A sleep regulation method based on magnetic field technology

Abstract

The present application provides a sleep regulation method based on magnetic field technology, including: constructing a user sleep stage judgment model according to the sleeping time set by the user and the current flight status, and judging the user's current sleep stage; obtaining the best magnetic field strength and sound wave frequency combination in different sleep stages, and dynamically adjusting the magnetic field and sound wave parameters; dynamically adjusting the parameters of the sound wave generator according to the noise suppression strategy database, and generating sound waves that match the noise frequency to offset the noise; obtaining the user's physiological parameters and magnetic field and sound wave frequency data, and predicting the optimal magnetic field strength and sound wave frequency parameter values ​​in different sleep stages; determining the expected landing and dining time according to the user's flight information and electronic schedule, performing brain wave wakefulness detection, and adjusting the magnetic field and sound wave parameters; establishing a user sleep quality prediction model, judging whether the user has sleep quality problems in a preset time period in the future, and determining the optimal magnetic field strength and sound wave frequency adjustment strategy.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a sleep regulation method based on magnetic field technology. Background Art

[0002] With the acceleration of globalization and the development of modern transportation, long-distance flights have become a part of life and work for many people. However, long flights are often accompanied by various sleep problems, which seriously affect the health and quality of life of passengers. On the plane, the main problems faced by passengers include environmental noise, insufficient seat comfort, and biological clock disorders caused by time zone changes. These factors together lead to generally poor sleep quality during flights. Environmental noise in the aircraft cabin is an important factor affecting sleep quality. The noise of aircraft engines, air conditioning systems, and passenger activities have caused serious interference to passengers' sleep. Although some airlines have begun to use noise reduction technology in the cabin, the effect is still limited and it is impossible to completely eliminate the impact of noise on sleep. Seat comfort during long-distance flights is also an urgent problem to be solved. The small seat space in the economy class makes it difficult for passengers to find a comfortable sleeping posture, which not only affects the quality of sleep, but also may cause muscle soreness and fatigue. In addition, frequent seat vibrations during flight can also interfere with sleep, affecting passengers' deep sleep and rapid eye movement sleep stages. The circadian rhythm disorder caused by flying across time zones is also an important factor affecting sleep quality. The human body's circadian rhythm is affected by many factors such as light, temperature, and diet. Long-term flights break these rules, causing passengers to face severe jet lag after landing. Jet lag not only affects passengers' sleep, but may also cause daytime sleepiness, lack of concentration, and decreased work efficiency. Although traditional sleep adjustment methods, such as providing eye masks, earplugs, neck pillows, etc., can alleviate these problems to a certain extent, the effect is limited and cannot fundamentally solve the sleep problem in flight. Moreover, these methods cannot be dynamically adjusted according to the individual differences and real-time status of passengers, resulting in many passengers still having difficulty getting high-quality sleep on the plane. In recent years, the application of EEG technology and artificial intelligence technology in sleep research has become more and more extensive, providing new ideas to solve these problems. However, how to apply these advanced technologies to sleep adjustment in flight is still a complex technical problem. Current research and applications mainly focus on improving the static sleeping environment, lacking dynamic monitoring and adjustment of the real-time status of passengers. Therefore, how to provide a quiet, comfortable sleeping environment that can dynamically adapt to the individual differences and real-time status of passengers during flight is still a technical problem that needs to be solved urgently. Summary of the invention

[0003] The present invention provides a sleep regulation method based on magnetic field technology, which mainly includes:

[0004] According to the user's set sleeping time and current flight status, combined with the time zone information of the departure and destination, the user's brain waves are detected in real time through the EEG sensor, the fast Fourier transform algorithm is used to perform frequency domain analysis on the brain wave signal, and a user sleep stage judgment model is constructed to judge the user's current sleep stage;

[0005] According to the user's current sleep stage information and destination time zone information, the optimal magnetic field strength and sound wave frequency combination for the sleep stage is obtained through the user's personalized sleep parameter database, and the magnetic field strength and frequency of the magnetic field generator built into the eye mask, as well as the sound wave frequency and amplitude of the sound wave generator, are dynamically adjusted;

[0006] According to the real-time acquired ambient noise audio signal and the user's sleep state, the parameters of the sound wave generator are dynamically adjusted in combination with the noise suppression strategy database to generate sound waves that match the noise frequency to offset the noise;

[0007] The user's physiological parameters, magnetic field and sound wave frequency data are obtained through the user's historical sleep monitoring database, and the optimal magnetic field strength and sound wave frequency parameter values ​​under different sleep stages are predicted using a recurrent neural network model, and updated to the user's personalized sleep parameter database;

[0008] Determine the estimated landing and meal time based on the user's flight information and electronic schedule, perform brain wave wakefulness detection, and adjust the magnetic field and sound wave parameters based on the user's wakefulness, continuously monitoring and optimizing the user's sleeping environment;

[0009] Based on the user's historical sleep data and current sleep status, a long short-term memory neural network algorithm is used to establish a user sleep quality prediction model to determine whether the user has sleep quality problems in the future preset time period, and determine the optimal magnetic field strength and sound wave frequency adjustment strategy based on the reinforcement learning algorithm.

[0010] Furthermore, according to the sleeping time set by the user and the current flight status, combined with the time zone information of the departure and destination, the user's brain waves are detected in real time through the electroencephalogram sensor, the brain wave signal is analyzed in the frequency domain using the fast Fourier transform algorithm, and a user sleep stage judgment model is constructed to judge the user's current sleep stage, including:

[0011] Obtain the user's set sleep time, current flight status, and time zone information of the departure and destination. The current flight status includes flight altitude, speed, and in-flight environment. Calculate the time zone difference between the departure and destination based on the time zone information of the departure and destination, and estimate the user's current biological clock deviation. Obtain the user's brain wave signal in real time through the EEG sensor, apply bandpass filter and wavelet transform to pre-process the brain wave signal to improve the quality of the brain wave signal. The pre-processing includes removing low-frequency interference, high-frequency noise, and environmental noise. Use the fast Fourier transform algorithm to convert the time domain brain wave signal into a frequency domain signal to obtain the amplitude and phase information of different frequency components. Extract key features from the frequency domain signal, including the frequency and amplitude of the main frequency components. The main frequency components include α wave, β wave, θ wave, and δ wave. According to the preset sampling rate and window length, the frequency and amplitude characteristics of brain waves are extracted and optimized; the historical user brain wave signal data is obtained through the user's historical sleep monitoring database, and the historical user brain wave signal data is labeled with sleep stages; based on the historical user brain wave signal data, the fast Fourier transform algorithm is used to obtain the amplitude and phase information of different frequency components, and key features are extracted from the frequency domain signal, and the model is trained using the random forest algorithm to build a user sleep stage judgment model; based on the user's current brain wave frequency domain characteristics, the user's sleep stage is judged using the user's sleep stage judgment model, and the sleep stages include light sleep, deep sleep or rapid eye movement; the user's sleep stage judgment result is compared with the sleep state estimated based on the time zone information and the time of falling asleep, to evaluate whether the user's sleep state meets expectations.

[0012] Furthermore, according to the user's current sleep stage information and the destination time zone information, the optimal magnetic field strength and sound wave frequency combination in the sleep stage is obtained from the user's personalized sleep parameter database, and the magnetic field strength and frequency of the magnetic field generator built into the eye mask, as well as the sound wave frequency and amplitude of the sound wave generator are dynamically adjusted, including:

[0013] Through the user's historical sleep monitoring database, the historical user's sleep stage information and the magnetic field strength and sound wave frequency combination parameters under the sleep stage are obtained, and a user's personalized sleep quality prediction model is built using a deep neural network to predict the optimal magnetic field strength and sound wave frequency combination parameters under different sleep stages of the user; based on the user's current sleep stage information detected in real time by the electroencephalogram sensor, the user's personalized sleep quality prediction model is used to obtain the optimal magnetic field strength and sound wave frequency combination parameters under the sleep stage; the magnetic field strength and sound wave frequency parameters predicted by the user's personalized sleep quality prediction model are input into the magnetic field generator and sound wave generator built into the eye mask, and the magnetic field generator is controlled to adjust the coil current according to the input parameters to generate a sound wave with the optimal magnetic field strength The system generates a magnetic field that matches the user's degree of sleep, controls the sound wave generator to adjust the amplitude and frequency according to the input parameters, and generates sound waves that match the optimal audio frequency; monitors the user's physiological signals in real time, including heart rate, breathing rate, and body movement times, and compares them with the user's personal normal range; uses the destination time zone information to calculate the day and night time corresponding to the current moment at the destination, and obtains the optimal magnetic field and sound wave parameter values ​​for each sleep stage of the user in the destination time zone from the user's personalized sleep parameter database; sets a time window for gradual adjustment based on the time difference between the current moment and the preset arrival time, and gradually adjusts the magnetic field and sound wave parameters to the optimal value that matches the destination's circadian rhythm based on the time difference, and continuously monitors the user's physiological signal feedback during the adjustment process.

[0014] Furthermore, the parameters of the sound wave generator are dynamically adjusted according to the real-time acquired ambient noise audio signal and the user's sleep state in combination with the noise suppression strategy database to generate a sound wave that matches the noise frequency to cancel the noise, including:

[0015] By placing multiple MEMS digital noise sensors at different locations in the aircraft cabin, the noise audio signal is acquired in real time and preprocessed, including removing DC bias, removing low-frequency noise, and removing high-frequency noise; the preprocessed noise audio signal is analyzed in the frequency domain using the fast Fourier transform algorithm to obtain the frequency spectrum distribution of the noise and determine the main noise frequency components and their corresponding amplitudes; the frequency spectrum data of the noise is combined with the user's current sleep state, and the optimal audio signal processing strategy and sound wave generator parameter combination are selected from the noise suppression strategy database; according to the selected audio signal processing strategy, the noise audio signal is filtered and attenuated to obtain the processed noise audio signal; according to the selected sound wave generator parameters, the sound wave is controlled The generator generates a specified sound wave with the same frequency, amplitude and phase difference of 180° as the processed noise signal; the specified sound wave generated by the sound wave generator is superimposed on the actual noise in the aircraft cabin by using the principle of sound wave interference, and the noise is canceled by destructive interference; a reference microphone is set around the user's ear to obtain an audio signal mixed with anti-noise and residual noise, and the audio signal obtained by the reference microphone is adaptively filtered using the LMS algorithm. The filter coefficient is estimated based on the reference signal and the error signal, and the filter coefficient is iteratively updated using the gradient descent method until the mean square value of the error signal is minimized, where the reference signal is the anti-noise and the error signal is the signal obtained by the reference microphone; the sound wave parameters are continuously iterated and optimized until the residual noise meets the comfort requirements.

[0016] It also includes dynamically adjusting the sound wave generator parameters to optimize noise suppression and predict changes in sleep quality based on real-time environmental noise audio data and user sleep status.

[0017] The method of dynamically adjusting the sound wave generator parameters to optimize noise suppression and predicting changes in sleep quality based on the real-time acquired ambient noise audio data and the user's sleep state specifically includes:

[0018] Audio acquisition equipment is used to acquire the ambient noise in the aircraft cabin in real time to obtain the original noise audio data; the short-time Fourier transform algorithm is used to convert the original noise audio data into a spectrum representation to obtain noise feature data; the user's sleep state data, including sleep stage, brain wave frequency and heart rate, is acquired in real time through the portable EEG headset and smart bracelet worn by the user. The brain wave frequencies include delta waves, theta waves, alpha waves and beta waves; the acquired sleep state data is synchronized with the noise feature data to construct a noise-sleep state mapping data set; based on the noise-sleep state mapping data set, the long short-term memory network algorithm is used for model training to build a sleep quality prediction model to predict the impact score of noise on sleep quality; the noise characteristics and current sleep state data during the flight are acquired in real time, Use the sleep quality prediction model to predict the impact score of noise on sleep quality in a preset time period in the future; if the predicted impact score is lower than the preset threshold, it is necessary to further optimize the noise reduction measures; according to the prediction results, adaptively adjust the parameters of the sound wave generator, including the sound wave frequency, amplitude and phase; minimize the difference between the predicted score and the actual score through the gradient descent algorithm, solve the optimal combination of sound wave parameters, and minimize the impact of noise; input the optimized parameters into the sound wave generator to generate sound waves with the same frequency, equal amplitude and opposite phase as the original noise, and achieve noise reduction effect after superposition with environmental noise through the principle of sound wave interference; continuously monitor the changes in the user's sleep state and environmental noise, dynamically update the noise-sleep state mapping data set, and regularly retrain the sleep quality prediction model and optimize the sound wave parameters.

[0019] Furthermore, the user's physiological parameters and magnetic field and sound wave frequency data are obtained through the user's historical sleep monitoring database, and the optimal magnetic field strength and sound wave frequency parameter values ​​under different sleep stages are predicted using a recurrent neural network model, and updated to the user's personalized sleep parameter database, including:

[0020] Through the user's historical sleep monitoring database, the user's physiological parameters per minute, as well as the magnetic field strength and sound wave frequency parameters at the corresponding moment, are extracted. The physiological parameters include brain wave characteristics, heart rate, and breathing rate; the electroencephalogram device is used to obtain the user's EEG signal during sleep, and the EEG signal is filtered at 50Hz to remove power frequency interference. The EEG signal is decomposed into 4 layers using wavelet transform to extract the energy characteristics of the four frequency bands of δ, θ, α, and β; the piezoelectric sensor is used to obtain the vibration signal of the user's mattress, the breathing frequency characteristics are extracted, and the photoelectric volume pulse wave sensor is used to obtain the user's heart rate signal; the acquired physiological signal characteristics are compared with the current The magnetic field strength and sound wave frequency parameter values ​​are associated to construct training data samples; the training data is preprocessed, including data cleaning and normalization; based on the training data samples, a recurrent neural network algorithm is used to train the model to predict the optimal magnetic field strength and sound wave frequency parameter values ​​under different sleep stages; the optimal magnetic field strength and sound wave frequency parameter values ​​predicted by the recurrent neural network model are updated to the user's personalized sleep parameter database, and stored in association with the user's historical sleep data; the recurrent neural network model is retrained and optimized regularly, and the optimal magnetic field strength and sound wave frequency parameter values ​​in the user's personalized sleep parameter database are updated.

[0021] It also includes dynamically adjusting the parameters of the magnetic field generator and sound wave generator based on the optimal magnetic field strength and sound wave frequency parameter values ​​in different sleep stages of the user, combined with real-time sleep monitoring data, and calculating the user's sleep quality score.

[0022] According to the optimal magnetic field strength and sound wave frequency parameter values ​​in different sleep stages of the user, the magnetic field generator and sound wave generator parameters are dynamically adjusted in combination with real-time sleep monitoring data, and the user's sleep quality score is calculated, including:

[0023] Obtain the historical optimal magnetic field strength and sound wave frequency parameter values ​​of the user in different sleep stages from the user's personalized sleep parameter database; use the user's sleep stage judgment model to judge the user's current sleep stage based on the user's real-time sleep data obtained by the sleep monitoring device, and obtain the optimal magnetic field strength and sound wave frequency parameter range corresponding to the sleep stage; when the user enters the deep sleep stage, the magnetic field strength is gradually increased from the current value to the upper limit of the optimal magnetic field strength parameter range by controlling the output current of the magnetic field generator; if the current value of the magnetic field strength is lower than the lower limit of the optimal magnetic field strength parameter range, the magnetic field strength is first adjusted to the lower limit, and then gradually increased to the upper limit; if in the process of increasing the magnetic field strength, the delta wave energy proportion does not show an upward trend in multiple consecutive adjustment cycles, the adjustment is stopped and maintained at the current value; when the user is in the light sleep or rapid eye movement sleep stage, the sound wave frequency is maintained at the middle value of the optimal sound wave frequency parameter range by controlling the output frequency of the sound wave generator, and the sound wave frequency deviation does not exceed If the actual frequency exceeds 5%, the calculation method of the sound wave frequency deviation is sound wave frequency deviation = |(actual frequency-optimal frequency middle value) / optimal frequency middle value|; dynamically adjust the sound wave amplitude according to the changes in the user's heart rate and breathing rate. If the instantaneous value of the heart rate or breathing rate exceeds the normal range when adjusting the sound wave amplitude, the adjustment is suspended until the physiological indicators return to normal. The physiological indicator parameters include the energy proportion of δ waves, θ waves, α waves and β waves, the mean and coefficient of variation of the heart rate, and the mean and coefficient of variation of the breathing rate; during the sleep process, continuously monitor the user's physiological indicator parameters as sleep quality evaluation indicators, and use a weighted average algorithm to comprehensively score the sleep quality evaluation indicators to obtain the user's current sleep quality score, wherein the weight coefficient of the sleep quality evaluation indicator is determined according to its correlation coefficient with the overall sleep quality; after the user finishes sleeping, the optimal magnetic field strength, sound wave frequency and amplitude parameters of each sleep stage recorded during the sleep process are entered into the user's personalized sleep parameter database, and merged with the user's historical sleep parameter data.

[0024] Furthermore, the estimated landing and dining time is determined according to the user's flight information and electronic schedule, the brain wave awakening level is detected, and the magnetic field and sound wave parameters are adjusted based on the user's awakening level, and the user's sleeping environment is continuously monitored and optimized, including:

[0025] Determine the user's estimated landing time and meal time through the user's flight information and electronic schedule; extract the flight number, take-off time and landing time from the user's electronic ticket, and combine the real-time dynamic data of the flight to estimate the user's actual landing time; identify dining arrangements with dining keywords from the user's electronic calendar and extract meal times, where dining keywords include dining, dining, lunch and dinner; if the user has dining arrangements before takeoff, perform an OR operation on the two conditions of landing time and meal time, and use the earlier meal time as the trigger condition for starting brain wave wakefulness detection; if the trigger condition is met, use the built-in dry electrode EEG sensor to obtain the user's single-lead EEG signal in real time, and use the Mallat algorithm to perform wavelet decomposition on the EEG signal, square and sum the wavelet coefficients of each layer to obtain the wavelet coefficient energy value of the frequency band, including δ waves, θ waves, α waves, and β waves; normalize the wavelet coefficient energy value The relative energy distribution of each frequency band is obtained as the time-frequency domain feature of the brain wave; the brain wave data of the user at different levels of wakefulness are obtained through the user's historical sleep monitoring database, and the decision tree algorithm is used for model training to build a wakefulness discrimination model, where the wakefulness levels include wakefulness, sleepiness, and micro-sleep; according to the extracted brain wave time-frequency domain features, the wakefulness discrimination model is used to judge the user's wakefulness; if the output result of the wakefulness discrimination model is the awake state, the parameter adjustment mechanism of the magnetic field generator and the sound wave generator is triggered, and the changes in the user's brain wave characteristics are continuously monitored to judge the user's wakefulness; if the output results of the wakefulness discrimination model are maintained in the awake state within a continuous preset time window, the speed of parameter adjustment of the magnetic field generator and the sound wave generator is accelerated to shorten the time for the device to be turned off; after the parameter adjustment of the magnetic field generator and the sound wave generator is completed, the magnetic field generator and the sound wave generator are turned off to stop interfering with the user's sleeping environment.

[0026] Furthermore, the user's sleep quality prediction model is established based on the user's historical sleep data and current sleep state using a long short-term memory neural network algorithm to determine whether the user has sleep quality problems in a future preset time period, and the optimal magnetic field strength and sound wave frequency adjustment strategy is determined based on a reinforcement learning algorithm, including:

[0027] The user's historical sleep data is obtained through the user's historical sleep monitoring database, and the sleep data includes the time of falling asleep, sleep duration, ratio of deep and light sleep, rapid eye movement sleep duration and number of awakenings during sleep, as well as the sleep quality score of the day; the historical sleep data is arranged in chronological order to construct a time series data set, and the historical sleep data is normalized; based on the time series data set of the historical sleep data, a long short-term memory neural network is used for model training to construct a sleep quality score prediction model to predict the sleep quality score within a preset time period in the future; based on the user's real-time sleep data of the day, the sleep quality score prediction model is used to predict the sleep quality score within a preset time period in the future; if the predicted score within the preset time period in the future shows a downward trend and the lowest score is lower than the preset score threshold, it is judged that the user has a sleep quality problem; based on the prediction results of the sleep quality score prediction model, the optimal magnetic field strength and sound wave frequency adjustment strategy is obtained through the policy gradient algorithm in reinforcement learning; the user's sleep data is continuously monitored, the training data of the sleep quality score prediction model is regularly updated, and the model is retrained and optimized.

[0028] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0029] The invention discloses a sleep adjustment method based on magnetic field technology. By real-time detection and analysis of the user's brain wave signal, combined with the user's personalized sleep parameter database, the magnetic field strength and sound wave frequency can be dynamically adjusted, thereby significantly improving the user's sleep quality at different sleep stages. The invention can accurately predict and adjust the optimal magnetic field strength and sound wave frequency at different sleep stages, realize personalized sleep adjustment strategies, and meet the different sleep needs of users in flight. The invention also dynamically adjusts the parameters of the sound wave generator through the real-time acquired environmental noise audio signal, combined with the noise suppression strategy database, generates sound waves matching the noise frequency, can effectively offset the environmental noise, and further improve the user's sleep environment quality. By combining the user's flight information and electronic schedule, the user's brain wave awakening degree can be dynamically detected, and the magnetic field and sound wave parameters can be adjusted based on the user's awakening degree to ensure that the user wakes up at the right time and optimize the user's sleep and awakening process. Through the established sleep quality prediction system, the invention can judge in advance the sleep quality problems that may occur to the user in the future preset time period, and determine the optimal magnetic field strength and sound wave frequency adjustment strategy, timely intervene and adjust, and prevent and solve sleep problems. The present invention provides a comprehensive, dynamic and personalized sleep adjustment solution through the comprehensive application of multiple advanced technologies, which significantly improves the user's sleep quality and flight experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 The present invention is a flow chart of a sleep regulation method based on magnetic field technology.

[0031] Figure 2 It is a schematic diagram of a sleep regulation method based on magnetic field technology of the present invention.

[0032] Figure 3 This is another schematic diagram of a sleep regulation method based on magnetic field technology according to the present invention. DETAILED DESCRIPTION

[0033] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0034] like Figure 1-3 In this embodiment, a sleep adjustment method based on magnetic field technology may specifically include:

[0035] S101. Based on the sleeping time set by the user and the current flight status, combined with the time zone information of the departure and destination, the user's brain waves are detected in real time through the electroencephalogram sensor, the brain wave signal is analyzed in the frequency domain using the fast Fourier transform algorithm, and a user sleep stage judgment model is constructed to judge the user's current sleep stage.

[0036] Obtain the user's set sleep time, current flight status, and time zone information of the departure and destination. The current flight status includes flight altitude, speed, and in-flight environment. According to the time zone information of the departure and destination, calculate the time zone difference between the departure and destination, and estimate the user's current biological clock deviation. Obtain the user's brain wave signal in real time through the EEG sensor, apply bandpass filter and wavelet transform to preprocess the brain wave signal to improve the quality of the brain wave signal. The preprocessing includes removing low-frequency interference, high-frequency noise and environmental noise. Use the fast Fourier transform algorithm to convert the time domain brain wave signal into a frequency domain signal to obtain the amplitude and phase information of different frequency components. Extract key features from the frequency domain signal, including the frequency and amplitude of the main frequency components, which include α wave, β wave, θ wave, and δ wave. Extract and optimize the frequency and amplitude characteristics of the brain wave according to the preset sampling rate and window length. Obtain historical user brain wave signal data through the user's historical sleep monitoring database, and mark the sleep stage of the historical user brain wave signal data. Based on the historical user brain wave signal data, the fast Fourier transform algorithm is used to obtain the amplitude and phase information of different frequency components, and the key features are extracted from the frequency domain signal. The random forest algorithm is used for model training to build a user sleep stage judgment model. Based on the user's current brain wave frequency domain characteristics, the user sleep stage judgment model is used to judge the user's sleep stage, which includes light sleep, deep sleep, or rapid eye movement. The user's sleep stage judgment result is compared with the sleep state estimated based on the time zone information and the time of falling asleep to evaluate whether the user's sleep state meets expectations.

[0037] Specifically, the user's set sleeping time is 22:00, the current flight status is a flight altitude of 10,000 meters, a speed of 800 kilometers per hour, and an in-flight ambient temperature of 22 degrees Celsius. The departure place is city A, the time zone is UTC-5, and the destination is city B, the time zone is UTC+0. According to the time zone information of the departure place and the destination, the time zone difference is calculated to be 5 hours, so the user's biological clock may be offset by 5 hours. The user's brain wave signal is obtained in real time through the electroencephalogram sensor, and a bandpass filter is applied to remove low-frequency interference, high-frequency noise and environmental noise, and the signal quality is improved by wavelet transform. The fast Fourier transform algorithm is used to convert the brain wave signal in the time domain into a frequency domain signal to obtain the amplitude and phase information of different frequency components. The frequency and amplitude of the main frequency components are extracted from the frequency domain signal, and the main frequency components include α wave, β wave, θ wave, and δ wave. According to the preset sampling rate of 256Hz and the window length of 2 seconds, the frequency and amplitude characteristics of the brain wave are extracted and optimized. The brain wave signal data of the past year is obtained through the user's historical sleep monitoring database, and the data is labeled with sleep stages. The fast Fourier transform algorithm is used to analyze the amplitude and phase information of different frequency components in the historical data, extract key features, and use these features to train the model through the random forest algorithm to build a user sleep stage judgment model. According to the current brain wave frequency domain characteristics, the judgment model determines that the user is in deep sleep, and compares this judgment result with the sleep state estimated based on the time zone information and the time of falling asleep to confirm that the user's current sleep state meets expectations.

[0038] S102. According to the user's current sleep stage information and the destination time zone information, the optimal magnetic field strength and sound wave frequency combination for the sleep stage is obtained from the user's personalized sleep parameter database, and the magnetic field strength and frequency of the magnetic field generator built into the eye mask, as well as the sound wave frequency and amplitude of the sound wave generator are dynamically adjusted.

[0039] Through the user's historical sleep monitoring database, the historical user sleep stage information and the magnetic field strength and sound wave frequency combination parameters under the sleep stage are obtained, and the user's personalized sleep quality prediction model is constructed using a deep neural network to predict the optimal magnetic field strength and sound wave frequency combination parameters under different sleep stages of the user. According to the user's current sleep stage information detected in real time by the electroencephalogram sensor, the user's personalized sleep quality prediction model is used to obtain the optimal magnetic field strength and sound wave frequency combination parameters under the sleep stage. The magnetic field strength and sound wave frequency parameters predicted by the user's personalized sleep quality prediction model are input into the magnetic field generator and sound wave generator built into the eye mask, and the magnetic field generator is controlled to adjust the coil current according to the input parameters to generate a magnetic field matching the optimal magnetic field strength, and the sound wave generator is controlled to adjust the amplitude and frequency according to the input parameters to generate sound waves matching the optimal audio frequency. The user's physiological signals are monitored in real time, including heart rate, breathing rate, and body movement times, and compared with the user's personal normal range. The destination time zone information is used to calculate the day and night time corresponding to the current time at the destination, and the optimal magnetic field and sound wave parameter values ​​of the user in each sleep stage under the destination time zone are obtained from the user's personalized sleep parameter database. According to the time difference between the current moment and the preset arrival time, a time window for gradual adjustment is set. The magnetic field and sound wave parameters are gradually adjusted to the optimal value for matching the circadian rhythm of the destination according to the time difference. During the adjustment process, the user's physiological signal feedback is continuously monitored.

[0040] Specifically, 1000 hours of sleep data were obtained from 100 users, including brain wave features, such as the relative energy values ​​of delta waves, theta waves, alpha waves, and beta waves, magnetic field strength, sound wave frequency, and corresponding sleep quality scores. The data was randomly divided into training set, validation set, and test set in a ratio of 8:1:1. A deep neural network model was used with an input dimension of 10 (4 dimensions of brain wave features, 1 dimension of magnetic field strength, 1 dimension of sound wave frequency, and 4 time steps), 128 hidden layer neurons, and 1 output dimension. The mean square error was used as the loss function, Adam was used as the optimizer, the learning rate was 0.001, 500 epochs were trained, and the batch_size was 32. The model performance was evaluated on the test set. If the mean absolute error was less than 0.5, the model was considered usable. When the user falls asleep wearing an eye mask, the EEG sensor collects the user's brain wave signals at a sampling frequency of 500Hz, calculates the relative energy values ​​of the delta wave, theta wave, alpha wave, and beta wave every 30 seconds, and inputs them into the prediction model together with the magnetic field strength and sound wave frequency. If the user is currently in deep sleep, the EEG characteristics are [0.6, 0.2, 0.1, 0.1]. At this time, the optimal magnetic field strength and sound wave frequency combination parameters output by the model are [30mT, 5kHz]. The coil current of the magnetic field generator is controlled to be 1.5A to generate a 30mT magnetic field. At the same time, the amplitude of the sound wave generator is controlled to be 2V and the frequency is 5kHz, and the corresponding sound pressure level is about 40dB. Real-time monitoring of the user's physiological signals, including using photoplethysmography to obtain heart rate, pressure sensor to obtain breathing rate, and acceleration sensor to obtain body movement times. If the user's average heart rate is 110 beats / minute, breathing rate is 35 beats / minute, and body movement times are 25 times / hour for 5 consecutive minutes, it is judged that the sleep quality has decreased. At this time, according to the parameter change direction output by the prediction model, the magnetic field strength is reduced to 28mT, the sound wave frequency is reduced to 4.5kHz, and the user's physiological signals are continuously monitored until they return to the normal range. The destination time zone information is used to calculate the day and night time corresponding to the current time at the destination. If the destination time is 2 a.m., the optimal magnetic field and sound wave parameter values ​​for each sleep stage of the user in the destination time zone are obtained from the user's personalized sleep parameter database. If the current time is 4 hours away from the preset arrival time, a 4-hour progressive adjustment time window is set, and the magnetic field strength is gradually adjusted from the current 5 millitesla to 3 millitesla that matches the destination's circadian rhythm according to the time difference, and the sound wave frequency is adjusted from 100 Hz to 200 Hz. During this process, the user's physiological signal feedback is continuously monitored to ensure that the adjustment process has the least impact on the user's sleep quality.

[0041] S103. According to the real-time acquired ambient noise audio signal and the user's sleep state, the parameters of the sound wave generator are dynamically adjusted in combination with the noise suppression strategy database to generate sound waves that match the noise frequency to cancel out the noise.

[0042] By arranging multiple MEMS digital noise sensors at different positions in the aircraft cabin, the noise audio signal is acquired in real time, and the noise audio signal is preprocessed, including removing DC bias, removing low-frequency noise, and removing high-frequency noise. The preprocessed noise audio signal is analyzed in the frequency domain using the fast Fourier transform algorithm to obtain the frequency spectrum distribution of the noise, and determine the main noise frequency components and their corresponding amplitudes. The frequency spectrum data of the noise is combined with the current sleep state of the user, and the optimal audio signal processing strategy and sound wave generator parameter combination are selected from the noise suppression strategy database. According to the selected audio signal processing strategy, the noise audio signal is filtered and attenuated to obtain the processed noise audio signal. According to the selected sound wave generator parameters, the sound wave generator is controlled to generate a specified sound wave with the same frequency, equal amplitude, and a phase difference of 180° as the processed noise signal. Using the principle of sound wave interference, the specified sound wave generated by the sound wave generator is superimposed on the actual noise in the aircraft cabin, and the noise is offset by destructive interference. A reference microphone is set around the user's ear to obtain an audio signal mixed with anti-noise and residual noise. The LMS algorithm is used to adaptively filter the audio signal obtained by the reference microphone. The filter coefficient is estimated based on the reference signal and the error signal. The filter coefficient is iteratively updated using the gradient descent method until the mean square value of the error signal is minimized. The reference signal is the anti-noise signal and the error signal is the signal obtained by the reference microphone. The sound wave parameters are continuously iterated and optimized until the residual noise meets the comfort requirements.

[0043] Specifically, four MEMS digital noise sensors with a sensitivity of -40dB and a frequency response of 20Hz-20kHz are arranged in the aircraft cabin, with a sampling frequency of 44.1kHz and a bit depth of 16bit. The real-time acquired ambient noise audio signal is preprocessed, and the DC bias is removed by subtracting the mean of the signal. A 6th-order Butterworth high-pass filter is used to remove low-frequency noise below 30Hz, and a 12th-order Butterworth low-pass filter is used to remove high-frequency noise above 18kHz. The preprocessed noise audio signal is analyzed in the frequency domain using the fast Fourier transform algorithm to obtain the frequency spectrum distribution of the noise, and the main noise frequency component is determined to be 500Hz, with a corresponding amplitude of 70dB. Combined with the user's current sleep state, the optimal audio signal processing strategy and sound wave generator parameter combination are selected from the noise suppression strategy database. The noise suppression strategy database contains parameter combinations constructed through a large number of simulation experiments and user feedback data, such as filter type, order, cutoff frequency, etc., as well as sound wave generator parameters such as waveform, amplitude, phase, etc. For a given noise frequency distribution and user sleep state, the noise suppression strategy database stores multiple sets of audio signal processing parameters and sound wave generator parameter combinations. These parameter combinations are sorted by evaluating indicators such as noise attenuation and user comfort, and the combination with the highest comprehensive score is selected as the optimal strategy. According to the selected audio signal processing strategy, the noise audio signal is filtered and attenuated to obtain the processed noise audio signal. According to the selected sound wave generator parameters, the sound wave generator is controlled to generate a specified sound wave with the same frequency, equal amplitude, and a phase difference of 180° as the processed noise signal, that is, a 500Hz, 70dB anti-phase sound wave. Using the principle of sound wave interference, the specified sound wave generated by the sound wave generator is superimposed with the actual noise in the aircraft cabin, and noise cancellation is achieved through destructive interference. A reference microphone is set around the user's ear to obtain an audio signal mixed with anti-noise and residual noise. The LMS algorithm is used to perform adaptive filtering on the audio signal obtained by the reference microphone. The reference signal is the anti-noise, and the error signal is the signal obtained by the reference microphone. The filter coefficients are estimated and updated iteratively using the gradient descent method until the mean square value of the error signal is minimized. After multiple iterations to optimize the sound wave parameters, the residual noise was finally reduced to 30dB, meeting the user's comfort requirements.

[0044] Based on the real-time acquired ambient noise audio data and the user's sleep status, the sound wave generator parameters are dynamically adjusted to optimize noise suppression and predict changes in sleep quality.

[0045] The audio acquisition equipment is used to acquire the ambient noise in the aircraft cabin in real time to obtain the original noise audio data. The short-time Fourier transform algorithm is used to convert the original noise audio data into a spectrum representation to obtain the noise feature data. The user's sleep state data, including sleep stage, brain wave frequency and heart rate, is acquired in real time through the portable EEG headset and smart bracelet worn by the user. The brain wave frequency includes δ wave, θ wave, α wave and β wave. The acquired sleep state data is synchronized with the noise feature data to construct a noise-sleep state mapping data set. According to the noise-sleep state mapping data set, the long short-term memory network algorithm is used for model training to build a sleep quality prediction model to predict the impact score of noise on sleep quality. The noise characteristics and current sleep state data during the flight are acquired in real time, and the sleep quality prediction model is used to predict the impact score of noise on sleep quality in the future preset time period. If the predicted impact score is lower than the pre-threshold, the noise reduction measures need to be further optimized. According to the prediction results, the parameters of the sound wave generator are adaptively adjusted, including the sound wave frequency, amplitude and phase. The gradient descent algorithm is used to minimize the difference between the predicted score and the actual score, and the optimal combination of sound wave parameters is solved to minimize the impact of noise. The optimized parameters are input into the sound wave generator to generate sound waves with the same frequency, equal amplitude, and opposite phase as the original noise. Through the principle of sound wave interference, the noise reduction effect is achieved after superposition with the ambient noise. The user's sleep state changes and ambient noise changes are continuously monitored, the noise-sleep state mapping data set is dynamically updated, and the sleep quality prediction model and sound wave parameters are regularly retrained.

[0046] Specifically, four microphones are arranged in the aircraft cabin with a sampling rate of 44.1kHz and a bit depth of 16bit to collect environmental noise in real time. Through the short-time Fourier transform algorithm, these original noise audio data are converted into a spectrum diagram to obtain the characteristic data of the noise, among which the main frequency component is 600Hz and the amplitude is 75dB. When the user wears a portable EEG headset and a smart bracelet, the sleep state data is obtained in real time, including that the user is currently in a deep sleep stage, the EEG frequency shows that the δ wave is 0.5Hz, the θ wave is 4Hz, the α wave is 8Hz, the β wave is 16Hz, and the heart rate is 60 times / minute. These sleep state data are synchronized with the noise characteristic data to construct a noise-sleep state mapping data set. Based on the mapping data set, the long short-term memory network algorithm is used for model training to construct a sleep quality prediction model to predict the impact score of noise on sleep quality. If the noise characteristics acquired in real time during the flight are 600Hz and 75dB, and the sleep state data acquired at the same time is the light sleep stage, the brain wave frequency is theta wave 4Hz, and the heart rate is 65 times / minute. Using the trained sleep quality prediction model, the impact of noise on sleep quality in the next 30 minutes is predicted to be 45 points, which is lower than the set threshold of 50 points, indicating that noise may significantly affect the user's sleep quality. According to the prediction results, the parameters of the sound wave generator are adaptively adjusted, including the sound wave frequency of 600Hz, the amplitude of 75dB, and the opposite phase. The gradient descent algorithm is used to minimize the difference between the predicted score and the actual score, and the optimal combination of sound wave parameters is solved to minimize the impact of noise. The optimized parameters are input into the sound wave generator to generate sound waves with the same frequency, equal amplitude, and opposite phase as the original noise. Through the principle of sound wave interference, the noise reduction effect is achieved after superposition with the ambient noise. During the entire sleep process, the user's sleep state changes and environmental noise changes are continuously monitored, the noise-sleep state mapping dataset is dynamically updated, and the sleep quality prediction model is retrained and the sound wave parameters are optimized regularly to ensure that the noise reduction effect can adapt to different noise environments and changes in the user's sleep state, thereby maintaining the user's high-quality sleep.

[0047] S104. Obtain the user's physiological parameters and magnetic field and sound wave frequency data through the user's historical sleep monitoring database, use the recurrent neural network model to predict the optimal magnetic field strength and sound wave frequency parameter values ​​in different sleep stages, and update them to the user's personalized sleep parameter database.

[0048] Through the user's historical sleep monitoring database, the user's physiological parameters per minute, as well as the magnetic field strength and sound wave frequency parameters at the corresponding moment, are extracted. The physiological parameters include brain wave characteristics, heart rate, and respiratory rate. The electroencephalogram device is used to obtain the user's EEG signal during sleep, and the EEG signal is filtered at 50Hz to remove power frequency interference. The EEG signal is decomposed into 4 layers using wavelet transform to extract the energy characteristics of the four frequency bands of δ, θ, α, and β. The vibration signal of the user's mattress is obtained using a piezoelectric sensor to extract the respiratory frequency characteristics, and the user's heart rate signal is obtained using a photoelectric volume pulse wave sensor. The acquired physiological signal characteristics are associated with the current magnetic field strength and sound wave frequency parameter values ​​to construct a training data sample. The training data is preprocessed, including data cleaning and normalization. According to the training data samples, the recurrent neural network algorithm is used for model training to predict the optimal magnetic field strength and sound wave frequency parameter values ​​under different sleep stages. The optimal magnetic field strength and sound wave frequency parameter values ​​predicted by the recurrent neural network model are updated to the user's personalized sleep parameter database and stored in association with the user's historical sleep data. The recurrent neural network model is retrained and optimized regularly to update the optimal magnetic field strength and sound wave frequency parameter values ​​in the user's personalized sleep parameter database.

[0049] Specifically, the user's sleep monitoring data for 7 consecutive days is extracted from the user's historical sleep monitoring database, and the user's brain wave characteristics per minute are extracted, including the energy proportion of δ waves, the energy proportion of θ waves, the energy proportion of α waves, and the energy proportion of β waves. The calculation is performed using a 4th-order Butterworth filter and a 256-point fast Fourier transform. At the same time, the heart rate and respiratory rate at the corresponding moment, as well as environmental parameters such as magnetic field strength and sound wave frequency, are extracted. The acquired physiological signal characteristics are associated with the current parameter values ​​of the magnetic field strength of 3mT and the sound wave frequency of 100Hz to construct a training data sample. The training data sample includes physiological parameters, magnetic field strength, and sound wave frequency parameters. The physiological parameters include brain wave characteristics, heart rate, and respiratory rate. The training data is preprocessed, including data cleaning and normalization, and each feature is scaled to between 0 and 1. The normalized 7-day data is randomly divided into a training set, a validation set, and a test set with a data ratio of 7:1.5:1.5, totaling about 300,000 samples. According to the training data samples, the recurrent neural network algorithm is used to train the model, and the optimal magnetic field strength and sound wave frequency parameter values ​​under different sleep stages are predicted. The model hyperparameters are tuned on the validation set, and the classification accuracy is the highest when the penalty coefficient C = 10 and the kernel function parameter γ = 0.01, reaching 92.5%. The model is trained on the entire training set using the optimal parameters, and the performance is evaluated on the test set. The classification accuracy is 91.8%, indicating that the model has good generalization ability. The trained recurrent neural network model is deployed to the server side, and the brain wave and physiological characteristic data uploaded by the user in real time are received. The user's current sleep stage is predicted once a minute. If the user's δ wave energy accounts for 60%, the θ wave energy accounts for 20%, the α wave energy accounts for 15%, the β wave energy accounts for 5%, the heart rate is 55 times / min, and the breathing rate is 12 times / min, then the user is predicted to be in the deep sleep stage. According to the prediction results, the optimal magnetic field strength range and sound wave frequency range corresponding to the deep sleep stage are queried from the user's personalized sleep parameter database as [20-30] mT and [1-5] kHz. If the current magnetic field strength is 25mT and the sound wave frequency is 6kHz, the sound wave frequency is adjusted to 4kHz, while the magnetic field strength remains unchanged, and is output to the magnetic field generator and sound wave generator control interface. The adjusted parameter value (25mT, 4kHz) is updated to the optimal parameter record of the user's deep sleep stage. Whenever the user's nighttime sleep monitoring data accumulates to 7 days, the support vector machine model is retrained, the model parameters and the personalized sleep parameter database are updated, and the cycle continues to iterate to achieve self-optimization of the model and database.

[0050] According to the optimal magnetic field strength and sound wave frequency parameter values ​​in different sleep stages of the user, combined with real-time sleep monitoring data, the magnetic field generator and sound wave generator parameters are dynamically adjusted, and the user's sleep quality score is calculated.

[0051] The historical best magnetic field strength and sound wave frequency parameter values ​​of the user under different sleep stages are obtained from the user personalized sleep parameter database. According to the real-time sleep data of the user obtained by the sleep monitoring device, the user's current sleep stage is judged using the user sleep stage judgment model to obtain the best magnetic field strength and sound wave frequency parameter range corresponding to the sleep stage. When the user enters the deep sleep stage, the magnetic field strength is gradually increased from the current value to the upper limit of the optimal magnetic field strength parameter range by controlling the output current of the magnetic field generator. If the current value of the magnetic field strength is lower than the lower limit of the optimal magnetic field strength parameter range, the magnetic field strength is first adjusted to the lower limit, and then gradually increased to the upper limit. If the delta wave energy ratio does not show an upward trend in a plurality of consecutive adjustment cycles during the process of increasing the magnetic field strength, the adjustment is stopped and maintained at the current value. When the user is in the shallow sleep or rapid eye movement sleep stage, the sound wave frequency is maintained at the middle value of the optimal sound wave frequency parameter range by controlling the output frequency of the sound wave generator, and the sound wave frequency deviation does not exceed 5%. The sound wave frequency deviation is calculated as sound wave frequency deviation = | (actual frequency - optimal frequency middle value) / optimal frequency middle value |. The sound wave amplitude is dynamically adjusted according to the changes in the user's heart rate and breathing rate. If the instantaneous value of the heart rate or breathing rate exceeds the normal range when adjusting the sound wave amplitude, the adjustment is suspended until the physiological indicators return to normal. The physiological indicator parameters include the energy proportion of δ waves, θ waves, α waves and β waves, the mean and coefficient of variation of the heart rate, and the mean and coefficient of variation of the breathing rate. During sleep, the user's physiological indicator parameters are continuously monitored as sleep quality evaluation indicators, and the weighted average algorithm is used to comprehensively score the sleep quality evaluation indicators to obtain the user's current sleep quality score, where the weight coefficient of the sleep quality evaluation indicator is determined according to its correlation coefficient with the overall sleep quality. After the user finishes sleeping, the optimal magnetic field intensity, sound wave frequency and amplitude parameters of each sleep stage recorded during the sleep process are entered into the user's personalized sleep parameter database and merged with the user's historical sleep parameter data.

[0052] Specifically, the data obtained from the user's personalized sleep parameter database show that the historical best magnetic field strength of the user in the deep sleep stage is 4 to 6mT, and the sound wave frequency is 80 to 100Hz, while in the shallow sleep stage and rapid eye movement sleep stage, the best sound wave frequency is 200Hz. According to the user data obtained in real time by the sleep monitoring device, the user's sleep stage judgment model is used to judge that the current user is in the deep sleep stage, the best magnetic field strength range is 4 to 6mT, and the best sound wave frequency is 80 to 100Hz. By controlling the output current of the magnetic field generator, the magnetic field strength is gradually increased from the current 3mT to 4mT, and then gradually increased to 6mT. If the proportion of delta wave energy does not show an upward trend in three consecutive adjustment cycles during the process of increasing the magnetic field strength, the adjustment is stopped and maintained at the current value. When the user enters the shallow sleep or rapid eye movement sleep stage, the system controls the output frequency of the sound wave generator to keep the sound wave frequency at 200Hz, ensuring that the sound wave frequency deviation does not exceed 5%, that is, the sound wave frequency deviation = |(actual frequency-200Hz) / 200Hz|. According to the changes in the user's heart rate and breathing rate, the sound wave amplitude is dynamically adjusted. If the user's heart rate exceeds 75 times / minute or the breathing rate exceeds 20 times / minute when adjusting the sound wave amplitude, the adjustment is suspended until the heart rate and breathing rate return to the normal range. During sleep, the user's physiological indicators are continuously monitored, including 40% of delta wave energy, 25% of theta wave energy, 15% of alpha wave energy, 10% of beta wave energy, the average heart rate is 60 times / minute, the heart rate variation coefficient is 5%, the average breathing rate is 16 times / minute, and the breathing rate variation coefficient is 3%. Using these physiological indicators, the system calculates the user's current sleep quality score through a weighted average algorithm, where the weight coefficient of each indicator is determined according to its correlation coefficient with the overall sleep quality. After the user's sleep ends, the optimal magnetic field intensity, sound wave frequency and amplitude parameters recorded in each sleep stage during the sleep process are entered into the user's personalized sleep parameter database and merged with the user's historical sleep parameter data.

[0053] S105. Determine the estimated landing and dining time according to the user's flight information and electronic schedule, perform brain wave wakefulness detection, and adjust the magnetic field and sound wave parameters based on the user's wakefulness, continuously monitor and optimize the user's sleeping environment.

[0054] The user's estimated landing time and dining time are determined through the user's flight information and electronic schedule. The flight number, take-off time and landing time are extracted from the user's electronic ticket, and the actual landing time of the user is estimated in combination with the real-time dynamic data of the flight. The dining arrangements with dining keywords are identified from the user's electronic calendar, and the dining time is extracted. The dining keywords include dining, dining, lunch, and dinner. If the user has a dining arrangement before takeoff, the landing time and the dining time are ORed, and the earlier dining time point is used as the trigger condition for starting the brain wave awakeness detection. If the trigger condition is met, the built-in dry electrode electroencephalogram sensor is used to obtain the user's single-lead brain wave signal in real time, and the Mallat algorithm is used to perform wavelet decomposition on the brain wave signal, and the square sum of the wavelet coefficients of each layer is performed to obtain the wavelet coefficient energy value of the frequency band, including δ waves, θ waves, α waves, and β waves. The wavelet coefficient energy value is normalized to obtain the relative energy distribution of each frequency band as the time-frequency domain feature of the brain wave. The brain wave data of the user at different levels of wakefulness are obtained through the user's historical sleep monitoring database, and the decision tree algorithm is used for model training to build a wakefulness discrimination model. The wakefulness levels include wakefulness, sleepiness, and micro-sleep. According to the extracted brain wave time-frequency domain characteristics, the wakefulness discrimination model is used to judge the user's wakefulness. If the output result of the wakefulness discrimination model is the awake state, the parameter adjustment mechanism of the magnetic field generator and the sound wave generator is triggered, and the changes in the user's brain wave characteristics are continuously monitored to judge the user's wakefulness. If the output results of the wakefulness discrimination model remain in the awake state within the continuous preset time window, the speed of parameter adjustment of the magnetic field generator and the sound wave generator is accelerated to shorten the time for the device to shut down. After the parameter adjustment of the magnetic field generator and the sound wave generator is completed, the magnetic field generator and the sound wave generator are turned off to stop interfering with the user's sleeping environment.

[0055] Specifically, through natural language processing technology, the flight number AB123, the take-off time 10:00 and the landing time 15:00 are extracted from the user's electronic ticket, and then combined with the real-time dynamic information interface of the flight, the actual landing time of the user is estimated to be 15:30. At the same time, regular expressions are used to match the time and dining keywords in the user's schedule, and the schedule information such as "12:30 lunch" is extracted. The landing time and the most recent dining time are added to the wakefulness detection trigger condition queue. The dining keywords include dining, dining, lunch, and dinner. Since the user has a meal arrangement on the plane before takeoff, the dining time 12:30 is used as the trigger condition for starting the brain wave wakefulness detection. When the trigger condition is met, the built-in dry electrode electroencephalogram sensor is used to obtain the user's single-lead brain wave signal in real time, and the Mallat algorithm is used to perform wavelet decomposition on the brain wave signal. The wavelet coefficients of each layer are squared and summed to obtain the wavelet coefficient energy value of the frequency band, including δ waves, θ waves, α waves, and β waves. The energy values ​​of these wavelet coefficients are normalized to obtain the relative energy distribution of each frequency band as the time-frequency domain characteristics of the brain wave. The brain wave data of the user at different levels of wakefulness are obtained through the user's historical sleep monitoring database. The decision tree algorithm is used for model training to build a wakefulness discrimination model. The wakefulness is divided into wakefulness, sleepiness, and micro-sleep. According to the extracted brain wave time-frequency domain characteristics, the wakefulness discrimination model is used to judge the user's wakefulness. If the model output result is a wakeful state, the parameter adjustment mechanism of the magnetic field generator and the sound wave generator is triggered, and the changes in the user's brain wave characteristics are continuously monitored. If during the parameter adjustment process, the user's wakefulness discrimination model always judges that the user is in a wakeful state within a continuous 10-minute time window, the system speeds up the parameter adjustment speed of the magnetic field generator and the sound wave generator, and shortens the time for the device to shut down. When the parameter adjustment is completed, the magnetic field generator and the sound wave generator are completely turned off, and the intervention in the user's sleeping environment is stopped, ensuring that the user can successfully complete the meal or cope with the activities after landing.

[0056] S106. Based on the user's historical sleep data and current sleep status, a long short-term memory neural network algorithm is used to establish a user sleep quality prediction model to determine whether the user has sleep quality problems in a preset time period in the future, and determine the optimal magnetic field strength and sound wave frequency adjustment strategy based on a reinforcement learning algorithm.

[0057] The user's historical sleep data is obtained through the user's historical sleep monitoring database. The sleep data includes the time of falling asleep, sleep duration, deep and shallow sleep ratio, rapid eye movement sleep duration, number of awakenings during sleep, and the sleep quality score of the day. The historical sleep data is arranged in chronological order to construct a time series data set, and the historical sleep data is normalized. According to the time series data set of historical sleep data, a long short-term memory neural network is used for model training to construct a sleep quality score prediction model to predict the sleep quality score in the future preset time period. According to the user's real-time sleep data of the day, the sleep quality score prediction model is used to predict the sleep quality score in the future preset time period. If the predicted score in the future preset time period shows a downward trend and the lowest score is lower than the preset score threshold, it is judged that the user has a sleep quality problem. According to the prediction results of the sleep quality score prediction model, the optimal magnetic field strength and sound wave frequency adjustment strategy is obtained through the policy gradient algorithm in reinforcement learning. The user's sleep data is continuously monitored, the training data of the sleep quality score prediction model is regularly updated, and the model is retrained and optimized.

[0058] Specifically, the sleep data obtained from the user's historical sleep monitoring database includes falling asleep at 22:30, sleeping for 8 hours, a ratio of deep and shallow sleep of 60:40, a rapid eye movement sleep duration of 1.5 hours, waking up twice during sleep, and a sleep quality score of 85 points on that day. These historical sleep data are arranged in chronological order, constructed into a time series data set, and the data is normalized. According to the processed time series data set of historical sleep data, a long short-term memory neural network is used for model training to construct a sleep quality score prediction model to predict the sleep quality score in the future preset time period. The trained model can predict the sleep quality score in the next 7 days. If on a certain night, through the user's real-time sleep data of the day, the time to fall asleep is 23:00, the current sleep duration is 6 hours, the ratio of deep and shallow sleep is 50:50, the rapid eye movement sleep duration is 1 hour, and the number of wake-ups during sleep is 3 times, the sleep quality score prediction model is used to predict the future preset time period, such as the sleep quality score in the next week. If the prediction results show that the sleep quality score in the next week is on a downward trend, and the lowest score drops to 65 points, which is lower than the preset score threshold of 70 points, the user is judged to have sleep quality problems. According to the prediction results of the sleep quality score prediction model, the magnetic field strength and sound wave frequency parameters are dynamically adjusted. If the current magnetic field strength is 3mT and the sound wave frequency is 100Hz, the magnetic field strength and sound wave frequency parameters are adjusted through the policy gradient algorithm in reinforcement learning, and finally the optimal adjustment strategy is obtained, and the magnetic field strength is adjusted to 5mT and the sound wave frequency is adjusted to 80Hz to improve the user's sleep quality. In this process, the user's real-time sleep data is continuously monitored, including the current time of falling asleep, sleep duration, deep and shallow sleep ratio, rapid eye movement sleep duration, and the number of wake-ups during sleep. The training data of the sleep quality score prediction model is regularly updated with the latest data, and the model is retrained and optimized to ensure the accuracy and effectiveness of the model prediction.

[0059] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them. Although the present invention has been described in detail with reference to the above embodiments, a person skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some or all of the technical features can be replaced by equivalents. However, these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A sleep regulation method based on magnetic field technology, characterized in that: The method comprises: According to the user's set sleeping time and current flight status, combined with the time zone information of the departure and destination, the user's brain waves are detected in real time through the EEG sensor, the fast Fourier transform algorithm is used to perform frequency domain analysis on the brain wave signal, and a user sleep stage judgment model is constructed to judge the user's current sleep stage; According to the user's current sleep stage information and destination time zone information, the optimal magnetic field strength and sound wave frequency combination for the sleep stage is obtained through the user's personalized sleep parameter database, and the magnetic field strength and frequency of the magnetic field generator built into the eye mask, as well as the sound wave frequency and amplitude of the sound wave generator, are dynamically adjusted; According to the real-time noise audio signal and the user's sleep state, the parameters of the sound wave generator are dynamically adjusted in combination with the noise suppression strategy database to generate sound waves that match the noise frequency to offset the noise; The user's physiological parameters, magnetic field and sound wave frequency data are obtained through the user's historical sleep monitoring database, and the optimal magnetic field strength and sound wave frequency parameter values ​​under different sleep stages are predicted using a recurrent neural network model, and updated to the user's personalized sleep parameter database; Determine the estimated landing and meal time based on the user's flight information and electronic schedule, perform brain wave wakefulness detection, and adjust the magnetic field and sound wave parameters based on the user's wakefulness, continuously monitoring and optimizing the user's sleeping environment; Based on the user's historical sleep data and current sleep status, a long short-term memory neural network algorithm is used to establish a user sleep quality prediction model to determine whether the user has sleep quality problems in the future preset time period, and determine the optimal magnetic field strength and sound wave frequency adjustment strategy based on the reinforcement learning algorithm.

2. The method according to claim 1, wherein: The method uses an electroencephalogram sensor to detect the user's brain waves in real time based on the user's set sleeping time and current flight status, combined with the time zone information of the departure and destination, uses a fast Fourier transform algorithm to perform frequency domain analysis on the brain wave signal, and constructs a user sleep stage judgment model to judge the user's current sleep stage, including: Obtain the user's set sleep time, current flight status, and time zone information of the departure and destination. The current flight status includes flight altitude, speed, and in-flight environment. Calculate the time zone difference between the departure and destination based on the time zone information of the departure and destination, and estimate the user's current biological clock deviation. Obtain the user's brain wave signal in real time through the EEG sensor, apply bandpass filter and wavelet transform to pre-process the brain wave signal to improve the quality of the brain wave signal. The pre-processing includes removing low-frequency interference, high-frequency noise, and environmental noise. Use the fast Fourier transform algorithm to convert the time domain brain wave signal into a frequency domain signal to obtain the amplitude and phase information of different frequency components. Extract key features from the frequency domain signal, including the frequency and amplitude of the main frequency components. The main frequency components include α wave, β wave, θ wave, and δ wave. According to the preset sampling rate and window length, the frequency and amplitude characteristics of brain waves are extracted and optimized; the historical user brain wave signal data is obtained through the user's historical sleep monitoring database, and the historical user brain wave signal data is labeled with sleep stages; based on the historical user brain wave signal data, the fast Fourier transform algorithm is used to obtain the amplitude and phase information of different frequency components, and key features are extracted from the frequency domain signal, and the model is trained using the random forest algorithm to build a user sleep stage judgment model; based on the user's current brain wave frequency domain characteristics, the user's sleep stage is judged using the user's sleep stage judgment model, and the sleep stages include light sleep, deep sleep or rapid eye movement; the user's sleep stage judgment result is compared with the sleep state estimated based on the time zone information and the time of falling asleep, to evaluate whether the user's sleep state meets expectations.

3. The method according to claim 1, wherein: The method dynamically adjusts the magnetic field strength and frequency of the magnetic field generator built into the eye mask, and the sound wave frequency and amplitude of the sound wave generator, according to the user's current sleep stage information and the destination time zone information, and obtains the optimal magnetic field strength and sound wave frequency combination in the sleep stage from the user's personalized sleep parameter database, including: Through the user's historical sleep monitoring database, the historical user's sleep stage information and the magnetic field strength and sound wave frequency combination parameters under the sleep stage are obtained, and a user's personalized sleep quality prediction model is built using a deep neural network to predict the optimal magnetic field strength and sound wave frequency combination parameters under different sleep stages of the user; based on the user's current sleep stage information detected in real time by the electroencephalogram sensor, the user's personalized sleep quality prediction model is used to obtain the optimal magnetic field strength and sound wave frequency combination parameters under the sleep stage; the magnetic field strength and sound wave frequency parameters predicted by the user's personalized sleep quality prediction model are input into the magnetic field generator and sound wave generator built into the eye mask, and the magnetic field generator is controlled to adjust the coil current according to the input parameters to generate a sound wave with the optimal magnetic field strength The system generates a magnetic field that matches the user's degree of sleep, controls the sound wave generator to adjust the amplitude and frequency according to the input parameters, and generates sound waves that match the optimal audio frequency; monitors the user's physiological signals in real time, including heart rate, breathing rate, and body movement times, and compares them with the user's personal normal range; uses the destination time zone information to calculate the day and night time corresponding to the current moment at the destination, and obtains the optimal magnetic field and sound wave parameter values ​​for each sleep stage of the user in the destination time zone from the user's personalized sleep parameter database; sets a time window for gradual adjustment based on the time difference between the current moment and the preset arrival time, and gradually adjusts the magnetic field and sound wave parameters to the optimal value that matches the destination's circadian rhythm based on the time difference, and continuously monitors the user's physiological signal feedback during the adjustment process.

4. The method according to claim 1, wherein: The method dynamically adjusts the parameters of the sound wave generator according to the noise audio signal obtained in real time and the user's sleep state in combination with the noise suppression strategy database to generate a sound wave that matches the noise frequency to offset the noise, including: By arranging multiple MEMS digital noise sensors at different positions in the aircraft cabin, the noise audio signal is acquired in real time, and the noise audio signal is preprocessed, including removing DC bias, removing low-frequency noise, and removing high-frequency noise; the preprocessed noise audio signal is analyzed in the frequency domain using the fast Fourier transform algorithm to obtain the frequency spectrum distribution of the noise, and determine the main noise frequency components and their corresponding amplitudes; the frequency spectrum data of the noise is combined with the current sleep state of the user, and the optimal audio signal processing strategy and sound wave generator parameter combination are selected from the noise suppression strategy database; according to the selected audio signal processing strategy, the noise audio signal is filtered and attenuated to obtain the processed noise audio signal; according to the real-time acquired noise audio signal and the user's sleep state, the sound wave generator parameters are dynamically adjusted to optimize noise suppression and predict changes in sleep quality; according to the selected sound wave generator parameters, the sound wave generator is controlled to generate a specified sound wave with the same frequency, equal amplitude, and a phase difference of 180° as the processed noise signal; Using the principle of sound wave interference, the specified sound waves generated by the sound wave generator are superimposed on the actual noise in the aircraft cabin, and the noise is canceled out through destructive interference; a reference microphone is set around the user's ear to obtain an audio signal mixed with anti-noise and residual noise, and the LMS algorithm is used to perform adaptive filtering on the audio signal obtained by the reference microphone. The filter coefficients are estimated based on the reference signal and the error signal, and the filter coefficients are iteratively updated using the gradient descent method until the mean square value of the error signal is minimized, where the reference signal is the anti-noise and the error signal is the signal obtained by the reference microphone; the sound wave parameters are continuously iterated and optimized until the residual noise meets the comfort requirements.

5. The method according to claim 4, wherein: The method dynamically adjusts the sound wave generator parameters to optimize noise suppression and predicts changes in sleep quality based on the noise audio signal obtained in real time and the user's sleep state, including: The ambient noise in the aircraft cabin is acquired in real time using audio acquisition equipment to obtain the original noise audio data; the original noise audio data is converted into a spectrum representation using the short-time Fourier transform algorithm to obtain noise characteristic data; Through the portable EEG headset and smart bracelet worn by the user, the user's sleep status data is obtained in real time, including sleep stage, brain wave frequency and heart rate. The brain wave frequencies include δ wave, θ wave, α wave and β wave. The obtained sleep status data is synchronized with the noise characteristic data to build a noise-sleep status mapping data set. According to the noise-sleep status mapping data set, the model is trained using the long short-term memory network algorithm to build a sleep quality prediction model to predict the impact score of noise on sleep quality. The noise characteristics and current sleep status data during the flight are obtained in real time, and the sleep quality prediction model is used to predict the impact score of noise on sleep quality in the future preset time period. If the predicted impact If the sound score is lower than the pre-threshold, it is necessary to further optimize the noise reduction measures; according to the prediction results, the parameters of the sound wave generator, including the sound wave frequency, amplitude and phase, are adaptively adjusted; the difference between the predicted score and the actual score is minimized through the gradient descent algorithm, and the optimal combination of sound wave parameters is solved to minimize the impact of noise; the optimized parameters are input into the sound wave generator to generate sound waves with the same frequency, equal amplitude and opposite phase as the original noise, and the noise reduction effect is achieved after superposition with the ambient noise through the principle of sound wave interference; the user's sleep state changes and environmental noise changes are continuously monitored, the noise-sleep state mapping data set is dynamically updated, and the sleep quality prediction model and sound wave parameters are regularly retrained.

6. The method according to claim 1, wherein: The method obtains the user's physiological parameters and magnetic field and sound wave frequency data through the user's historical sleep monitoring database, uses a recurrent neural network model to predict the optimal magnetic field strength and sound wave frequency parameter values ​​under different sleep stages, and updates them to the user's personalized sleep parameter database, including: Through the user's historical sleep monitoring database, the user's physiological parameters per minute, as well as the magnetic field strength and sound wave frequency parameters at the corresponding moment, are extracted. The physiological parameters include brain wave characteristics, heart rate, and breathing rate. The EEG device is used to obtain the user's EEG signal during sleep, and the EEG signal is filtered at 50Hz to remove power frequency interference. The EEG signal is decomposed into 4 layers using wavelet transform to extract the energy characteristics of the four frequency bands of δ, θ, α, and β. Use a piezoelectric sensor to obtain the vibration signal of the user's mattress, extract the breathing frequency characteristics, and use a photoplethysmographic sensor to obtain the user's heart rate signal; The acquired physiological signal characteristics are associated with the current magnetic field strength and sound wave frequency parameter values ​​to construct training data samples; the training data is preprocessed, including data cleaning and normalization; Based on the training data samples, a recurrent neural network algorithm is used to train the model and predict the optimal magnetic field strength and sound wave frequency parameter values ​​under different sleep stages; based on the optimal magnetic field strength and sound wave frequency parameter values ​​under different sleep stages of the user, the magnetic field generator and sound wave generator parameters are dynamically adjusted in combination with real-time sleep monitoring data, and the user's sleep quality score is calculated; the optimal magnetic field strength and sound wave frequency parameter values ​​predicted by the recurrent neural network model are updated to the user's personalized sleep parameter database, and stored in association with the user's historical sleep data; the recurrent neural network model is retrained and optimized regularly, and the optimal magnetic field strength and sound wave frequency parameter values ​​in the user's personalized sleep parameter database are updated.

7. The method according to claim 6, wherein: The method dynamically adjusts the parameters of the magnetic field generator and the sound wave generator according to the optimal magnetic field strength and sound wave frequency parameter values ​​under different sleep stages of the user in combination with real-time sleep monitoring data, and calculates the user's sleep quality score, including: Obtain the historical optimal magnetic field strength and sound wave frequency parameter values ​​of the user in different sleep stages from the user personalized sleep parameter database; use the user sleep stage judgment model to judge the user's current sleep stage based on the user's real-time sleep data obtained by the sleep monitoring device, and obtain the optimal magnetic field strength and sound wave frequency parameter range corresponding to the sleep stage; when the user enters the deep sleep stage, by controlling the output current of the magnetic field generator, the magnetic field strength is gradually increased from the current value to the upper limit of the optimal magnetic field strength parameter range; if the current value of the magnetic field strength is lower than the lower limit of the optimal magnetic field strength parameter range, the magnetic field strength is first adjusted to the lower limit, and then gradually increased to the upper limit; If, in the process of increasing the magnetic field strength, the δ wave energy ratio does not show an upward trend in multiple consecutive adjustment cycles, the adjustment is stopped and maintained at the current value; when the user is in the light sleep or rapid eye movement sleep stage, the sound wave frequency is maintained at the middle value of the optimal sound wave frequency parameter range by controlling the output frequency of the sound wave generator, and the sound wave frequency deviation does not exceed 5%. The sound wave frequency deviation is calculated as sound wave frequency deviation = |(actual frequency-optimal frequency middle value) / optimal frequency middle value|; the sound wave amplitude is dynamically adjusted according to the changes in the user's heart rate and breathing rate. If the instantaneous value of the heart rate or breathing rate exceeds the normal range when adjusting the sound wave amplitude, the adjustment is suspended until the physiological indicators return to normal. The sleep quality index is restored to normal, and the physiological index parameters include the energy proportion of delta wave, theta wave, alpha wave and beta wave, the mean and coefficient of variation of heart rate, the mean and coefficient of variation of respiratory rate; during sleep, the user's physiological index parameters are continuously monitored as sleep quality evaluation indicators, and the weighted average algorithm is used to comprehensively score the sleep quality evaluation indicators to obtain the user's current sleep quality score, wherein the weight coefficient of the sleep quality evaluation index is determined according to its correlation coefficient with the overall sleep quality; after the user's sleep ends, the optimal magnetic field intensity, sound wave frequency and amplitude parameters of each sleep stage recorded during the sleep process are entered into the user's personalized sleep parameter database and merged with the user's historical sleep parameter data.

8. The method according to claim 1, wherein: The estimated landing and dining time is determined according to the user's flight information and electronic schedule, the brain wave awakening level is detected, and the magnetic field and sound wave parameters are adjusted based on the user's awakening level, and the user's sleeping environment is continuously monitored and optimized, including: Determine the user's estimated landing time and meal time based on the user's flight information and electronic schedule; The flight number, take-off time and landing time are extracted from the user's electronic ticket, and the user's actual landing time is estimated in combination with the real-time dynamic data of the flight; the dining arrangements with dining keywords are identified from the user's electronic calendar, and the dining times are extracted. The dining keywords include dining, dining, lunch and dinner; if the user has dining arrangements before take-off, the landing time and dining time are ORed together, and the earlier dining time is used as the trigger condition for starting the brain wave wakefulness detection; if the trigger condition is met, the built-in dry electrode EEG sensor is used to obtain the user's single-lead EEG signal in real time, and the Mallat algorithm is used to perform wavelet decomposition on the EEG signal, and the square sum of the wavelet coefficients of each layer is performed to obtain different The wavelet coefficient energy value of the frequency band, different frequency bands include δ wave, θ wave, α wave, and β wave; the wavelet coefficient energy value is normalized to obtain the relative energy distribution of each frequency band as the time-frequency domain feature of the brain wave; the brain wave data of the user at different levels of wakefulness are obtained through the user's historical sleep monitoring database, and the decision tree algorithm is used for model training to build a wakefulness discrimination model, and the wakefulness level includes wakefulness, sleepiness, and micro-sleep; according to the extracted brain wave time-frequency domain features, the wakefulness discrimination model is used to judge the user's wakefulness; if the output result of the wakefulness discrimination model is a wakefulness state, the parameter adjustment mechanism of the magnetic field generator and the sound wave generator is triggered, and the changes in the user's brain wave characteristics are continuously monitored to judge the user's wakefulness level; If the output results of the wakefulness discrimination model remain in the wakefulness state within a continuous preset time window, the speed of adjusting the parameters of the magnetic field generator and the sound wave generator will be accelerated, and the time it takes to shut down the device will be shortened; after the parameter adjustment of the magnetic field generator and the sound wave generator is completed, the magnetic field generator and the sound wave generator will be turned off to stop interfering with the user's sleeping environment.

9. The method according to claim 1, wherein: The method uses a long short-term memory neural network algorithm to establish a user sleep quality prediction model based on the user's historical sleep data and current sleep state, determines whether the user has sleep quality problems in a preset time period in the future, and determines the optimal magnetic field strength and sound wave frequency adjustment strategy based on a reinforcement learning algorithm, including: The user's historical sleep data is obtained through the user's historical sleep monitoring database, and the sleep data includes the time of falling asleep, sleep duration, ratio of deep and light sleep, rapid eye movement sleep duration and number of awakenings during sleep, as well as the sleep quality score of the day; the historical sleep data is arranged in chronological order to construct a time series data set, and the historical sleep data is normalized; based on the time series data set of the historical sleep data, a long short-term memory neural network is used for model training to construct a sleep quality score prediction model to predict the sleep quality score within a preset time period in the future; based on the user's real-time sleep data of the day, the sleep quality score prediction model is used to predict the sleep quality score within a preset time period in the future; if the predicted score within the preset time period in the future shows a downward trend and the lowest score is lower than the preset score threshold, it is judged that the user has a sleep quality problem; based on the prediction results of the sleep quality score prediction model, the optimal magnetic field strength and sound wave frequency adjustment strategy is obtained through the policy gradient algorithm in reinforcement learning; the user's sleep data is continuously monitored, the training data of the sleep quality score prediction model is regularly updated, and the model is retrained and optimized.

Images