction time PTT. The output layer has only 1 neuron, which is used to output the stress index. The number of hidden layers and neurons is determined by experiments. The activation function is selected in the hidden layer and the output layer. The ReLU function is used for the hidden layer, and its expression is: The output layer selects the activation function according to the demand. If the value range of the stress index is any real number, the activation function is not used.

[0008] Preferably, before training the neural network, the connection weights and biases between neurons are initialized. In each training iteration, a sample in the training set is input into the neural network, starting from the input layer, and the signal is propagated forward layer by layer by calculating the weighted sum of the neurons and the output of the activation function until the output layer obtains the predicted stress index.

[0009] Preferably, the loss between the predicted stress index and the true stress index is calculated by the following loss function, which is expressed as follows: where is the number of samples, the value of the loss function reflects the degree of difference between the model prediction result and the true result, and the training goal is to minimize this loss function; the gradient of the weight and bias is calculated according to the loss function, the error is back-propagated from the output layer to the input layer through the back-propagation algorithm, and the weight and bias are updated using the gradient descent algorithm; the number of training rounds is set, that is, the number of times the entire training set is completely traversed, and after each round of training, the validation set is used to evaluate the performance of the model, the samples in the validation set are input into the trained model, the loss and accuracy on the validation set are calculated, and the model is judged whether it is overfitting or underfitting by observing the change in the validation set loss; When the model training is completed, the newly collected pulse signal is processed according to the above-mentioned feature extraction and preprocessing steps, and then input into the model to calculate the stress index.

[0010] Compared with the prior art, the present invention has the following beneficial effects: The present invention can quickly calculate the stress index value by real-time acquisition and processing of pulse signals, and has the characteristics of strong real-time performance. In addition, the accuracy of stress index calculation is improved by extracting characteristic parameters of pulse rate variability PRV, waveform morphology change and pulse wave conduction time PTT. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 The figure is a flow chart of a method for calculating a stress index based on a pulse signal according to the present invention. DETAILED DESCRIPTION

[0012] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0013] See also Figure 1 The present invention provides a technical solution: a method for calculating a stress index based on a pulse signal, the method comprising the following steps: Step 1: Signal acquisition, using pulse signal acquisition equipment to collect individual pulse signals in real time; ensure the stability and accuracy of the signal during the acquisition process, and reduce noise interference. Pulse signal acquisition equipment includes finger-clip pulse detectors, wrist pulse detectors, and pulse modules in multi-parameter monitors, etc., which can be selected according to needs and are devices in the prior art.

[0014] Step 2: Preprocessing: preprocessing the collected pulse signal, including filtering, denoising and signal enhancement, removing useless information and noise in the pulse signal and retaining useful information related to the stress state; Step 3: Feature extraction, extracting characteristic parameters closely related to the stress state from the preprocessed pulse signal; the characteristic parameters include pulse rate variability PRV, waveform morphology changes and pulse wave transmission time PTT, which can reflect the response of the individual's cardiovascular system to stress.

[0015] For the feature extraction of pulse rate variability PRV: the pulse rate is calculated from the collected pulse waveform data, usually by detecting the time interval between adjacent pulse wave peaks (RR interval, similar to the concept of RR interval in electrocardiogram). For example, if the time interval between two adjacent pulse wave peaks is 1 second, the pulse rate is 60 times / minute. The pulse rate sequence is analyzed for variability, and statistical indicators such as standard deviation (SDNN) and root mean square of adjacent pulse rate differences (RMSSD) are calculated. SDNN reflects the overall standard deviation of heart rate changes over a longer period of time, while RMSSD focuses more on reflecting the changes between adjacent heartbeats.

[0016] Feature extraction of waveform morphology changes: Analyze the slope of the ascending and descending branches, as well as the peak and valley values ​​of the pulse waveform. For example, the slope of the ascending branch can reflect the speed of blood ejection when the heart contracts, and the slope of the descending branch is related to factors such as peripheral vascular resistance; calculate the amplitude characteristics of the waveform, such as the main wave height and the dicrotic wave height. The main wave height is mainly related to the heart's contractile force, and the dicrotic wave height is affected by factors such as vascular elasticity.

[0017] Feature extraction of pulse wave transmission time PTT: Determine the starting time point or feature point of the pulse wave at two different locations (such as the heart and a point in the peripheral blood vessels). This can be achieved by collecting pulse signals at two locations at the same time, or by using electrocardiogram (ECG) and PPG signals for joint analysis. For example, PTT is calculated using the peak value of the R wave of the ECG and the starting point of the PPG waveform; the time difference of pulse wave transmission between the two locations is calculated, and PTT can be used as an important indicator to reflect changes in vascular elasticity and blood pressure.

[0018] Step 4: Stress index calculation: Based on the extracted characteristic parameters, a stress index calculation model is constructed, and the characteristic parameters are accurately mapped to stress index values ​​through the stress index calculation model; The construction of the stress index calculation model is as follows: The extracted feature parameters are normalized according to the formula, where is the original data, and is the minimum and maximum value of the original data value range, and is the normalized data; The normalized feature parameter data and the corresponding stress index label data are divided into a training set and a validation set. The stress index label data is obtained by measuring after experimentally inducing a stress state. The above data are then organized into a format suitable for neural network input. For example, if there are samples, each sample has three feature parameters, pulse rate variability, waveform morphology change, and pulse wave conduction time. The data can be organized into a matrix of shape, where each row represents a feature vector of a sample. The stress index label data is organized into a vector of length.

[0019] A multilayer perceptron is designed, including an input layer, a hidden layer and an output layer. The number of neurons in the input layer is the same as the dimension of the input feature, that is, 3 neurons correspond to the three characteristic parameters of pulse rate variability PRV, waveform morphology change and pulse wave conduction time PTT. The output layer has only 1 neuron, which is used to output the stress index. The number of hidden layers and neurons is determined by experiments. The activation function is selected in the hidden layer and the output layer. The ReLU function is used for the hidden layer, and its expression is: The output layer selects the activation function according to the demand. If the value range of the stress index is any real number, the activation function is not used.

[0020] Before training the neural network, the connection weights and biases between neurons are initialized, usually using a random initialization method. For example, for weights, they can be initialized by random sampling from a normal distribution with a mean of 0 and a small standard deviation (such as 0.01); for biases, they can be initialized to 0. This ensures that at the beginning of training, each neuron in the network has an initial state, and the initial states of different neurons are different, which is conducive to the network learning different features; in each training iteration, a sample in the training set is input into the neural network, starting from the input layer, by calculating the weighted sum of the neurons and the output of the activation function, the signal is propagated forward layer by layer until the output layer obtains the predicted stress index.

[0021] The loss between the predicted stress index and the true stress index is calculated using the following loss function, expressed as, where is the number of samples. The value of the loss function reflects the degree of difference between the model's predicted results and the true results. The goal of training is to minimize this loss function. The gradient of the weights and biases is calculated based on the loss function. The error is back-propagated from the output layer to the input layer through the back-propagation algorithm, and the weights and biases are updated using the gradient descent algorithm. The number of training rounds is set, that is, the number of times the entire training set is fully traversed. After each round of training, the validation set is used to evaluate the performance of the model. The samples in the validation set are input into the trained model, and the loss and accuracy on the validation set are calculated. By observing the changes in the validation set loss, it is determined whether the model is overfitting or underfitting, and overfitting and underfitting processing optimization are performed. If the model's loss on the validation set begins to rise, while the loss on the training set continues to decrease, this may indicate that the model is overfitting, which can be solved by regularization methods. If the model's loss on both the training set and the validation set is high, it may be underfitting. At this time, you can consider increasing the complexity of the network, such as increasing the number of hidden layers or neurons, extending the training time, etc., which will not be repeated here.

[0022] When the model training is completed, the newly collected pulse signal is processed according to the above-mentioned feature extraction and preprocessing steps, and then input into the model to calculate the stress index.

[0023] Step 5: Result output and evaluation: output the stress index value calculated in step 4 to the user or monitoring system, and evaluate the stress state according to the preset threshold; the user or monitoring system takes intervention measures according to the evaluation results to reduce the individual's stress level.

[0024] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for calculating a stress index based on a pulse signal, characterized in that: The method comprises the following steps: Step 1: signal acquisition, using pulse signal acquisition equipment to collect individual pulse signals in real time; Step 2: Preprocessing: preprocessing the collected pulse signal to remove useless information and noise in the pulse signal; Step 3: Feature extraction, extracting characteristic parameters related to the stress state from the preprocessed pulse signal; Step 4: Stress index calculation: Based on the extracted characteristic parameters, a stress index calculation model is constructed, and the characteristic parameters are accurately mapped into stress index values ​​through the stress index calculation model; Step 5: Result output and evaluation: output the stress index value calculated in step 4 to the user or monitoring system, and evaluate the stress state according to the preset threshold; the user or monitoring system takes intervention measures according to the evaluation results to reduce the individual's stress level.

2. The method for calculating a stress index based on a pulse signal according to claim 1, characterized in that: The preprocessing in step 2 includes filtering, denoising and signal enhancement.

3. The method for calculating a stress index based on a pulse signal according to claim 1, characterized in that: The characteristic parameters in step three include pulse rate variability PRV, waveform morphology change and pulse wave transmission time PTT.

4. The method for calculating a stress index based on a pulse signal according to claim 3, characterized in that: In step 4, the stress index calculation model is constructed as follows: The extracted feature parameters are normalized according to the formula, where is the original data, and is the minimum and maximum value of the original data value range, and is the normalized data; The normalized characteristic parameter data and the corresponding stress index label data are divided into a training set and a validation set, and the stress index label data is obtained by measuring after the stress state is induced experimentally; and the above data are then sorted into a format suitable for neural network input; Design a multi-layer perceptron, including input layer, hidden layer and output layer. The number of neurons in the input layer is the same as the dimension of the input feature, that is, 3 neurons correspond to the three characteristic parameters of pulse rate variability PRV, waveform morphology change and pulse wave transmission time PTT. The output layer has only 1 neuron, which is used to output the stress index. The number of hidden layers and neurons is determined by experiments. Select activation functions in the hidden layer and output layer. The ReLU function is used for the hidden layer, and its expression is: The output layer selects the activation function according to the requirements. If the value range of the stress index is any real number, no activation function is used.

5. The method for calculating stress index based on pulse signal according to claim 4, characterized in that: Before training the neural network, the connection weights and biases between neurons are initialized. In each training iteration, a sample in the training set is input into the neural network, starting from the input layer, and the signal is propagated forward layer by layer by calculating the weighted sum of neurons and the output of the activation function until the output layer obtains the predicted stress index.

6. The method for calculating a stress index based on a pulse signal according to claim 5, characterized in that: Calculate the loss between the predicted stress index and the true stress index through the following loss function, expressed as, where is the number of samples, the value of the loss function reflects the degree of difference between the model prediction result and the true result, and the training goal is to minimize this loss function; calculate the gradient of the weight and bias according to the loss function, propagate the error from the output layer to the input layer through the back propagation algorithm, and use the gradient descent algorithm to update the weight and bias; set the number of training rounds, that is, the number of times the entire training set is fully traversed. After each round of training, use the validation set to evaluate the performance of the model, input the samples in the validation set into the trained model, calculate the loss and accuracy on the validation set, and judge whether the model is overfitting or underfitting by observing the change in the validation set loss; When the model training is completed, the newly collected pulse signal is processed according to the above-mentioned feature extraction and preprocessing steps, and then input into the model to calculate the stress index.

Images