Sleep disorder monitoring, regulating and controlling system and monitoring, regulating and controlling method

Abstract

The invention relates to a sleep disorder monitoring, regulating and controlling system and a monitoring, regulating and controlling method, an upper computer is used as a data processing and control core, a handheld intelligent terminal, an intelligent bracelet, a monitoring unit, an environment monitoring unit and a cloud database are integrated, and the sleep disorder is monitored by judging the condition of people in a monitored area and the sleep condition of people with sleep disorder. Environment data and physiological data are combined to judge whether the person with sleep disorder really falls asleep or not, multi-parameter fusion monitoring regulation and control are achieved, data sources are wider, and regulation and control are more accurate; in a scene that a plurality of persons exist in a monitoring area, auxiliary regulation and control can be carried out on people with sleep disorder and the people with sleep disorder can be helped to fall asleep quickly without disturbing sleep of normal people, so that intelligent management of healthy sleep is realized; the cloud database is enriched and perfected, so that the sleep disorder and the regulation and control strategy are more accurate, and people with sleep disorder can be better served.

Description

Technical Field

[0001] The present invention belongs to the technical field of health management, and specifically relates to a sleep disorder monitoring and control system and a monitoring and control method. Background Art

[0002] Due to factors such as work pressure, daily chores, and environmental changes, modern people face increasing stress, leading to declining sleep quality, which in turn impacts their health and daily lives. Statistics show that the average person falls asleep after midnight, sleeps an average of 6.75 hours, and sleep duration is generally shortened. Their overall sleep score is 75, with an average of 1.4 awakenings. Twenty-eight percent of people sleep less than six hours a night, and 64% report poor sleep quality, with 22% experiencing very poor sleep quality. Therefore, how to effectively intervene in sleep disorders and promote healthy sleep has become a hot topic in society. Interventions for sleep disorders include both pharmacological and non-pharmacological interventions. Pharmacological interventions, such as hypnotics and sedatives, can help improve sleep, but long-term use can lead to drug dependence and negatively impact the digestive system and kidneys. Non-pharmacological treatments primarily include cognitive behavioral therapy (CBTI), relaxation therapy, music therapy, vision therapy, psychological counseling, and mindfulness therapy.

[0003] While existing non-drug treatments are more widely accepted by those with sleep disorders, many overlook a key issue: they are often designed for scenarios where only one person is in the room, particularly relaxation therapy, music therapy, and visual therapy. For scenarios where there are multiple people in the room, such as those with sleep disorders and those with normal sleep, using relaxation therapy, music therapy, or visual therapy to intervene in the sleep of the person with sleep disorders may disrupt the sleep of the person with normal sleep, leading to insomnia for the person with normal sleep. Therefore, in scenarios where there are multiple people in the room, how to intervene in the sleep of the person with sleep disorders without disrupting the sleep of the person with normal sleep has become a technical challenge that urgently needs to be addressed by those skilled in the art. Summary of the Invention

[0004] In order to solve the technical difficulties in the above background technology, the present invention proposes a sleep disorder monitoring and control system, which is as follows:

[0005] A sleep disorder monitoring and control system comprises at least an information collection unit, a host computer, a cloud, and a smart home terminal, wherein the information collection unit is communicatively connected to an input terminal of the host computer, the host computer is bidirectionally connected to the cloud, and the output terminal of the host computer is communicatively connected to the smart home terminal. The host computer receives data uploaded by the information collection unit and queried cloud storage data, performs deep learning and judgment, and issues sleep monitoring and control instructions to control the operating status of the smart home and assist people with sleep disorders to fall asleep.

[0006] The information collection unit includes a lower computer, a monitoring unit, an environmental monitoring unit, and a smart bracelet. The information collection unit is used to upload input data, image data, environmental data, and physiological data of sleep disorder patients to the upper computer. The lower computer is used for sleep disorder patients to manually input data. The monitoring unit is used to collect image data within the monitoring range in real time, identify the number of people and their sleep status. The environmental monitoring unit is used to monitor environmental data, and the smart bracelet is used to monitor the physiological data of sleep disorder patients in real time.

[0007] The input data includes, but is not limited to, the identity information of sleep disorder patients, work information (work type, working hours), and medical examination information. The environmental data includes the temperature H1, humidity H2, air cleanliness H3, and noise level H4 within the monitoring area. The physiological data includes blood pressure X1, blood oxygen level X2, heart rate X3, respiratory rate X4, pulse frequency X5, and acceleration X6.

[0008] The upper computer is responsible for performing data preprocessing, data fusion, and deep learning training on the data uploaded by the information collection unit, obtaining the sleep disorder monitoring results, querying the most matching regulation plan from the cloud database as the regulation strategy, and on this basis, forming a sleep monitoring regulation control instruction to control the operation status of the smart home and assist sleep disorder patients in falling asleep.

[0009] The sleep disorder monitoring results include: the number of people in the monitoring area, pupil recognition results, sleeping postures, the correlation θ1 between the sleep situation and environmental data, and the correlation θ2 between the sleep situation and physiological data.

[0010] The cloud stores historical data such as the input information and regulation plans of sleep disorder patients. These historical data are in a formatted form, which is convenient for the upper computer to call.

[0011] The smart home includes two categories. One category is the smart home related to physiology, such as headphones and massage pillows. The other category is the smart home related to the environment, such as lighting atmosphere lights, music playing devices, air conditioners, air humidifiers, air purifiers, and noise shields.

[0012] Based on the above-mentioned sleep disorder monitoring and regulation system, further, the present invention proposes a sleep disorder monitoring and regulation method, which includes the following steps:

[0013] Step 1: Input data: Sleep disorder patients fill in the input data through the lower computer APP. The input data is stored in the upper computer. At the same time, the upper computer can encrypt or perform secondary processing on the input data and then upload it to the cloud for storage.

[0014] Step 2: Monitoring data input: Obtain image data within the monitoring area through the monitoring unit, obtain environmental data through the environmental monitoring unit, and obtain physiological data of sleep disorder patients through the smart bracelet;

[0015] Among them, the environmental data includes environmental temperature H1, humidity H2, air cleanliness H3, and noise level H4; the physiological data includes blood pressure X1, blood oxygen level X2, heart rate X3, respiratory rate X4, pulse frequency X5, and acceleration X6;

[0016] Step 3: Data preprocessing and data fusion: The host computer preprocesses and fuses the image data, environmental data, and physiological data to obtain a target parameter set, and randomly divides the target parameter set into a 70% training package and a 30% test set;

[0017] Step 4: Convolution training: Input the training set into the Transformer model, add positional encoding and masking, perform encoding-decoding operations, and obtain the first training set through the self-attention structure in the Transformer model. On this basis, import the first training set into the neural network for in-depth learning training, and obtain the sleep disorder monitoring result through classification;

[0018] Step 5: Test optimization: Input the test set into the objective function test model for testing. The corresponding objective function Obj() is shown in the following formula (1):

[0019]

[0020] In the formula: N represents the amount of test data imported into the pooling layer, K represents the amount of query data imported into the pooling layer, the first term L() on the right side of the equation represents the prediction function, and the prediction function judges the sleep disorder prediction accuracy by solving the difference between the true value y i and the predicted value The second term β(f k ) represents the matching result queried through the cloud database. The specific calculation formula of β(f k ) is as follows:

[0021] β(fk ) = ρ X，F，t [X(m), F(p)] + ρ H，F，t [H(n), F(p)] + 0.5 (2)

[0022] In the formula, X(m) represents the m-item physiological data set, the value range of m is from 1 to 6, H(n) represents the n-item environmental data set, the value range of n is from 1 to 4, F(p) represents the image parameter set, p represents the number of pixels; ρ represents the correlation coefficient between two parameters, the subscript "t" and the superscript "t" both represent time, and f k represents the correlation function;

[0023] The prediction value is calculated using the proximity algorithm. Specifically, the KNN tree structure is used to calculate the prediction value of sleep disorders. In the KNN tree structure, the current prediction value is iteratively generated from the prediction value at the previous moment. L() is expressed as:

[0024]

[0025] To solve the objective function Obj(), assume that the prediction value at the initial moment is equal to the true value y1. Substitute equations (2) and (3) into equation (1) above, and solve the objective function Obj() through repeated iteration. According to the objective function Obj(), adjust the weights of the convolutional layer in the neural network;

[0026] To verify the accuracy of the prediction of the objective function Obj(), the residual index is used for evaluation. The error index includes the mean absolute error (MAE) and the mean absolute percentage error (MAPE). The corresponding calculation formulas are as follows:

[0027]

[0028] When both residual indices exceed 95%, stop the calculation. Import the objective function Obj() calculated from the test set into the neural network as an adjustment index to adjust the weights of the convolutional layer of the neural network;

[0029] Step 6: Result output and regulation strategy: When the neural network training meets the classification accuracy requirements, output the sleep disorder monitoring results. The host computer queries the regulation plan in the cloud database according to the sleep disorder monitoring results, and selects the regulation plan with the highest matching degree as the regulation strategy;

[0030] The above sleep disorder monitoring results include: the number of people in the monitoring area, the pupil recognition result, the sleeping posture, the correlation degree θ1 between the sleep situation and environmental data, and the correlation degree θ2 between the sleep situation and physiological data.

[0031] Step 7: Smart home control: According to the regulation strategy, formulate regulation parameters, generate control instructions to control the operating state of the smart home, and perform regulation when sleep disorder patients encounter sleep disorders.

[0032] On the basis of step 7, in order to enrich and improve the cloud database, make the sleep disorder and the regulation strategy more accurate, and better serve more sleep disorder patients through big data, steps 8 and 9 are also included, specifically as follows:

[0033] Step 8: Monitor the execution of the current regulation strategy during the monitoring period. If the execution effect is good, proceed to Step 8. If the execution effect is not good, return to Step 6 and continue to query other regulation schemes in the cloud database according to the matching degree. The other regulation schemes are different from the current regulation scheme.

[0034] Step 9: Upload and archive the current regulation scheme to the cloud database through the host computer.

[0035] In summary, compared with the prior art, the sleep disorder monitoring and regulation system and the monitoring and regulation method of the present invention have the following advantages:

[0036] 1) The present invention takes the host computer as the data processing and control core, integrates the handheld intelligent terminal, the smart bracelet, the monitoring unit, the environment monitoring unit, and the cloud database. By judging the personnel situation in the monitoring area and the sleep situation of sleep disorder patients, and combining the environmental data and physiological data to judge whether the sleep disorder patients are truly asleep, it realizes multi-parameter integrated monitoring and regulation, with a wider data source and more accurate regulation.

[0037] 2) For the scenario where there are multiple people in the monitoring area, it can assist in regulating sleep disorder patients without disturbing the sleep of normal sleepers, helping them fall asleep quickly and realizing intelligent management of healthy sleep.

[0038] 3) The present invention combines the Transformer model with the neural network, and uses the objective function for test optimization. The objective function is used to adjust the weights of the neural network convolutional layer, thereby greatly reducing the computational amount in the deep learning process and improving the computational performance of the host computer on the premise of ensuring the accuracy of sleep disorder prediction and sleep disorder monitoring results.

[0039] 4) In order to enrich and improve the cloud database and make the sleep disorder and regulation strategy more accurate, within a monitoring period, if the execution effect of the regulation strategy is good, it will be uploaded and archived to the cloud database, and the sleep big data can better serve sleep disorder patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a schematic structural diagram of a sleep disorder monitoring and regulation system of the present invention;

[0041] Figure 2 is a flow block diagram of a sleep disorder monitoring and regulation method of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0042] The present invention will be described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the technical solution of the present invention, but should not be regarded as a limitation to the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention.

[0043] Please examine Figure 1 , a sleep disorder monitoring and regulation system, which at least includes an information collection unit, a host computer, a cloud, and a smart home terminal. Among them, the information collection unit is communicatively connected to the input end of the host computer, the host computer is communicatively connected to the cloud bidirectionally, the output end of the host computer is communicatively connected to the smart home terminal. The host computer receives the data uploaded by the information collection unit and the data stored in the cloud queried, and performs deep learning and judgment, issues sleep monitoring and regulation control instructions, controls the operating state of the smart home, and assists sleep disorder patients to fall asleep.

[0044] The information collection unit includes a slave computer, a monitoring unit, an environmental monitoring unit, and a smart bracelet. The information collection unit is used to upload input data, image data, environmental data, and physiological data of sleep disorder patients to the host computer. Among them, the slave computer is used for sleep disorder patients to manually input data, and the input data includes the identity information, work information (work type, work duration), and medical examination information of sleep disorder patients. The slave computer can use handheld terminals such as smartphones and smart watches, and sleep disorder patients can manually input data through the APP on the operation interface of the slave computer; the monitoring unit is used to collect image data within the monitoring range in real time and identify the number of people and the sleep state of people. The environmental monitoring unit is used to monitor the environmental temperature H1, humidity H2, air cleanliness H3, and noise level H4. The smart bracelet is a wearable device for people, which can monitor the physiological data of sleep disorder patients in real time; the physiological data includes blood pressure X1, blood oxygen level X2, heart rate X3, respiratory rate X4, pulse frequency X5, and acceleration X6.

[0045] The host computer can use a microcontrol unit (MCU controller), a smart home central control screen, etc. The host computer is the core of the entire sleep disorder monitoring and regulation system, responsible for preprocessing data, data fusion, neural network training on the uploaded data, outputting classification results, querying the intervention strategy matching it from the cloud database based on the classification results, forming sleep monitoring and regulation control instructions on this basis, controlling the operating state of the smart home, and assisting sleep disorder patients to fall asleep.

[0046] The cloud, as a big data storage terminal, stores historical data such as the input information of sleep disorder patients and the monitoring and regulation plans. These historical data are formatted data, which is convenient for the host computer to call.

[0047] Smart home devices can include headphones, lighting mood lights, music players, air conditioners, air humidifiers, air purifiers, noise blockers, massage pillows, etc. Smart home devices are connected to the host computer through wireless local area networks such as WIFI, ZigBee, LoRa, Bluetooth, and NB-IoT, and accept the control of the host computer. Among them, lighting mood lights, music players, air conditioners, air humidifiers, air purifiers, and noise blockers belong to smart home devices related to the environment, while headphones and massage pillows belong to smart home devices related to physiology.

[0048] If the host computer identifies that the person with sleep disorder has not fallen asleep and there is only one person with sleep disorder based on the monitoring unit and the smart bracelet, and the environmental monitoring unit detects that the indoor temperature is high, the humidity is high, and the environmental noise exceeds 40 dB, the host computer will control the air conditioner to turn on the dehumidification mode, and at the same time turn on the music player to play soft music, or turn on the noise blocker to block the noise coming in from outside the monitoring area; when the image data shows that the person with sleep disorder has fallen asleep for a certain period of time, control the air conditioner, music player or noise blocker to turn off. If the host computer identifies that the person with sleep disorder has not fallen asleep and other people in the monitoring area have entered the sleep state based on the monitoring unit and the smart bracelet, and the environmental data all meet the standards, the host computer will only turn on the headphones worn by the person with sleep disorder and play soft music through the headphones. When the image data shows that the person with sleep disorder has fallen asleep for a certain period of time, control the headphones to stop playing. At this time, without affecting the normal sleep of others, the sleep disorder monitoring and regulation system can assist the person with sleep disorder to fall asleep quickly.

[0049] On the basis of the above content, further, the present invention proposes a sleep disorder monitoring and regulation method based on a sleep disorder monitoring and regulation system, and the method includes the following steps:

[0050] Step 1: Enter data: The person with sleep disorder fills in and enters data through the lower computer APP. The entered data includes at least identity information, work information, and medical examination information. The entered data is stored in the host computer. At the same time, the host computer can encrypt or perform secondary processing on the entered data and then upload it to the cloud for storage;

[0051] Step 2: Input monitoring data: Obtain image data in the monitoring area through the monitoring unit, obtain environmental data through the environmental monitoring unit, and obtain physiological data of the person with sleep disorder through the smart bracelet;

[0052] Among them, the environmental data includes environmental temperature H1, humidity H2, air cleanliness H3, and noise level H4, which are obtained through the environmental monitoring unit; the physiological data includes blood pressure X1, blood oxygen level X2, heart rate X3, respiratory rate X4, pulse frequency X5, and acceleration X6, which are obtained through the smart bracelet;

[0053] Step 3: Data preprocessing and data fusion: The host computer preprocesses and fuses the image data, environmental data, and physiological data. Specifically, it segments and denoises the image data, searches for the human body contour area according to the edge search algorithm. On this basis, the background area is discarded, and only the human body contour area is retained. The numerical image of the human body contour area has pixel values, grayscale values, and coordinate values. The reshape function is used to transform the numerical image into a one-dimensional image parameter set. At the same time, the environmental data and physiological data are respectively normalized to obtain a dimensionless one-dimensional environmental parameter set and a one-dimensional physiological parameter set. On this basis, the Kalman filtering algorithm is used to fuse the image parameter set, environmental parameter set, and physiological parameter set according to the time series characteristics to obtain a target parameter set, and the target parameter set is randomly divided into a 70% training package and a 30% test set;

[0054] Step 4: Convolution training: The training set is input into the Transformer model, position encoding and masking are added, and encoding-decoding operations are performed. The first training set is obtained through the self-attention structure in the Transformer model. On this basis, the first training set is imported into the convolutional layer, pooling layer, and attention layer in the neural network for in-depth learning training, and finally the classification result is obtained;

[0055] Step 5: Test and optimization: The test set is input into the objective function test model for testing. The objective function test model aims to solve the objective function through repeated iteration. The corresponding objective function Obj() is shown in the following formula (1):

[0056]

[0057] In the formula: N represents the amount of test data imported into the pooling layer, K represents the amount of query data imported into the pooling layer. The first term L() on the right side of the equation represents the prediction function. The prediction function judges the prediction accuracy of sleep disorders by solving the difference between the true value y i and the predicted value . The second term β(f k ) represents the matching result queried through the cloud database. The specific calculation formula of β(f k ) is as follows:

[0058] β(f k ) = ρ X，F，t [X(m), F(p)] + ρ H，F，t [H(n), F(p)] + 0.5 (2)

[0059] Wherein, X(m) represents the m - item physiological data set (m ranges from 1 to 6), H(n) represents the n - item environmental data set (n ranges from 1 to 4), F(p) represents the image parameter set, p represents the number of pixels; ρ represents the correlation coefficient between two parameters, the subscript "t" and the superscript "t" both represent time, and f k represents the correlation function;

[0060] The prediction value is calculated using the nearest - neighbor algorithm. Specifically, the KNN tree - type structure is used to calculate the sleep disorder prediction value. In the KNN tree - type structure, the current prediction value is generated by iterating the prediction value at the previous moment. L() is expressed as:

[0061]

[0062] To solve the objective function Obj(), assume that the prediction value at the initial moment is equal to the true value y1. Substitute equations (2) and (3) into equation (1) above, and solve the objective function Obj() through repeated iteration. According to the objective function Obj(), adjust the weights of the convolutional layer in the neural network.

[0063] To verify the accuracy of the prediction of the objective function Obj(), the error index is used for evaluation. The error index includes the mean absolute error (MAE) and the mean absolute percentage error (MAPE). The corresponding calculation formulas are as follows:

[0064]

[0065] When both of the two residual indices exceed 95%, it indicates that the objective function Obj() meets the prediction accuracy requirements. At this time, stop the calculation, and import the objective function Obj() calculated from the test set into the neural network as the adjustment index to adjust the weights of the convolutional layer of the neural network;

[0066] Step 6: Result output and regulation strategy: When the neural network training meets the classification accuracy requirements, output the sleep disorder monitoring results. The host computer queries the regulation plan in the cloud database according to the sleep disorder monitoring results, and selects the regulation plan with the highest matching degree as the regulation strategy;

[0067] The above - mentioned sleep disorder monitoring results include: the number of people in the monitoring area, the pupil recognition result, the sleeping posture, the correlation degree θ1 between the sleep situation and the environmental data, and the correlation degree θ2 between the sleep situation and the physiological data.

[0068] Step 7: Smart home control: According to the regulation strategy, formulate regulation parameters, generate control instructions to control the operation state of the smart home, and conduct regulation when sleep - disordered people encounter sleep disorders.

[0069] Step 7 above further includes steps 71 to 78, which are specifically as follows:

[0070] Step 71: When it is judged from the sleep disorder monitoring results that there is only one person with sleep disorder and the person has not fallen asleep, continue to judge the correlation degree θ1 between the falling asleep situation and the environmental data and the correlation degree θ2 between the falling asleep situation and the physiological data; if θ1 > θ2 + 0.2, go to step 72; if θ1 < θ2 - 0.2, go to step 73; if |θ1 - θ2| ≤ 0.2, go to step 74;

[0071] Step 72: The host computer only controls the operation of the smart home related to the environment;

[0072] For example, control the operation of the lighting atmosphere lamp, control the air conditioner to cool down when the temperature is relatively high in the monitoring area, control the air humidifier to operate when the indoor air is relatively dry, or control the air conditioner to turn on the "humidification" mode. When the indoor humidity is relatively high, control the air conditioner to turn on the "dehumidification" mode. When the noise level exceeds the night sleep noise level standard, control the noise shield to turn on, or control the music playing device to play soothing music. Specifically, which smart home devices to control can be set by the person with sleep disorder through the lower computer APP, and the setting results are uploaded to the host computer as input data for storage;

[0073] Step 73: The host computer only controls the operation of the smart home related to the physiology;

[0074] Step 74: Control the smart home related to the environment and the smart home related to the physiology to run simultaneously;

[0075] Step 75: When it is judged from the sleep disorder monitoring results that there are multiple people in the monitoring area and only the person with sleep disorder has not fallen asleep, continue to judge the correlation degree θ1 between the falling asleep situation and the environmental data and the correlation degree θ2 between the falling asleep situation and the physiological data; if θ1 > θ2 + 0.2, go to step 76; if θ1 < θ2 - 0.2, go to step 73; if |θ1 - θ2| ≤ 0.2, go to step 74;

[0076] Step 76: The host computer only controls the operation of the smart home related to the environment, and the smart home is limited to one or more of the air conditioner, air humidifier, air purifier, and noise shield;

[0077] Step 77: Control the smart home related to the environment and the smart home related to the physiology to run simultaneously, and the smart home is limited to one or more of the air conditioner, air humidifier, air purifier, and noise shield;

[0078] Step 78: When it is judged from the sleep disorder monitoring results that the person with sleep disorder has fallen asleep for the preset time, control the smart home devices in steps 72 to 77 to turn off.

[0079] Based on step 7, in order to enrich and improve the cloud database, make the sleep disorder and regulation strategies more accurate, and better serve more people with sleep disorders through big data, steps 8 and 9 are also included, as follows:

[0080] Step 8: Monitor the execution of the current regulation strategy during the monitoring period. If the execution effect is good, proceed to step 8. If the execution effect is not good, return to step 6 and continue to query other regulation schemes in the cloud database according to the matching degree, and the other regulation schemes are different from the current one;

[0081] The monitoring period can be 3 days or one week, which is entered by the person with sleep disorder through the lower computer;

[0082] Step 9: Upload and archive the current regulation scheme to the cloud database through the upper computer.

[0083] In summary, a sleep disorder monitoring and regulation system of the present invention takes the upper computer as the data processing and control core, integrates a handheld intelligent terminal, a smart bracelet, a monitoring unit, an environmental monitoring unit, and a cloud database. By judging the situation of people in the monitoring area and the sleep situation of people with sleep disorders, and combining environmental data and physiological data to judge whether the people with sleep disorders are really asleep. If not, whether the insomnia of the people with sleep disorders is caused by environmental factors or other psychological and pathological factors. If the upper computer identifies that the person with sleep disorder has not fallen asleep and there is only one person with sleep disorder according to the monitoring unit and the smart bracelet, and the environmental monitoring unit detects that the environmental data exceeds the comfort value for normal people to sleep, turn on the smart home to assist sleep; when it is detected that the person with sleep disorder has fallen asleep, control the smart home to turn off. This intelligent adjustment method not only helps sleep but also is conducive to energy conservation. If the upper computer identifies that the person with sleep disorder has not fallen asleep and other people have fallen asleep according to the monitoring unit and the smart bracelet, and the environmental data all meet the standards, the upper computer will only turn on the smart home without sound / lighting external playback to assist the person with sleep disorder to fall asleep. At this time, without affecting the normal sleep of others, the sleep disorder monitoring and regulation system can assist the person with sleep disorder to fall asleep quickly and achieve intelligent adjustment and control.

[0084] The above embodiments are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various deformations and optimization methods. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A sleep disorder monitoring and regulation system, at least comprising an information collection unit, a host computer, a cloud, and a smart home terminal, characterized in that: The information collection unit is communicatively connected to the input end of the host computer. The host computer is in two-way communicative connection with the cloud. The output end of the host computer is communicatively connected to the smart home end. The host computer receives the data uploaded by the information collection unit and the cloud storage data queried, performs deep learning and judgment, and issues a sleep monitoring and regulation control instruction to control the operating state of the smart home and assist sleep disorder patients in falling asleep.

2. The sleep disorder monitoring and regulation system according to claim 1, characterized in that: The information collection unit includes a slave computer, a monitoring unit, an environmental monitoring unit, and a smart bracelet. The information collection unit is used to upload input data, image data, environmental data, and the physiological data of sleep disorder patients to the host computer. The slave computer is used for sleep disorder patients to manually input data. The monitoring unit is used to collect image data within the monitoring range in real time, identify the number of people and the sleep state of the people. The environmental monitoring unit is used to monitor environmental data. The smart bracelet monitors the physiological data of sleep disorder patients in real time.

3. The sleep disorder monitoring and regulation system according to claim 2, characterized in that: The input data includes, but is not limited to, the identity information of sleep disorder patients, work information (work type, working hours), and medical examination information. The environmental data includes the temperature H1, humidity H2, air cleanliness H3, and noise level H4 in the monitoring area. The physiological data includes blood pressure X1, blood oxygen level X2, heart rate X3, respiratory rate X4, pulse frequency X5, and acceleration X6.

4. The sleep disorder monitoring and regulation system according to claim 1, wherein: The host computer is responsible for performing data preprocessing, data fusion, and deep learning training on the data uploaded by the information collection unit, outputting the sleep disorder monitoring results, and querying the regulation plan with the highest matching degree from the cloud database as the regulation strategy to control the operating state of the smart home.

5. A sleep disorder monitoring and regulation system according to claim 4, characterized in that: The sleep disorder monitoring results include: the number of people in the monitoring area, the pupil recognition result, the sleeping posture, the correlation degree θ1 between the sleep onset situation and the environmental data, and the correlation degree θ2 between the sleep onset situation and the physiological data.

6. The sleep disorder monitoring and regulation system according to claim 1, characterized in that: The smart home includes two categories. One category is the smart home related to physiology, such as earphones and massage pillows. The other category is the smart home related to the environment, such as lighting atmosphere lights, music playing devices, air conditioners, air humidifiers, air purifiers, and noise shields.

7. A method for monitoring and regulating sleep disorders, characterized in that: The method includes the following steps: Step 1: Input data: Sleep disorder patients fill in and input data through the slave computer APP. The input data is stored in the host computer. At the same time, the host computer can encrypt or perform secondary processing on the input data and then upload it to the cloud for storage. Step 2: Input of monitoring data: Obtain the image data in the monitoring area through the monitoring unit, obtain the environmental data through the environmental monitoring unit, and obtain the physiological data of sleep disorder patients through the smart bracelet. Step 3: Data preprocessing and data fusion: The host computer performs preprocessing and data fusion on the image data, environmental data, and physiological data to obtain a target parameter set, and randomly divides the target parameter set into a 70% training package and a 30% test set. Step 4: Convolution Training: Input the training set into the Transformer model, add positional encoding and masks, perform encoding-decoding operations, and obtain the first training set through the self-attention structure in the Transformer model. On this basis, import the first training set into the neural network for in-depth learning training, and obtain the sleep disorder monitoring result through classification; Step 5: Testing and Optimization: Input the test set into the objective function test model for testing, calculate the objective function, and adjust the weights of the convolutional layer of the neural network through the objective function; Step 6: Result Output and Regulation Strategy: When the neural network training meets the classification accuracy requirements, output the sleep disorder monitoring result. The host computer queries the regulation plan in the cloud database according to the sleep disorder monitoring result, and selects the regulation plan with the highest matching degree as the regulation strategy; The sleep disorder monitoring result includes: the number of people in the monitoring area, the pupil recognition result, the sleeping posture, the correlation degree θ1 between the falling asleep situation and the environmental data, and the correlation degree θ2 between the falling asleep situation and the physiological data. Step 7: Smart Home Control: According to the regulation strategy, formulate regulation parameters, generate control instructions to control the running state of the smart home, and perform regulation when sleep disorder patients encounter sleep disorders.

8. A method for monitoring and regulating sleep disorders according to claim 7, characterized in that: After step 7, there are also steps 8 and 9: Step 8: Monitor the execution situation of the current regulation strategy within the monitoring period. If the execution effect is good, enter step 8. If the execution effect is not good, return to step 6, and continue to query other regulation plans in the cloud database according to the matching degree. And the other regulation plan is different from the current regulation plan; Step 9: Upload and archive the current regulation plan to the cloud database through the host computer.

9. A method for monitoring and regulating sleep disorders according to claim 7, characterized in that: The specific content of step 5 is: Input the test set into the objective function test model for testing, and the corresponding objective function Obj() is shown in the following formula (1): Where: N represents the amount of test data imported into the pooling layer, K represents the amount of query data imported into the pooling layer, and the first term L() on the right side of the equation represents the prediction function, which determines the prediction accuracy of sleep disorders by solving the difference between the true value y i and the predicted value The second term β(f k ) represents the matching result queried through the cloud database. The specific calculation formula of β(f k ) is as follows: β(f t k ) = ρ X，F，t [X(m), F(p)] + ρ H，F,t [H(n), F(p)] + 0.5(2) Wherein, X(m) represents the m physiological data sets, the value range of m is from 1 to 6, H(n) represents the n environmental data sets, the value range of n is from 1 to 4, F(p) represents the image parameter set, and p represents the number of pixels; ρ represents the correlation coefficient between two parameters, the subscript "t" and the superscript "t" both represent time, and f k represents the correlation function; Use the proximity algorithm to calculate the predicted value, and L() is expressed as the following formula: To solve the objective function Obj(), assume that the predicted value at the initial moment is equal to the true value y1. Substitute equations (2) and (3) into equation (1) above, and solve the objective function Obj() through repeated iteration. According to the objective function Obj(), adjust the weights of the convolutional layer in the neural network.

10. A sleep disorder monitoring and regulation method according to claim 9, characterized in that: In order to verify the accuracy of the prediction of the objective function Obj() function, the residual index is used for evaluation. The error index includes the mean absolute error (MAE) and the mean absolute percentage error (MAPE). The corresponding calculation formulas are as follows: When both of the two residual indexes exceed 95%, stop the calculation and output the objective function Obj().

Images