Selective attention quantitative evaluation method based on pulse wave signals

Abstract

The invention discloses a selective attention quantitative evaluation method based on pulse wave signals, which comprises the following steps: acquiring pulse wave signals of a testee in a static sitting state, extracting physiological indexes such as pulse variability and oxyhemoglobin saturation, and objectively quantifying the selective attention of the testee by using the indexes. The method comprises the following steps: recording a pulse wave signal of a testee before executing a visual search task and the accuracy rate and reaction time of executing the task; extracting physiological indexes from the denoised pulse wave signals; constructing a data set based on the data of multiple testees, and obtaining a feature subset and an importance weight thereof through random forest and cross validation; eliminating the dimensions of the physiological parameter features of the to-be-detected object through dimensionless processing; and performing weighted summation on the feature data through the importance weight to obtain a quantitative evaluation result of selective attention. According to the invention, a comfortable and lossless measurement method can be provided for quantitative evaluation of selective attention.

Description

Technical Field

[0001] The present invention relates to the field of information processing technology, and in particular to a method for quantitatively evaluating selective attention based on pulse wave signals. Background Art

[0002] In real-life situations, multiple stimuli often impact the human senses simultaneously, yet individuals can only process a limited number of them within a given timeframe. This cognitive process of prioritizing specific information from the external environment while ignoring irrelevant information is known as selective attention. Selective attention has important evolutionary significance, ensuring that individuals can dedicate limited mental resources within a limited timeframe to processing stimuli or events that are crucial to their survival.

[0003] Currently, the commonly used measurement methods for selective attention can be divided into three categories: subjective scale assessment, task performance assessment, and physiological indicator assessment. Subjective scale assessment relies on questionnaires or self-assessment scales. This method is cumbersome, time-consuming, has a time lag, cannot reflect the state of the subject in real time, and is easily affected by the subject's subjectivity. Although task performance can indirectly assess selective attention, its assessment criteria need to be set according to the specific task and also have limitations. Physiological indicator assessment mainly involves contact physiological parameter measurement, including eye tracking technology, electroencephalogram (EEG), event-related potential (ERP), etc., which directly measures the subject's physiological response through wearable devices or sensors. However, these devices are highly restrictive and can easily cause discomfort to the subject. In addition, the equipment is expensive, bulky, and has high maintenance costs, and requires professional personnel for operation and data analysis.

[0004] Pulse wave signals contain a wealth of cardiovascular information. Selective attention correlates with both the time-domain and frequency-domain characteristics of heart rate variability. Therefore, detecting the time-frequency characteristics of pulse wave signals can indirectly analyze autonomic nervous system activity. Pulse wave signal acquisition is simple and convenient, and wearable devices for measuring pulse wave signals are readily available, mature, and possess high practical value. Therefore, using pulse wave signals to objectively and quantitatively assess selective attention is of great significance. Summary of the Invention

[0005] The purpose of the present invention is to solve the defects in the prior art and provide a method for quantitatively evaluating selective attention based on pulse wave signals.

[0006] The purpose of the present invention can be achieved by taking the following technical solutions:

[0007] A method for quantitatively evaluating selective attention based on pulse wave signals comprises the following steps:

[0008] S1. Construct a test scenario, collect pulse wave signals from a sitting subject, and then record the accuracy and reaction time of a visual search task to calculate a task score; wherein the visual search task involves finding a specified character among several similar characters;

[0009] S2. Decomposing the pulse wave signal collected in the sitting state and denoising it using a wavelet threshold denoising method to obtain a denoised pulse wave signal;

[0010] S3. Identifying cycle-by-cycle feature points of the denoised pulse wave signal and extracting pulse variability characteristic indicators based on the identification results, wherein the pulse variability characteristic indicators include time domain features and nonlinear features; calculating the power spectrum of the denoised pulse wave signal to obtain frequency domain features, and calculating the subject's blood oxygen saturation according to a preset formula;

[0011] S4. Testing multiple subjects with the visual search task and collecting pulse wave signals in a sitting state to construct a data set, and then obtaining feature subsets and their importance weights through random forest and cross-validation;

[0012] S5. Having the subject perform the visual search task, repeating steps S1-S3, obtaining the subject's task score, the resting blood oxygen saturation, and the pulse rate variability characteristic index, and removing the dimension of the subject's physiological parameter characteristics by dimensionless transformation;

[0013] S6. Perform weighted summation through the importance weights to obtain a quantitative evaluation result of the selective attention of the subject to be tested.

[0014] Furthermore, the process of step S1 is as follows:

[0015] S101. The subject sits in front of a display screen, L centimeters away from the screen. The display screen is controlled by an external laptop computer, which instructs the subject to follow the test steps. The external pulse wave signal acquisition module collects the subject's resting pulse wave signal. The resting pulse wave signal collected in this step serves as the basis for subsequent analysis of the correlation between physiological indicators and selective attention.

[0016] S102, presenting the visual search task on a display screen. First, specify the characters to be found in the visual search task, and then begin executing the visual search task. The reaction time from the display of the characters to the subject clicking on any character is defined as the reaction time to complete one task. Clicking the specified character is considered correct, and clicking on other characters is considered incorrect. After clicking on any character, the subject immediately proceeds to the next task. N rounds of tasks are performed in total, and the accuracy rate and average reaction time are recorded. This step uses the visual search task for testing. The tester's distance from the screen is fixed, and the characters appear at random positions, thereby ensuring the scientific nature of the test.

[0017] S103: Set up several visual search tasks of varying difficulty levels. Repeat step S102, recording the subject's accuracy and average reaction time for each visual search task at each difficulty level. Calculate the subject's total score based on the weights assigned to each task at each difficulty level. This step is intended to increase the dimensionality of the test, minimize oversimplification, and ensure scientific accuracy.

[0018] Furthermore, the process of step S2 is as follows:

[0019] S201. Adaptively decompose the collected pulse wave signal using the PEUPEMD method to obtain multiple IMF (Intrinsic Mode Function) components. This step takes into account that the PEUPEMD method can solve the problems of residual noise and incomplete decomposition that other empirical mode decomposition methods have, and is superior to other related methods in terms of decomposition accuracy and mode mixing suppression;

[0020] S202, performing wavelet denoising on the selected λ high-frequency IMF components. This step takes into account that the high-frequency IMF component signals decomposed by the PEUPEMD method contain high-frequency noise such as motion artifacts, and wavelet denoising has a good effect on removing other noises such as environmental and power frequency noises;

[0021] S203, superimposing the high-frequency IMF component obtained in step S202 with the remaining IMF components to reconstruct the pulse wave signal;

[0022] S204: Use cubic spline interpolation to fit the pulse starting point of the reconstructed pulse wave signal to obtain a baseline drift curve. Subtract the baseline drift curve from the reconstructed pulse wave signal to obtain a denoised pulse wave signal. This step is to remove low-frequency interference, improve signal quality, and facilitate subsequent extraction of physiological parameters.

[0023] Furthermore, the process of step S3 is as follows:

[0024] S301, extracting pulse wave signal feature points cycle by cycle from the processed pulse wave signal; this step is to extract the maximum rising point of each cycle, which is used to calculate the interval time of each cycle to facilitate the extraction of physiological features;

[0025] S302, extracting time domain features of the pulse wave signal based on the pulse wave signal feature points extracted in step S301, obtaining the mean RR interval MeanNN, the standard deviation SDNN of the RR interval, and the proportion pNN20 of adjacent RR interval differences exceeding 20ms, and extracting nonlinear features to obtain approximate entropy ApEn and sample entropy SampEn; wherein the RR interval refers to the time interval between two consecutive heartbeats;

[0026] S303. Calculate the power spectrum of the denoised pulse wave signal using the Lomb-Scargle periodogram method, extract frequency domain features, and obtain the low-frequency band power LF, the high-frequency band power HF, and the ratio of the low-frequency band to the high-frequency band power LF / HF. The indicators selected in this step can reflect the status of the human body's sympathetic and parasympathetic nervous systems, and thus reflect the level of attention under selection from the perspective of the human nervous system.

[0027] S304 : Calculate the blood oxygen saturation of the denoised pulse wave signal according to a formula corresponding to the MAX30100 heart rate and blood oxygen sensor module.

[0028] Furthermore, the process of step S4 is as follows:

[0029] S401, performing a visual search task test on multiple subjects, repeating steps S1 to S3, obtaining task scores, resting blood oxygen saturation, and pulse rate variability characteristic indicators for the multiple subjects, and dividing the subjects into three categories based on their visual search task scores;

[0030] S402, constructing a feature subset of the above-mentioned blood oxygen saturation and pulse variability features by randomly extracting them;

[0031] S403. Evaluate and score each feature subset, using the classified f1-weighted as the scoring indicator for the feature subset. The final score is proportional to the importance of the feature subset. Upon completion, the correspondence between the number of features in different feature subsets and the score is obtained. Common feature evaluation indicators such as precision and accuracy can be used for evaluation in this step. Precision represents the proportion of positive samples that are actually positive, while recall represents the proportion of all positive samples that are correctly predicted to be positive. f1-weighted is the most commonly used feature subset scoring indicator, which comprehensively considers factors such as class frequency.

[0032] S404, calculating the number of features with the highest score based on the correspondence between the number of features and the score in step S403;

[0033] S405. Process the features in the feature subset by removing different features and calculating the scores. Calculate the importance weight of each feature based on the change in the scores, remove the feature with the lowest weight, and iterate multiple times until the number of features reaches the number of features with the highest score in step S404. This reduces the influence of irrelevant features and prevents the curse of dimensionality. Finally, the feature subset with the strongest correlation is obtained and its importance weight is recorded.

[0034] Furthermore, the process of step S5 is as follows:

[0035] S501, collecting the pulse wave signal of the subject in a resting state, repeating steps S1 to S3, and obtaining the subject's task score, resting blood oxygen saturation, and pulse variability characteristic index;

[0036] S502, selecting features from the feature subset in step S405 from the physiological features of the subject to be tested;

[0037] S503. For the above different feature dimensions, analyze the corresponding value distribution characteristics, transform the feature direction, and offset the feature value by a reference value to transform each feature into a positive value range, which is positively correlated with the human physiological state;

[0038] S504: Compress the value range of the feature selected in step S502 so that the compressed value range is between [0, 1].

[0039] Furthermore, the process of step S6 is as follows:

[0040] S601, multiplying the physiological parameter characteristics obtained in step S5 by the importance weights obtained in step S4 to obtain weighted values of each physiological parameter characteristic;

[0041] S602: Add the weighted values of the physiological parameter features to obtain a quantitative evaluation score of selective attention, where the score ranges from [0, 1].

[0042] The present invention has the following advantages and beneficial effects compared to the prior art:

[0043] (1) The present invention quantitatively evaluates selective attention from the perspective of physiological parameters. It is an objective and convenient evaluation method, which improves the subjective tendency and insufficient accuracy of evaluation methods such as questionnaires and inquiries.

[0044] (2) The present invention uses an external heart rate and blood oxygen sensor module to measure the pulse wave signal, which does not rely on complex medical instruments and can obtain quantitative scoring results of selective attention in a low-cost, convenient and fast method.

[0045] (3) In the denoising of pulse wave signals, the present invention proposes a method of improved uniform phase empirical mode decomposition based on permutation entropy combined with wavelet threshold denoising, which has a higher signal-to-noise ratio than other existing signal denoising methods and can improve the accuracy of pulse wave signal feature extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0047] Figure 1 This is a flow chart of a method for quantitatively evaluating selective attention based on pulse wave signals disclosed in the present invention;

[0048] Figure 2 is a flowchart of denoising the original pulse wave in Example 1 of the present invention;

[0049] Figure 3 is a flow chart of extracting physiological features from a denoised pulse wave signal in Example 1 of the present invention;

[0050] Figure 4 is a waveform diagram of the original pulse wave signal in Example 2 of the present invention;

[0051] Figure 5 Schematic diagram of pulse wave signal feature point extraction in Example 2 of the present invention;

[0052] Figure 6 Schematic diagram of the corresponding relationship between the number of features in different feature subsets and the f1-weighted score in Example 2 of the present invention. DETAILED DESCRIPTION

[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings 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 making creative efforts shall fall within the scope of protection of the present invention.

[0054] References to "embodiments" in this application mean that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described in this application may be combined with other embodiments.

[0055] Example 1

[0056] This embodiment discloses a method for quantitatively evaluating selective attention based on pulse wave signals. Figure 1 As shown, Figure 1 The feature weight in refers to the importance weight. The specific steps are as follows:

[0057] S1. Build a test scenario, collect pulse wave signals while sitting still, and then record the accuracy and reaction time of the visual search task and calculate the score. The specific steps are as follows:

[0058] S101. The subject sits in front of a display screen, 50 cm away from the screen. The display screen is controlled by an external laptop computer, which instructs the subject to follow the test steps. The external pulse wave signal acquisition module collects the subject's resting pulse wave signal for six minutes.

[0059] S102, displaying a visual search task on the screen. First, specify the characters to be found in the visual search task, and then start the visual search task. The reaction time from the display of the characters to the subject clicking on any character is considered to complete one task. Clicking the specified character is considered correct, and clicking on other characters is considered incorrect. After clicking on any character, the subject immediately enters the next task. A total of 100 rounds of tasks are performed, and the task duration is more than 6 minutes. The accuracy rate and average reaction time are recorded;

[0060] S103. For the beginner visual search task, specify that the character "T" must be found, and 15 "L"s of the same size and color and 1 "T" are randomly displayed on the screen. For the upgraded visual search task, specify that the character "LT" must be found, and 15 "LL," "TT," and "TL" of the same size and color and 1 "LT" are randomly displayed on the screen. Record the accuracy and average reaction time of the subjects in the visual search tasks at these two difficulty levels, and calculate the total score according to the formula;

[0061] The total score of the visual search task is calculated as:

[0062] Score=W1*T1+W2*T2

[0063] Among them, Score is the total score, W1 is the weight of the score of the primary difficulty visual search task, T1 is the score of the primary difficulty visual search task, W2 is the weight of the score of the visual search task with increased difficulty, and T2 is the score of the visual search task with increased difficulty.

[0064] T i The calculation formula for (i=1e2) is:

[0065]

[0066] Among them, MRT i is the median reaction time of the visual search task of all subjects, RT i is the reaction time of the visual search task, A i is the accuracy of the visual search task.

[0067] S2. Decompose the pulse wave signal collected in the sitting state and denoise it using a wavelet threshold denoising method to obtain a denoised pulse wave signal. The specific steps are as follows:

[0068] like Figure 2 This is the denoising flowchart of the original pulse wave mentioned above.

[0069] S201, using the PEUPEMD method to adaptively decompose the collected pulse wave signal to obtain 9 components;

[0070] S202, performing wavelet threshold denoising on the first three components with higher frequencies, wherein the wavelet basis function in the wavelet threshold denoising is the "db9" wavelet, the number of wavelet decomposition layers is 3, the threshold is the global threshold and the threshold function soft threshold function;

[0071] S203, superimposing the three IMF components of step S202 with the remaining six IMF components to reconstruct the pulse wave signal;

[0072] S204: Use cubic spline interpolation to fit the pulse starting position of the pulse wave signal reconstructed in step S203 to obtain a baseline drift curve. Subtract the baseline drift curve from the reconstructed pulse wave signal to obtain a denoised pulse wave signal.

[0073] The PEUPEMD method is an improved version of the Uniform Phase Empirical Mode Decomposition (UPEMD) algorithm, which is an improvement on the classic Empirical Mode Decomposition (EMD) algorithm. The execution steps of the EMD algorithm are as follows:

[0074] 1) Find all extreme points of the original signal x(t);

[0075] Use the cubic spline interpolation method to fit all the maximum and minimum points respectively to obtain the upper and lower envelopes of the signal, and average the upper and lower envelopes to obtain the signal m1(t);

[0076] 2) Calculate the difference h1(t) between the original signal x(t) and m1(t), as shown in the following formula:

[0077] h1(t)=x(t)-m1(t)

[0078] Determine whether h1(t) has reached the convergence conditions. There are generally three convergence conditions: m1(t) is close to 0; the difference between the number of local extreme points and zero crossing points of h1(t) does not exceed 1; and the set iteration threshold is reached. As long as one of the three conditions is met, it is considered that the convergence condition has been reached.

[0079] If the convergence condition is not met, the previous operation is repeated, otherwise the intrinsic component IMF1 = m1(t) and the residual res1 = h1(t) are obtained;

[0080] 3) Take the remaining term res1 as the new original signal x(t) and execute the previous steps until the remaining term res n It is a monotonic function and cannot be decomposed further. At this time, res n It is the residual term. After EMD decomposition, the original signal x(t) becomes the sum of n IMF components and the residual term, as shown in the following formula:

[0081]

[0082] n represents the number of IMFs. i Represents all eigencomponents decomposed, res n is the final residual component.

[0083] Because EMD decomposition of signals can cause modal aliasing, the UPEMD algorithm improves it. In the UPEMD algorithm, a pair of sinusoidal signals with uniform phase are added as masking signals to even out the distribution of extreme values. The UPEMD method can effectively suppress the modal aliasing phenomenon of EMD and reduce residual errors. The steps of the UPEMD algorithm are as follows:

[0084] 1) Initial parameter M IMF , T w , n p , and α0 are pre-set. The maximum number of intrinsic mode functions IMFs M IMF and period T wIt is calculated based on the signal length n, where M IMF =log2(n), T w =2 j (j=1e2e…eM IMF ). The frequency of the masking signal f w =1 / T w , phase number n p =2 p (p∈N * ), the amplitude α0 is set appropriately.

[0085] 2) Let r0(t) = x(t), j = 1, masking signal w(t; j ;f w θ k ) is calculated as:

[0086] w(t;ɑ j ;f w θ k )=α j cos[2πf w t+θ k ]

[0087] Among them, α j =α0·std(r j-1 (t)), std(·) represents the calculated standard deviation, θ k Evenly divided in [0, 2π], θ k =2π(k-1) / n p , k=1,2,…,n p .

[0088] 3) Add the masking signal to the original signal x(t) to obtain a mixed signal. Then, calculate the first IMF component by EMD:

[0089] c k，j (t) = D1[r0(t)+w(t; α j ;f w θ k )]

[0090] where the operator Di[·] represents the i-th IMF component obtained by EMD decomposition.

[0091] 4) Then, the first component IMF1 of UPEMD is obtained:

[0092]

[0093] 5) Calculate the residual r1(t) = r0(t) - C1, and then set r1(t) as the new signal. Repeat steps (2) to (4) until all IMF components and a monotonically varying residual are obtained.

[0094] The UPEMD algorithm decomposes the signal, but modal aliasing still exists. The PEUPEMD algorithm improves on this. The key step of the PEUPEMD algorithm is to detect abnormal intermittent or high-frequency noise signals, which are the main cause of modal aliasing. Once the noise signal or abnormal intermittent is detected and subtracted from the original signal, the residual is usually composed of low-frequency components, and the EMD algorithm can be directly used to decompose the residual signal. The steps of the PEUPEMD algorithm are as follows:

[0095] 1) Define x(t) as the original signal to be decomposed, set r0(t) = x(t), j = 1

[0096] 2) Construct the mask signal and add it to the original signal:

[0097] r j (t) = r0(t) + w(t; α j ;f w θ k )

[0098] 3) EMD is used to decompose mixed signals r j (t), get the first layer intrinsic mode function component c k，j (t), by c k，j (t)Average to calculate the final intrinsic mode function components

[0099] 4) Calculate C j The PE value of (t) is recorded as P j If P j is greater than the preset threshold θ0, set θ0 to 0.55. j (t) = r j-1 (t)-C j (t) is set as a new signal, and then step 3) is repeated until P j Less than θ0. The remaining low-frequency signal is obtained by separating the first j-1 IMFs from the original signal:

[0100]

[0101] 5) Residual signal r j (t) is completely decomposed by EMD:

[0102]

[0103] 6) Finally, the original signal is decomposed into:

[0104]

[0105] S3. Perform cycle-by-cycle pulse wave signal feature point identification on the denoised pulse wave signal. Based on this, extract the time domain features and nonlinear features of the pulse variability, calculate the power spectrum of the denoised pulse wave signal using the Lomb-Scargle periodogram method, extract the frequency domain features, and calculate the blood oxygen saturation of the denoised pulse wave signal according to the formula. The specific steps are as follows:

[0106] like Figure 3 This is the flowchart for extracting physiological features from the denoised pulse wave signal.

[0107] S301, extracting pulse wave signal feature points cycle by cycle from the denoised pulse wave signal;

[0108] S302, extracting time domain features of the pulse wave signal based on the pulse wave signal feature points extracted in step S301 to obtain MeanNN, SDNN, and pNN20, and extracting nonlinear features to obtain ApEn and SampEn;

[0109] S303, using the Lomb-Scargle periodogram method to calculate the power spectrum of the denoised pulse wave signal, extract frequency domain features, and obtain LF, HF, and LF / HF;

[0110] S304. Calculate the blood oxygen saturation according to the formula corresponding to the MAX30100 heart rate and blood oxygen sensor module:

[0111] SpO2=A*R 2 +B*R+C

[0112] Where A = -45.060, B = 30.354, and C = 94.845. The formula for calculating R is:

[0113]

[0114] Among them, red AC is the AC part of the red light signal, red DC is the DC part of the red light signal, ir AC is the AC part of the infrared light signal, ir DC is the DC part of the infrared light signal. The above signals can all be extracted from the denoised pulse wave signal.

[0115] S4. 70 subjects were tested on a visual search task and pulse wave signal collection in a sitting state. The pulse wave signal collection lasted four minutes. Each collected pulse wave signal was divided into two parts, each lasting two minutes, resulting in 140 sets of data. Feature subsets and their importance weights were then obtained through random forest and cross-validation. The specific steps are as follows:

[0116] S401. Perform a visual search task test and collect pulse wave signals while sitting still on 70 subjects. The 70 subjects are divided into three categories based on their visual search task scores. The pulse wave signal collection lasts six minutes. Each collected pulse wave signal is divided into three parts, each lasting two minutes. Repeat steps S2 and S3 to obtain the visual search task test scores and blood oxygen and pulse rate variability characteristic indicators while sitting still for the 70 subjects. After division, a total of 140 sets of data are obtained.

[0117] S402: constructing feature subsets by randomly extracting the above features, performing random extraction multiple times to obtain feature subsets with different numbers of features;

[0118] S403. Evaluate and score each feature subset. The evaluation index uses the classification accuracy feature, and the scoring uses the classification f1-weighted as the feature subset scoring index. After completion, the corresponding relationship between the number of features in different feature subsets and the scoring scores is obtained.

[0119] S404: Based on the correspondence between the number of features and the scoring scores, it is calculated that the number of features with the highest score is 11, and the score is also high when the number of features is 5-6;

[0120] S405. Process the features by removing different features and calculating the scores. The importance weight of each feature is calculated from the change in the scores. The feature with the lowest weight is removed. Multiple iterations are performed until the number of features is 11. The feature parameter set at this time is 11 parameters, including blood oxygen saturation, MeanNN, SDNN, pNN20, ApEn, SampEn, short-term change SD1 of Poincare graph, LF, HF, and LF / HF. Considering that the score when the number of features is 5 is close to that when the number is 11, in order to reduce the feature dimension and improve the calculation efficiency, the five parameters of blood oxygen saturation, MeanNN, pNN20, SampEn, and LF / HF are finally selected as evaluation indicators, and the importance weights of these five feature parameters are recorded.

[0121] S5. Have the subject perform a visual search task and repeat steps S1-S3 to obtain the subject's task score, blood oxygen and pulse rate variability characteristic indicators at rest, and remove the dimension of the subject's physiological parameter characteristics by dimensionless transformation. The specific steps are as follows:

[0122] S501. Collect pulse wave signals and reaction time and accuracy of the visual search task from five test subjects in a sitting state, and repeat steps S1 to S3 to obtain characteristic indicators of blood oxygen and pulse variability and visual search task scores of the test subjects in a resting state.

[0123] S502, selecting the five parameters in step S405 from the physiological characteristics of the subject to be measured;

[0124] S503. For the above different feature dimensions, analyze the corresponding value distribution characteristics, transform the feature direction, and offset the feature value by a reference value to transform each feature into a positive value range, which is positively correlated with the human physiological state;

[0125] S504: Compress the value ranges of all features so that the value ranges of all features are between [0, 1].

[0126] S6. Obtain the quantitative results of selective attention by weighted summation of importance weights. The specific steps are as follows:

[0127] S601, multiplying the physiological parameter characteristics obtained in step S405 and their importance weights to obtain weighted values of the physiological parameter characteristics of each human body;

[0128] S602: Add the weighted values of the physiological parameter characteristics of each human body to obtain a quantitative evaluation score of selective attention, where the score ranges from [0, 1].

[0129] Example 2

[0130] This embodiment continues to disclose a method for quantitatively evaluating selective attention based on pulse wave signals, such as Figure 1 As shown, the specific steps are as follows:

[0131] S1. Construct a test scenario, collect pulse wave signals in a sitting state, and then record the accuracy and reaction time of the visual search task and calculate the score. The measured original pulse wave signal is as follows: Figure 4 As shown;

[0132] S2. Decomposing the pulse wave signal collected in the sitting state and denoising it using a wavelet threshold denoising method to obtain a denoised pulse wave signal;

[0133] S3. Identify the pulse wave signal feature points cycle by cycle on the denoised pulse wave signal. The schematic diagram of the feature point extraction of the pre-processed pulse wave signal is as follows: Figure 5As shown, on this basis, the time domain characteristics and nonlinear characteristics of the pulse variability are extracted, and the power spectrum of the denoised pulse wave signal is calculated using the Lomb-Scargle periodogram method to extract the frequency domain characteristics, and the blood oxygen saturation of the denoised pulse wave signal is calculated according to the formula;

[0134] S4. 70 subjects were tested on visual search tasks and pulse wave signals were collected in a sitting state. The pulse wave signal collection lasted four minutes. Each collected pulse wave signal was divided into two parts, each lasting two minutes, to obtain 140 sets of data. Then, random forest and cross-validation were used to obtain feature subsets and their importance weights (i.e. Figure 1 The corresponding relationship between the number of features in different feature subsets and the f1-weighted score is as follows: Figure 6 The importance weights of each feature in the feature subset are shown in Table 1;

[0135] Table 1. Importance weights of each feature in Example 2

[0136] Blood oxygen saturation MeanNN pNN20 SampEn LF / HF 0.1001 0.1652 0.2521 0.1863 0.2963

[0137] S5. Have the subject perform a visual search task and repeat steps S1-S3 to obtain the subject's task score, resting blood oxygen saturation, and pulse rate variability characteristic indicators. De-dimension the subject's physiological parameter characteristics by dimensionless transformation. For different feature dimensions in different feature subsets in step S4, process them using the corresponding dimensionless transformation methods shown in Table 2.

[0138] Table 2. Feature processing methods for different dimensions in Example 2

[0139] Target range Dimensionless method Blood oxygen saturation [0,1] naturalization MeanNN [0,1] Anti-naturalization pNN20 [0,1] naturalization SampEn [0,1] Anti-naturalization LF / HF [0,1] naturalization

[0140] S6. The quantitative evaluation results of selective attention were obtained by weighted summation of importance weights. The comparison results of the quantitative scores of selective attention of the five test subjects and their actual scores in the visual search task are shown in Table 3.

[0141] Table 3. Quantitative evaluation results of selective attention in Example 2

[0142] Formula evaluation results Task score Object 1 75.63 76.42 Object 2 72.44 69.25 Object 3 77.55 76.08 Object 4 77.02 77.43 Object 5 80.88 79.23

[0143] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be considered as equivalent replacement methods and are included in the scope of protection of the present invention.

Claims

1. A method for quantitatively evaluating selective attention based on pulse wave signals, characterized in that: The following steps are involved: S1. Construct a test scenario, collect pulse wave signals from a sitting subject, and then record the accuracy and reaction time of a visual search task to calculate a task score; wherein the visual search task involves finding a specified character among several similar characters; S2. Decomposing the pulse wave signal collected in the sitting state and denoising it using a wavelet threshold denoising method to obtain a denoised pulse wave signal; S3. Identifying cycle-by-cycle feature points of the denoised pulse wave signal and extracting pulse variability characteristic indicators based on the identification results, wherein the pulse variability characteristic indicators include time domain features and nonlinear features; calculating the power spectrum of the denoised pulse wave signal to obtain frequency domain features, and calculating the subject's blood oxygen saturation according to a preset formula; S4. Testing multiple subjects with the visual search task and collecting pulse wave signals in a sitting state to construct a data set, and then obtaining feature subsets and their importance weights through random forest and cross-validation; S5. Having the subject perform the visual search task, repeating steps S1-S3, obtaining the subject's task score, the resting blood oxygen saturation, and the pulse rate variability characteristic index, and removing the dimension of the subject's physiological parameter characteristics by dimensionless transformation; S6. Perform weighted summation through the importance weights to obtain a quantitative evaluation result of the selective attention of the subject to be tested.

2. The method for quantitatively evaluating selective attention based on pulse wave signals according to claim 1, wherein: The process of step S1 is as follows: S101, the subject sits in front of a display screen, L cm away from the screen. The display screen is controlled by an external laptop computer, instructing the subject to follow the test steps, and an external pulse wave signal acquisition module collects the subject's pulse wave signal in a resting state; S102, presenting the visual search task on a display screen, first specifying the characters to be found in the visual search task, and then starting the visual search task. The reaction time from the display of the characters to the subject clicking on any character is defined as the reaction time to complete one task. Clicking the specified character is considered a correct result, while clicking on other characters is considered an error. After clicking on any character, the subject immediately proceeds to the next task. A total of N rounds of tasks are performed, and the accuracy rate and average reaction time are recorded. S103. Set several visual search tasks of different difficulty levels, repeat step S102, record the accuracy and average reaction time of the subjects' visual search tasks of different difficulty levels, and calculate the subjects' total scores according to the weights of the tasks of different difficulty levels.

3. The method for quantitatively evaluating selective attention based on pulse wave signals according to claim 1, wherein: The process of step S2 is as follows: S201, using an improved uniform phase empirical mode decomposition method based on permutation entropy to adaptively decompose the collected pulse wave signal to obtain multiple IMF components; S202, performing wavelet threshold denoising on the first λ high-frequency IMF components; S203, superimposing the high-frequency IMF component obtained in step S202 with the remaining IMF components to reconstruct the pulse wave signal; S204 , fitting the pulse starting point position of the reconstructed pulse wave signal using a cubic spline interpolation method to obtain a baseline drift curve, and subtracting the baseline drift curve from the reconstructed pulse wave signal to obtain a denoised pulse wave signal.

4. The method for quantitatively evaluating selective attention based on pulse wave signals according to claim 1, wherein: The process of step S3 is as follows: S301, extracting pulse wave signal feature points cycle by cycle from the processed pulse wave signal; S302, extracting time domain features of the pulse wave signal based on the pulse wave signal feature points extracted in step S301, obtaining the mean RR interval MeanNN, the standard deviation SDNN of the RR interval, and the proportion pNN20 of adjacent RR interval differences exceeding 20ms, and extracting nonlinear features to obtain approximate entropy ApEn and sample entropy SampEn; wherein the RR interval refers to the time interval between two consecutive heartbeats; S303, using the Lomb-Scargle periodogram method to calculate the power spectrum of the denoised pulse wave signal, extract frequency domain features, and obtain the low-frequency band power LF, the high-frequency band power HF, and the ratio of the low-frequency band to the high-frequency band power LF / HF; S304 : Calculate the blood oxygen saturation of the denoised pulse wave signal according to a formula corresponding to the MAX30100 heart rate and blood oxygen sensor module.

5. The method for quantitatively evaluating selective attention based on pulse wave signals according to claim 1, wherein: The process of step S4 is as follows: S401, performing a visual search task test on multiple subjects, repeating steps S1 to S3, obtaining task scores, resting blood oxygen saturation, and pulse rate variability characteristic indicators for the multiple subjects, and classifying the subjects into three categories of low, medium, and high selective attention based on their task scores; S402: constructing a feature subset of the resting state blood oxygen saturation and pulse rate variability characteristic indicators by random sampling; S403. Evaluate and score each feature subset. The final score is proportional to the importance of the feature subset to the classification. After completion, the corresponding relationship between the number of features in different feature subsets and the score is obtained. S404: Calculate the number of features with the highest score based on the correspondence between the number of features and the score. S405. Process the features in the feature subset by removing different features and calculating the scoring scores. The importance weight of each feature is calculated based on the change in the scores. The feature with the lowest weight is removed. After multiple iterations, the feature subset with the strongest correlation is finally obtained and its importance weight is recorded.

6. The method for quantitatively evaluating selective attention based on pulse wave signals according to claim 5, wherein: The process of step S5 is as follows: S501, collecting the pulse wave signal of the subject in a resting state, repeating steps S1 to S3, and obtaining the subject's task score, resting blood oxygen saturation, and pulse variability characteristic index; S502, selecting features from the feature subset in step S405 from the physiological features of the subject to be tested; S503. For the above different feature dimensions, analyze the corresponding value distribution characteristics, transform the feature direction, and offset the feature value by a reference value to transform each feature into a positive value range, which is positively correlated with the human physiological state; S504: Compress the value range of the feature selected in step S502 so that the compressed value range is between [0, 1].

7. The method for quantitatively evaluating selective attention based on pulse wave signals according to claim 6, wherein: The process of step S6 is as follows: S601, multiplying the physiological parameter characteristics obtained in step S5 by the importance weights obtained in step S4 to obtain weighted values of the physiological parameter characteristics; S602: Add the weighted values of the physiological parameter features to obtain a quantitative evaluation score of selective attention, where the score ranges from [0, 1].

Images