Multi-mode biofeedback allergy defense massage control method

Abstract

The multi-modal biofeedback allergy defense massage control method comprises the following steps: acquiring multi-modal data of a user in a massage process in real time, determining whether the user has an allergy symptom before massage, and when the user has the allergy symptom before massage, outputting the acquired multi-modal data at a first moment as first multi-modal data; when the user does not have allergy symptoms before massage, constructing an antagonistic allergy risk model based on sympathetic / parasympathetic nerve activity indexes; performing real-time risk assessment by the antagonistic allergy risk model based on the multi-modal data acquired in real time, and outputting the multi-modal data as second multi-modal data when the risk assessment result is that the user has an allergy symptom; constructing a massage control model, and adjusting massage parameters of the massage equipment based on the first multi-modal data or the second multi-modal data by adopting the massage control model; according to the invention, the massage parameters of the user who is allergic or allergic in the massage process can be adjusted, so that the allergy symptom is relieved.

Description

Technical Field

[0001] The present invention relates to the technical field of massage control, and in particular to a multi-modal biofeedback allergy prevention massage control method. Background Art

[0002] With the rapid development of the economy and society and the significant improvement in living standards, the public's attention to health management has been increasing. Whether it is urban employees who work at desks for a long time and are under high pressure, or farmers who engage in high-intensity physical labor, they will face fatigue, pain and soreness in the joints and local muscles in their daily work. Taking office workers as an example, maintaining a fixed sitting posture for a long time can easily cause cervical stiffness and lumbar muscle strain; farmers often experience symptoms such as shoulder pain and knee discomfort after heavy farm work. These discomforts not only affect work efficiency, but also threaten physical health. Massage, as a safe and effective physical relief method, can be an important choice to improve physical condition, which has strongly promoted the vigorous development of the massage industry.

[0003] At present, the massage industry has mainly formed two major technical paths: manual massage and equipment massage. Manual massage relies on the professional skills and experience of technicians. Although it can meet some personalized needs through precise acupoint positioning and manipulation changes, it is essentially limited to mechanical pressure and kneading. There are bottlenecks in service efficiency and effect stability, and it is difficult to provide diversified intervention methods such as electrical stimulation, temperature regulation, and magnetic therapy. In contrast, equipment massage, with the advantages of modern technology, integrates multiple functions such as electric pulse stimulation, heat therapy, and magnetic therapy. It can deeply stimulate muscle tissue and promote blood circulation by simulating different massage techniques, significantly improving the efficiency of relieving physical fatigue and discomfort. For example, some high-end massage chairs can effectively activate deep muscle fibers through electrical stimulation, and with the precise temperature-controlled hot compress function, accelerate lactic acid metabolism, quickly relieve muscle soreness, and bring users a more efficient and comfortable massage experience.

[0004] However, the massage equipment currently on the market still has obvious technical shortcomings. Most devices only focus on relieving physical fatigue and general discomfort, and lack the ability to deeply perceive and analyze the user's physiological data. On the one hand, the equipment generally uses preset fixed programs or simple mode switching functions, which cannot capture the dynamic changes of the user's heart rate, skin electrical response, body temperature and other physiological indicators in real time, and it is difficult to intelligently adjust the massage parameters according to the individual's physical condition; on the other hand, for allergies, a common and widely affecting health problem, existing equipment completely ignores the relationship between allergic reactions and massage interventions. When the user has allergic symptoms, the equipment can neither identify the allergic risk in time by monitoring relevant physiological data, nor can it dynamically adjust the massage strength, frequency, temperature and action points based on the specific needs of the allergic reaction. As a result, the massage service cannot effectively relieve allergic symptoms, and may even aggravate the allergic reaction due to improper stimulation, and cannot meet the user's growing diversified health needs. Summary of the Invention

[0005] In view of this, the present invention proposes a multimodal biofeedback allergy prevention massage control method, which combines multimodal data to alleviate the user's allergic symptoms by changing the massage method.

[0006] The technical solution of the present invention is achieved as follows:

[0007] A multimodal biofeedback allergy prevention massage control method comprises the following steps:

[0008] Step S1, collecting multimodal data of the user during the massage process in real time, determining whether the user has allergic symptoms before the massage, and if the user has allergic symptoms before the massage, outputting the collected multimodal data at the first moment as first multimodal data;

[0009] Step S2: When the user does not experience allergic symptoms before the massage, an antagonistic allergy risk model is constructed based on the sympathetic / parasympathetic nerve activity index;

[0010] Step S3: The antagonistic allergy risk model performs real-time risk assessment based on the multimodal data collected in real time, and outputs the multimodal data when the user experiences allergic symptoms as a risk assessment result as second multimodal data;

[0011] Step S4: constructing a massage control model, and using the massage control model to adjust massage parameters of the massage device based on the first multimodal data or the second multimodal data.

[0012] Preferably, the specific steps of collecting multimodal data of the user during the massage process in real time in step S1 include:

[0013] Step S11: collecting skin conductance signals through electrodes attached to the user's skin, using wavelet threshold noise reduction, and extracting burst activity features through a convolutional neural network to obtain an EDA value;

[0014] Step S12: The user's respiratory variability rate (RVR) is collected through a respiratory sensor, and the ambient air in the user's environment is collected through an air sampler. The allergen concentration C is obtained by analysis, and the user's symptom score and emotion score are scored using an expert scoring method to obtain a symptom score S and an emotion score E, respectively.

[0015] Step S13: Collect the user's electrocardiogram signal ECG through an electrocardiograph, and convert the ECG into sympathetic nerve activity SA and parasympathetic nerve activity PA.

[0016] Preferably, the specific steps of step S13 are:

[0017] Step S131: sticking the electrodes of the electrocardiograph to the user's skin to collect the user's electrocardiogram (ECG) signal;

[0018] Step S132: extracting the RR interval sequence of the electrocardiogram signal ECG, removing ectopic beats, and interpolating missing data;

[0019] Step S133: Perform fast Fourier transform on the RR interval sequence to obtain frequency domain distribution, and obtain the high frequency band HF and the low frequency band LF from the frequency domain distribution;

[0020] Step S134 : output the high frequency band HF as the parasympathetic nerve activity PA, sum the high frequency band HF and the low frequency band LF, and output the ratio of the low frequency band LF to the summed result as the sympathetic nerve activity SA.

[0021] Preferably, the specific steps of determining whether the user has allergic symptoms before the massage in step S1 are: before the massage, the user inputs whether the user has allergic symptoms; if the user has allergic symptoms, the multimodal data collected at the first moment when the user uses the massage device for massage is output as the first multimodal data.

[0022] Preferably, the antagonistic allergy risk model of step S2 includes a feature extraction layer, an antagonistic fusion layer and a risk classifier connected in sequence, the feature extraction layer includes a CNN-based SA branch and an LSTM-based PA branch, the outputs of the SA branch and the PA branch are connected to the antagonistic fusion layer, the antagonistic fusion layer describes the antagonistic interaction between sympathetic nerve activity SA and parasympathetic nerve activity PA through an interspecies competition model, and obtains the antagonistic interaction update value of sympathetic nerve activity SA and parasympathetic nerve activity PA, and the risk classifier outputs the risk assessment result through a risk scoring function.

[0023] Preferably, the expression of the interspecific competition model is:

[0024]

[0025] in, and are the antagonistic interaction update values ​​of sympathetic nerve activity SA and parasympathetic nerve activity PA, α is the intrinsic growth rate of sympathetic nerve activity SA, β is the inhibitory effect of parasympathetic nerve activity PA on sympathetic nerve activity SA, γ is the promoting effect of allergen concentration C on sympathetic nerve activity SA, δ is the intrinsic growth rate of parasympathetic nerve activity PA, ∈ is the inhibitory effect of sympathetic nerve activity SA on parasympathetic nerve activity PA, and ζ is the inhibitory effect of sympathetic nerve activity SA on parasympathetic nerve activity PA.

[0026] Preferably, the risk scoring function is expressed as: Where σ is the Sigmoid function, W1 and W2 are the negative and positive weights reflecting the antagonistic orientation, b is the bias term, and NI is the antagonistic intensity. Its expression is:

[0027] Preferably, the specific steps of step S3 are:

[0028] Step S31: input the sympathetic nerve activity SA and the parasympathetic nerve activity PA into the SA branch and the PA branch respectively, and extract the SA feature vector and the PA feature vector;

[0029] Step S32: input the SA feature vector and the PA feature vector into the antagonistic fusion layer to obtain the antagonistic interaction update value of the SA feature vector and the PA feature vector;

[0030] In step S33, the antagonistic interaction update value of the SA feature vector and the PA feature vector is input into the risk classifier, and the risk classifier outputs the risk assessment result in real time. When the risk assessment result indicates that the user has allergic symptoms, the multimodal data when the user has allergic symptoms is output as the second multimodal data.

[0031] Preferably, the massage control model expression of step S4 is:

[0032]

[0033] Where F is the massage intensity, f is the massage frequency, T is the massage temperature, and ω1-ω6 are weights preset according to the user's physical condition.

[0034] Preferably, the specific steps of step S4 are:

[0035] Step S41: When the user has allergic symptoms before the massage, the first multimodal data is processed and input into the massage control model;

[0036] Step S42: When the user experiences an allergic symptom during the massage, the second multimodal data collected when the user experiences the allergic symptom is input into the massage control model;

[0037] Step S43: The massage control model processes and outputs the massage intensity, massage frequency and massage cycle, and sends them to the massage device. The massage device massages the user according to the massage intensity, massage frequency and massage cycle.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] The present invention provides a multimodal biofeedback allergy prevention massage control method, which can provide targeted massage relief for the user's allergic symptoms. The user's allergic symptoms include those occurring before and during massage. After collecting multimodal data during massage, the antagonistic allergy risk model can be used to evaluate whether the user has allergic symptoms during the massage. When the user has allergic symptoms before and during massage, the multimodal data can be processed by the constructed massage control model to obtain massage parameters, which can be fed back to the massage device so that the massage device can perform targeted massage, thereby allergic symptoms can be relieved. The antagonistic antagonistic mechanism is used to evaluate the allergic risk based on sympathetic / parasympathetic nerve activity indicators, which can ensure the accuracy of the risk assessment results, so that allergic symptoms can be promptly relieved through massage when they occur during the massage. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only preferred embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0041] Figure 1 This is a flow chart of a multimodal biofeedback allergy prevention massage control method of the present invention;

[0042] Figure 2 This is a flow chart of step S1 of a multimodal biofeedback allergy prevention massage control method of the present invention;

[0043] Figure 3 This is a flow chart of step S13 of the multimodal biofeedback allergy prevention massage control method of the present invention;

[0044] Figure 4 This is a flow chart of step S3 of the multimodal biofeedback allergy prevention massage control method of the present invention;

[0045] Figure 5 This is a flow chart of step S4 of the multimodal biofeedback allergy prevention massage control method of the present invention; DETAILED DESCRIPTION

[0046] In order to better understand the technical content of the present invention, a specific embodiment is provided below, and the present invention is further described in conjunction with the accompanying drawings.

[0047] See also Figures 1 to 5 The present invention provides a multimodal biofeedback allergy prevention massage control method, comprising the following steps:

[0048] Step S1, collecting multimodal data of the user during the massage process in real time, determining whether the user has allergic symptoms before the massage, and if the user has allergic symptoms before the massage, outputting the collected multimodal data at the first moment as first multimodal data;

[0049] Step S2: When the user does not experience allergic symptoms before the massage, an antagonistic allergy risk model is constructed based on the sympathetic / parasympathetic nerve activity index;

[0050] Step S3: The antagonistic allergy risk model performs real-time risk assessment based on the multimodal data collected in real time, and outputs the multimodal data when the user experiences allergic symptoms as a risk assessment result as second multimodal data;

[0051] Step S4: constructing a massage control model, and using the massage control model to adjust massage parameters of the massage device based on the first multimodal data or the second multimodal data.

[0052] The present invention provides a multimodal biofeedback allergy prevention massage control method, which is applied to the control and scheduling of intelligent massage equipment, wherein the massage equipment can massage different parts of the user's body, and in addition to traditional mechanical pressure massage, it can also provide electrical stimulation, magnetic therapy, hot and cold stimulation, etc. In addition, allergy relief regulation is also added. When the user has allergic symptoms, the massage parameters of the massage equipment can be adjusted to relieve the user's allergic symptoms. When the user's allergic symptoms are mild, the activities of the sympathetic and parasympathetic nerves can be regulated through massage to restore them to a balanced state. When the sympathetic nerves are excited, they will inhibit the overreaction of the immune system, while the moderate activation of the parasympathetic nerves will help the body relax and recover. For example, massage can reduce the tension of the sympathetic nerves, reduce symptoms such as bronchial constriction and vasodilation caused by allergies, and relieve problems such as difficulty breathing and skin redness and swelling.

[0053] The occurrence of the user's allergic symptoms includes two nodes. The first node is that the user has already experienced allergic symptoms before the massage, and the second node is that the user has experienced allergic symptoms during the massage. When the user has allergic symptoms before the massage, the multimodal data collected by the user at the first moment can be recorded as the first multimodal data. If the customer has allergic symptoms during the massage, the multimodal data when the allergic symptoms occur can be recorded as the second multimodal data. The first multimodal data and the second multimodal data are used for the massage control model to obtain massage parameters. The massage control model can obtain the massage parameters corresponding to the needle based on the first multimodal parameters or the second multimodal parameters, and after feeding back to the massage device, the massage device performs targeted massage to relieve the user's allergic symptoms.

[0054] In order to improve the accuracy of judging whether the user has allergic symptoms during the massage process, the present invention constructs an antagonistic allergy risk model based on the sympathetic / parasympathetic nerve activity index. The multimodal data collected by the user in real time during the massage process will be transmitted to the antagonistic allergy risk model. The antagonistic allergy risk model will perform non-invasive detection based on the sympathetic / parasympathetic nerve activity index and obtain risk assessment results. The risk assessment results can be used to judge whether the user has allergic symptoms, and can capture subtle changes in the body before the allergic symptoms are obvious, thereby achieving early and accurate warning.

[0055] Preferably, the specific steps of collecting multimodal data of the user during the massage process in real time in step S1 include:

[0056] Step S11: collecting skin conductance signals through electrodes attached to the user's skin, using wavelet threshold noise reduction, and extracting burst activity features through a convolutional neural network to obtain an EDA value;

[0057] Step S12: The user's respiratory variability rate (RVR) is collected through a respiratory sensor, and the ambient air in the user's environment is collected through an air sampler. The allergen concentration C is obtained by analysis, and the user's symptom score and emotion score are scored using an expert scoring method to obtain a symptom score S and an emotion score E, respectively.

[0058] Step S13: Collect the user's electrocardiogram signal ECG through an electrocardiograph, and convert the ECG into sympathetic nerve activity SA and parasympathetic nerve activity PA.

[0059] Regardless of whether the user experiences allergic symptoms before or during the massage, multimodal data of the user needs to be collected. In addition to the user's own body data, multimodal data also includes environmental data and user feedback. The user's own body data includes skin conductance signal, respiratory variability rate RVR and electrocardiogram signal ECG. Environmental data includes allergen concentration C, and user feedback includes symptom score S and emotion score E. The collected different modal data can be used for different models. For example, EDA value and respiratory variability rate RVR are used to select massage parameters in the massage control model, while the sympathetic nerve activity SA and parasympathetic nerve activity PA converted from electrocardiogram ECG can be used to construct an antagonistic allergy risk model.

[0060] Among them, the skin conductance signal mainly reflects the conductance changes caused by the secretion activity of sweat glands on the skin surface, which is closely related to sympathetic nerve activity. The signal is usually obtained by placing electrodes on the skin and measuring changes in skin resistance or conductance. The respiratory variability rate RVR refers to the changes in the respiratory cycle. It is usually calculated by monitoring the respiratory signal through a respiratory sensor and then analyzing the fluctuations in the respiratory cycle. The allergen concentration C is collected by an air sampler to analyze the allergen particles in the air. The symptom score S and the emotion score E can be obtained by expert scoring based on the patient's objective performance. The electrocardiogram ECG requires the use of an electrocardiograph. The collected electrocardiogram ECG can display the activity status of the human body. The sympathetic nerve activity SA and parasympathetic nerve activity PA can be extracted through special processing methods, which are used to determine whether allergic symptoms occur.

[0061] Preferably, the specific steps of step S13 are:

[0062] Step S131: sticking the electrodes of the electrocardiograph to the user's skin to collect the user's electrocardiogram (ECG) signal;

[0063] Step S132: extracting the RR interval sequence of the electrocardiogram signal ECG, removing ectopic beats, and interpolating missing data;

[0064] Step S133: Perform fast Fourier transform on the RR interval sequence to obtain frequency domain distribution, and obtain the high frequency band HF and the low frequency band LF from the frequency domain distribution;

[0065] Step S134 : output the high frequency band HF as the parasympathetic nerve activity PA, sum the high frequency band HF and the low frequency band LF, and output the ratio of the low frequency band LF to the summed result as the sympathetic nerve activity SA.

[0066] Electrocardiogram (ECG) data needs to be collected by an electrocardiograph. After the electrodes are attached to the user's skin, the electrocardiograph can automatically collect the ECG data. The Pan-Tompkins algorithm can then be used to extract the RR interval (intervals between adjacent heartbeats) sequence from the ECG. The extracted RR interval sequence needs to be preprocessed. The specific processing includes removing ectopic beats and interpolating missing data to ensure data accuracy. The preprocessed RR interval sequence can be subjected to fast Fourier transform to obtain the frequency domain distribution, which includes:

[0067] High frequency band (HF, 0.15–0.4 Hz): mainly regulated by the parasympathetic nervous system and directly reflects PA;

[0068] Low frequency band (LF, 0.04–0.15 Hz): affected by both sympathetic and parasympathetic nerves, but sympathetic nerves dominate in the resting state;

[0069] Very low frequency band (VLF, <0.04Hz): related to fluid regulation, etc., usually ignored or used for baseline correction;

[0070] The high-frequency band HF and low-frequency band LF can be directly obtained from the frequency domain distribution. The parasympathetic nerve activity PA is positively correlated with the high-frequency band HF. Therefore, the high-frequency band HF can be directly output as the parasympathetic nerve activity PA. The sympathetic nerve activity SA is expressed as follows:

[0071] Preferably, the specific steps of determining whether the user has allergic symptoms before the massage in step S1 are: before the massage, the user inputs whether the user has allergic symptoms; if the user has allergic symptoms, the multimodal data collected at the first moment when the user uses the massage device for massage is output as the first multimodal data.

[0072] Before the user performs a massage, relevant information input by the user is received through the interactive screen of the smart terminal or the massage device, including the user's basic information and whether allergic symptoms occur. That is, the user actively inputs whether allergic symptoms occur. When the user has allergic symptoms, the multimodal data collected by the collection device from the user can be output as the first multimodal data.

[0073] Preferably, the antagonistic allergy risk model of step S2 includes a feature extraction layer, an antagonistic fusion layer and a risk classifier connected in sequence, the feature extraction layer includes a CNN-based SA branch and an LSTM-based PA branch, the outputs of the SA branch and the PA branch are connected to the antagonistic fusion layer, the antagonistic fusion layer describes the antagonistic interaction between sympathetic nerve activity SA and parasympathetic nerve activity PA through an interspecies competition model, and obtains the antagonistic interaction update value of sympathetic nerve activity SA and parasympathetic nerve activity PA, and the risk classifier outputs the risk assessment result through a risk scoring function.

[0074] The expression of the interspecific competition model is:

[0075]

[0076] in, and are the antagonistic interaction update values ​​of sympathetic nerve activity SA and parasympathetic nerve activity PA, α is the inherent growth rate of sympathetic nerve activity SA, which is the rate of growth of sympathetic nerve activity SA itself when there is no influence of parasympathetic nerve activity PA and allergen concentration C, and is usually a positive value; β is the inhibitory effect of parasympathetic nerve activity PA on sympathetic nerve activity SA, and the larger the β value, the stronger the antagonistic effect of PA on SA; γ is the promoting effect of allergen concentration C on sympathetic nerve activity SA, and the larger the γ value, the greater the influence of the change of allergen concentration C on SA; δ is the parasympathetic nerve activity PA The intrinsic growth rate is the rate of growth of PA itself when there is no influence of sympathetic nerve activity SA, which is generally a positive value; ∈ is the inhibitory effect of sympathetic nerve activity SA on parasympathetic nerve activity PA. The larger the ∈ value, the stronger the antagonistic effect of SA on PA; ζ is the inhibitory effect of sympathetic score S on parasympathetic nerve activity PA. The larger the ζ value, the more obvious the inhibitory effect of sympathetic score S on PA. By measuring the changes in variables such as sympathetic nerve activity, parasympathetic nerve activity, allergen concentration, and symptom score, the parameter values ​​α, β, γ, δ, ∈, and ζ can be estimated through data analysis and fitting.

[0077] The expression of the risk scoring function is: Where σ is the Sigmoid function, W1 and W2 are the negative and positive weights that reflect the antagonistic orientation. The specific weights can be obtained through model training. For example, the gradient descent algorithm is used in the training process to minimize the difference between the predicted risk and the actual risk. The values ​​of W1 and W2 are continuously adjusted until the model converges to a better state, thereby obtaining the optimal weight value suitable for the data. b is the bias term, which is used to adjust the output of the function so that the model can better fit the data. NI is the antagonistic intensity, which is expressed as follows:

[0078] The antagonistic allergy risk model of the present invention includes three layers from top to bottom. The first layer is the feature extraction layer, the second layer is the antagonistic fusion layer, and the third layer is the risk classifier. The multimodal data passes through the feature extraction layer, the antagonistic fusion layer and the risk classifier in sequence. The feature extraction layer includes a dual-branch structure, namely the SA branch and the PA branch, which can respectively extract features of the sympathetic nerve activity SA and the parasympathetic nerve activity PA. The antagonistic fusion layer can describe the antagonistic interaction of the sympathetic nerve activity SA and the parasympathetic nerve activity PA through the interspecies competition model, and update the sympathetic nerve activity SA and the parasympathetic nerve activity PA thereby obtaining the antagonistic interaction update value of the sympathetic nerve activity SA and the parasympathetic nerve activity PA. Finally, the risk classifier can perform risk scoring based on its built-in risk scoring function, and determine whether the user has allergic symptoms based on the risk assessment results.

[0079] By introducing an antagonistic mechanism, non-invasive detection can be performed. At the same time, based on sympathetic nerve activity SA and parasympathetic nerve activity PA, an accurate assessment of whether an allergy is present can be made. Sympathetic / parasympathetic nerve activity indicators can be monitored continuously in real time, and changes in the user's nerve activity in different environments and at different times can be tracked in a timely manner, providing a comprehensive understanding of the dynamic development of allergic symptoms.

[0080] Preferably, the specific steps of step S3 are:

[0081] Step S31: input the sympathetic nerve activity SA and the parasympathetic nerve activity PA into the SA branch and the PA branch respectively, and extract the SA feature vector and the PA feature vector;

[0082] Step S32: input the SA feature vector and the PA feature vector into the antagonistic fusion layer to obtain the antagonistic interaction update value of the SA feature vector and the PA feature vector;

[0083] In step S33, the antagonistic interaction update value of the SA feature vector and the PA feature vector is input into the risk classifier, and the risk classifier outputs the risk assessment result in real time. When the risk assessment result indicates that the user has allergic symptoms, the multimodal data when the user has allergic symptoms is output as the second multimodal data.

[0084] The multimodal data collected during the massage process are input into the antagonistic allergy risk model, where the main input data are sympathetic nerve activity SA and parasympathetic nerve activity PA. The sympathetic nerve activity SA and parasympathetic nerve activity PA are respectively input into the SA branch and PA branch for parallel processing. The SA branch can process to obtain the SA feature vector, and the PA branch can process to obtain the PA feature vector. Then the SA feature vector and the PA feature vector can be input into the antagonistic fusion layer for antagonistic interaction update. The obtained antagonistic interaction update value can be input into the risk classifier, and the risk classifier outputs the risk assessment result. Finally, based on the risk assessment result, it can be determined whether the user has allergic symptoms and the severity of the user's allergic symptoms can be assessed.

[0085] Preferably, the massage control model expression of step S4 is:

[0086]

[0087] Where F is the massage intensity, f is the massage frequency, T is the massage temperature, and ω1-ω6 are weights preset according to the user's physical condition. The specific steps of step S4 are:

[0088] Step S41: When the user has allergic symptoms before the massage, the first multimodal data is processed and input into the massage control model;

[0089] Step S42: When the user experiences an allergic symptom during the massage, the second multimodal data collected when the user experiences the allergic symptom is input into the massage control model;

[0090] Step S43: The massage control model processes and outputs the massage intensity, massage frequency and massage cycle, and sends them to the massage device. The massage device massages the user according to the massage intensity, massage frequency and massage cycle.

[0091] Massage parameters include massage intensity, massage frequency and massage temperature. When the user experiences allergic symptoms, the first multimodal data or the second multimodal data can be input into the massage control model, which is processed by the massage control model to obtain massage parameters. The massage parameters can be fed back to the massage device, so as to control the massage head and other components to adjust the massage method, so as to alleviate the user's allergic symptoms in a targeted manner.

[0092] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A multimodal biofeedback allergy prevention massage control method, characterized in that: The following steps are involved: Step S1, collecting multimodal data of the user during the massage process in real time, determining whether the user has allergic symptoms before the massage, and if the user has allergic symptoms before the massage, outputting the collected multimodal data at the first moment as first multimodal data; Step S2: When the user does not experience allergic symptoms before the massage, an antagonistic allergy risk model is constructed based on the sympathetic / parasympathetic nerve activity index; Step S3: The antagonistic allergy risk model performs real-time risk assessment based on the multimodal data collected in real time, and outputs the multimodal data when the user experiences allergic symptoms as a risk assessment result as second multimodal data; Step S4: constructing a massage control model, and using the massage control model to adjust massage parameters of the massage device based on the first multimodal data or the second multimodal data.

2. The multimodal biofeedback allergy prevention massage control method according to claim 1, characterized in that: The specific steps of step S1 of collecting multimodal data of the user during massage in real time include: Step S11: collecting skin conductance signals through electrodes attached to the user's skin, using wavelet threshold noise reduction, and extracting burst activity features through a convolutional neural network to obtain an EDA value; Step S12: The user's respiratory variability rate (RVR) is collected through a respiratory sensor, and the ambient air in the user's environment is collected through an air sampler. The allergen concentration C is obtained by analysis, and the user's symptom score and emotion score are scored using an expert scoring method to obtain a symptom score S and an emotion score E, respectively. Step S13: Collect the user's electrocardiogram signal ECG through an electrocardiograph, and convert the ECG into sympathetic nerve activity SA and parasympathetic nerve activity PA.

3. The multimodal biofeedback allergy prevention massage control method according to claim 2, characterized in that: The specific steps of step S13 are: Step S131: sticking the electrodes of the electrocardiograph to the user's skin to collect the user's electrocardiogram (ECG) signal; Step S132: extracting the RR interval sequence of the electrocardiogram signal ECG, removing ectopic beats, and interpolating missing data; Step S133: Perform fast Fourier transform on the RR interval sequence to obtain frequency domain distribution, and obtain the high frequency band HF and the low frequency band LF from the frequency domain distribution; Step S134 : output the high frequency band HF as the parasympathetic nerve activity PA, sum the high frequency band HF and the low frequency band LF, and output the ratio of the low frequency band LF to the summed result as the sympathetic nerve activity SA.

4. The multimodal biofeedback allergy prevention massage control method according to claim 1, characterized in that: The specific steps of determining whether the user has allergic symptoms before the massage in step S1 are: before the massage, the user inputs whether the user has allergic symptoms; if the user has allergic symptoms, the multimodal data collected at the first moment when the user uses the massage device to massage is output as the first multimodal data.

5. The multimodal biofeedback allergy prevention massage control method according to claim 2, characterized in that: The antagonistic allergy risk model of step S2 includes a feature extraction layer, an antagonistic fusion layer and a risk classifier connected in sequence. The feature extraction layer includes a CNN-based SA branch and an LSTM-based PA branch. The outputs of the SA branch and the PA branch are connected to the antagonistic fusion layer. The antagonistic fusion layer describes the antagonistic interaction between sympathetic nerve activity SA and parasympathetic nerve activity PA through an interspecies competition model, and obtains the antagonistic interaction update value of sympathetic nerve activity SA and parasympathetic nerve activity PA. The risk classifier outputs the risk assessment result through a risk scoring function.

6. The multimodal biofeedback allergy prevention massage control method according to claim 5, characterized in that: The expression of the interspecific competition model is: in, and are the antagonistic interaction update values ​​of sympathetic nerve activity SA and parasympathetic nerve activity PA, α is the intrinsic growth rate of sympathetic nerve activity SA, β is the inhibitory effect of parasympathetic nerve activity PA on sympathetic nerve activity SA, γ is the promoting effect of allergen concentration C on sympathetic nerve activity SA, δ is the intrinsic growth rate of parasympathetic nerve activity PA, ∈ is the inhibitory effect of sympathetic nerve activity SA on parasympathetic nerve activity PA, and ζ is the inhibitory effect of sympathetic nerve activity SA on parasympathetic nerve activity PA.

7. The multimodal biofeedback allergy prevention massage control method according to claim 5, characterized in that: The risk score function is expressed as: Where σ is the Sigmoid function, W1 and W2 are the negative and positive weights reflecting the antagonistic orientation, b is the bias term, and NI is the antagonistic intensity. Its expression is:

8. The multimodal biofeedback allergy prevention massage control method according to claim 5, characterized in that: The specific steps of step S3 are: Step S31: input the sympathetic nerve activity SA and the parasympathetic nerve activity PA into the SA branch and the PA branch respectively, and extract the SA feature vector and the PA feature vector; Step S32: input the SA feature vector and the PA feature vector into the antagonistic fusion layer to obtain the antagonistic interaction update value of the SA feature vector and the PA feature vector; In step S33, the antagonistic interaction update value of the SA feature vector and the PA feature vector is input into the risk classifier, and the risk classifier outputs the risk assessment result in real time. When the risk assessment result indicates that the user has allergic symptoms, the multimodal data when the user has allergic symptoms is output as the second multimodal data.

9. The multimodal biofeedback allergy prevention massage control method according to claim 3, characterized in that: The massage control model expression of step S4 is: Where F is the massage intensity, f is the massage frequency, T is the massage temperature, and ω1-ω6 are weights preset according to the user's physical condition.

10. The multimodal biofeedback allergy prevention massage control method according to claim 9, characterized in that: The specific steps of step S4 are: Step S41: When the user has allergic symptoms before the massage, the first multimodal data is processed and input into the massage control model; Step S42: When the user experiences an allergic symptom during the massage, the second multimodal data collected when the user experiences the allergic symptom is input into the massage control model; Step S43: The massage control model processes and outputs the massage intensity, massage frequency and massage cycle, and sends them to the massage device. The massage device massages the user according to the massage intensity, massage frequency and massage cycle.

Images