Sleep assisting system and method for dynamic natural sound simulation

Abstract

The invention discloses a sleep assistance system and method for dynamic natural sound simulation, and particularly relates to the technical field of sleep assistance, comprising: acquiring historical preference data of a user and external environment data, generating a first reference coefficient based on the historical preference data of the user and the external environment data, the historical preference data comprises a sound type, a volume value and a playing duration; acquiring historical sleep cycle data of the user, and inputting the first reference coefficient and the historical sleep cycle data into a pre-constructed first sound training model; the sleep state is determined on the basis of the physiological evaluation coefficient obtained in real time, the sound playing strategy is optimized by comparing the sleep state with the preset sleep duration, and dynamic intelligent regulation and control of sleep assistance are achieved. In the sleeping process, sound parameters can be adjusted in time according to the actual duration condition of light sleep, deep sleep, dream or awakening state of the user.

Description

Technical Field

[0001] The present invention relates to the technical field of sleep assistance, and more specifically, to a sleep assistance system and method for dynamic natural sound simulation. Background Art

[0002] Traditional sleep aid devices or wake-up devices usually have a single function and are difficult to balance the simulation of natural sounds and a gentle wake-up experience. Sleep aid devices often use white noise, light music or other fixed sound effects to help users fall asleep, but ignore the simulation effect of dynamic natural sounds. Wake-up devices often wake up users with abrupt ringing or vibration, which is likely to interfere with the user's sleep cycle and cause bad moods and physical discomfort.

[0003] In existing methods, for example, Chinese Patent Application No. CN101940813A discloses a multi-functional sleep assistance system and its control method. An artistic conception sound device and an air humidity adjustment device are provided inside the cabinet of the system. An air quality adjustment device and an air purification and circulation device are provided inside the air humidity adjustment device. An artistic conception light device, a humidifier fog tube, a table lamp, an artistic conception light device, an artistic conception sound device, and a clock device are provided on the tabletop of the cabinet and are integrated with the cabinet. The above devices are controlled manually and automatically, and a computer control system is provided inside the cabinet. The above method expands the functions of the bedside table and the humidifier and improves the sleep quality. However, through research and application of the above method and the prior art, it is found that the above method and the prior art have at least the following partial defects:

[0004] The above method only simply controls the on / off of each device and the adjustment of basic parameters, lacking an automatic intelligent control mechanism after in-depth analysis of the sleep process and being unable to well adapt to the dynamic changes during sleep.

[0005] Therefore, there is an urgent need for a system that can dynamically simulate natural sounds, which can not only assist users in falling asleep through natural environmental sound effects but also gradually play natural wake-up sound effects within the user-set wake-up time range, thereby reducing the discomfort caused by waking up angry and improving the user's sleep quality and wake-up experience. Therefore, the present invention provides a sleep assistance system and method for dynamic natural sound simulation. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides a sleep assistance system and method for dynamic natural sound simulation to solve the problems raised in the above background art.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] In the first aspect, the present invention provides a sleep assistance method for dynamic natural sound simulation, including:

[0009] Step 1: Obtain the user's historical preference data and external environment data, and generate a first reference coefficient based on the user's historical preference data and external environment data. The historical preference data includes sound type, volume value, and playing duration.

[0010] Step 2: Obtain the user's historical sleep cycle data, and input the first reference coefficient and the historical sleep cycle data into a pre-constructed first sound training model to obtain the user's initial sound playback data.

[0011] Step 3: Obtain the user's physiological evaluation coefficient within the T time period, and determine the sleep state of the user within the T + a time period based on the physiological evaluation coefficient.

[0012] Step 4: Based on the sleep state of the user within the T + a time period, obtain the corresponding sleep duration, compare the corresponding sleep duration with the preset sleep duration, and optimize the initial sound playback data according to the comparison result to obtain a sound regulation strategy.

[0013] Furthermore, the external environment data includes the temperature difference, humidity difference, and sound intensity at which the user is located per unit time. The method for generating the first reference coefficient based on the user's historical preference data and external environment data includes:

[0014] Step s1: Analyze the historical preference data to obtain a historical preference coefficient HPC;

[0015] Step s2: Analyze the external environment data to obtain an environmental evaluation coefficient EAC;

[0016] Step s3: Let the current moment be t, the historical preference coefficient sequence be {HPC 1 , HPC 2 ,..., HPC t}, and HPC t be the historical preference coefficient at the current moment t;

[0017] The environmental evaluation coefficient sequence is {EAC 1 , EAC 2 ,..., EAC t}, and EAC t be the environmental evaluation coefficient at the current moment t; the attenuation factor is

[0018] Step s4: Calculate the weighted sum of the historical preference coefficients r = 1, 2, ……, t;

[0019] and calculate the weighted sum of the environmental evaluation coefficients

[0020] Step s5: Comprehensively process the historical preference coefficient HPC and the environmental assessment coefficient EAC to obtain the first reference coefficient, and its calculation formula is:

[0021]

[0022] In the formula, FRC represents the first reference coefficient, σ 1 and σ 2 are weight factors, and σ 1 +σ 2 = 1.

[0023] Furthermore, the method for obtaining the historical preference coefficient includes:

[0024] Step a1: Suppose there are n different sound types, and each sound type corresponds to a technical score S i , i = 1, 2,..., n;

[0025] Step a2: Perform weighted adjustment according to the selection frequency F i of the sound type by the user. The F i is the proportion of the number of times the i-th sound type is selected to the total number of selections;

[0026] Step a3: The volume value V is non-linearly mapped according to the logarithmic relationship. Let the minimum volume value be V min , and the maximum volume value be V max , and its mapping formula is:

[0027]

[0028] In the formula, V m represents the volume value obtained after non-linearly mapping the volume value V;

[0029] Step a4: The sound duration C is processed using a piecewise function. Let C 1 , C 2 , C 3 be different duration thresholds; when C < C 1 , C p = 0; when C 1 < C < C 2 , when C 2 < C < C 3 , when C > C 3 , C p = 1, C p is the playing duration;

[0030] Step a5: Combine the sound type S i , the selection frequency F i , the volume value V after non-linear mapping, and...m and the playback duration C p Perform a formulaic calculation to obtain the historical preference coefficient;

[0031]

[0032] In the formula, HPC represents the historical preference coefficient.

[0033] Furthermore, the method for obtaining the environmental assessment coefficient includes:

[0034] Step b1: The temperature standard deviation ΔT is processed using a Gaussian function. Let the standard temperature range be [T min , T max , and the real-time temperature is T s . The calculation formula is:

[0035]

[0036] Step b1: The humidity standard deviation ΔH is processed using a Gaussian function. Let the standard humidity range be [H min , H max , and the real-time humidity is H s . The calculation formula is:

[0037]

[0038] Step b3: The sound intensity standard deviation ΔQ is processed using a power function. Let the standard sound intensity range be [Q min , Q max , and the real-time sound intensity is Q s . The calculation formula is:

[0039]

[0040] Step b4: The calculation formula for the environmental assessment coefficient EAC is:

[0041] EAC = μ 1 ×ΔT + μ 2 ×ΔH + μ 3 ×ΔQ

[0042] In the formula, EAC represents the environmental assessment coefficient, and μ 1 , μ 2 and μ 3 are weighting factors, and μ 1 + μ 2 + μ 3 = 1.

[0043] Further, the historical sleep cycle data includes each sleep state and its corresponding sleep duration, and the sleep states include light sleep state, deep sleep state, dream state, and wake state; the initial sound playback data includes the initial music type, the initial sound volume, and the initial playback duration.

[0044] The method for obtaining the historical sleep cycle data of a user includes:

[0045] Step c1: Obtain historical electroencephalogram data, use the time in the historical electroencephalogram data as the horizontal axis, and the electroencephalogram in the historical electroencephalogram data as the vertical axis to construct an electroencephalogram time-domain graph, and the electroencephalogram includes θ wave, δ wave, β wave, and α wave.

[0046] Step c2: Divide the electroencephalogram time-domain graph according to the electroencephalogram to obtain a set of historical electroencephalograms, and the set of historical electroencephalograms includes Z historical electroencephalogram waveforms, and the historical electroencephalogram waveforms include θ wave waveforms, δ wave waveforms, β wave waveforms, and α wave waveforms.

[0047] Step c3: Extract the z-th historical electroencephalogram waveform in the set of historical electroencephalograms, where z is an integer greater than zero, and the initial value of z is 1.

[0048] Step c4: Extract the corresponding standard electroencephalogram waveform, calculate the similarity between the historical electroencephalogram waveform and the standard electroencephalogram waveform. If the similarity between the historical electroencephalogram waveform and the standard electroencephalogram waveform is greater than or equal to the preset electroencephalogram similarity threshold, then extract the corresponding sleep duration and jump to step c5; if the similarity between the historical electroencephalogram waveform and the standard electroencephalogram waveform is less than the preset electroencephalogram similarity threshold, then directly jump to step c5.

[0049] Step c5: Let z = z + 1 and jump back to step c3.

[0050] Step c6: Repeat steps c3 to c5 until z = Z, then end the loop and obtain the corresponding sleep duration.

[0051] Further, the construction method of the first sound training model includes:

[0052] Divide the historical voice training data into a voice training set and a voice test set, and construct a regression network model. The historical voice training data includes voice feature data and corresponding initial voice playback data. The voice feature data includes a first reference coefficient and historical sleep cycle data. Use the voice feature data in the voice training set as the input of the regression network model, and use the corresponding initial voice playback data in the voice training set as the output of the regression network model to train the regression network model to obtain an initial first regression network model. With minimizing the sum of prediction accuracies as the training objective, use the voice test set to evaluate the initial first regression network model. If the sum of prediction accuracies is less than the threshold of the preset sum of accuracies, then use the corresponding initial first regression network model as the first voice training model. If the sum of prediction accuracies is greater than or equal to the threshold of the preset sum of accuracies, then use the original voice training set to retrain the initial first regression network model until the test result meets the set threshold. The initial first regression network model is a decision tree regression network model, a random forest regression network model, a support vector machine regression network model, a linear regression network model, or a neural network model.

[0053] Further, the method for obtaining the physiological evaluation coefficient of the user within the T time period includes:

[0054] Obtain the real-time physiological representation data of the user. The real-time physiological representation data includes the absolute value of the heart rate difference, the absolute value of the respiratory rate difference, the type of brain waves, and the absolute value of the muscle electrical activity frequency difference.

[0055] Normalize the real-time physiological representation data to obtain the physiological evaluation coefficient.

[0056] The method for determining the sleep state of the user within the T + a time period based on the physiological evaluation coefficient includes:

[0057] Input the physiological evaluation coefficient into a pre-constructed state prediction model to obtain the sleep state of the user within the T + a time period. Among them, the specific training process of the state prediction model includes:

[0058] Pre-collect P groups of physiological evaluation coefficients, where P is an integer greater than 1. Set corresponding prediction results for the physiological evaluation coefficients. The prediction results include light sleep state, deep sleep state, dream state, and wake state. Set different digital labels for the light sleep state, deep sleep state, dream state, and wake state. Mark the digital label of the prediction result as the prediction label, and convert the physiological evaluation coefficient and the corresponding prediction label into a corresponding set of feature vectors.

[0059] Use each set of feature vectors as the input of the state prediction model. The state prediction model outputs a set of prediction labels corresponding to each set of physiological evaluation coefficients, and uses the actual prediction labels corresponding to each set of physiological evaluation coefficients as the prediction target. The actual prediction labels are the pre-set prediction labels corresponding to the physiological evaluation coefficients. Use minimizing the sum of the prediction errors of the prediction labels corresponding to all physiological evaluation coefficients as the training target. Train the state prediction model until the sum of the prediction errors converges and then stop training. The state prediction model is a deep neural network model.

[0060] Further, the comparison results include light sleep optimization instructions, deep sleep optimization instructions, dream optimization instructions, and wakefulness optimization instructions. The preset sleep duration includes a first sleep duration O 1 , a second sleep duration O 2 , a third sleep duration O 3 , and a fourth sleep duration O 4 ;

[0061] The method for obtaining the corresponding sleep duration based on the sleep state of the user within the T + a time period includes:

[0062] Record the sleep state output by the state prediction model.

[0063] If the sleep state is a light sleep state, record the light sleep duration of the user as K 1 , compare K 1 with O 1 . If K 1 < O 1 , then do not generate a light sleep optimization instruction. If K 1 > O 1 , then generate a light sleep optimization instruction;

[0064] If the sleep state is a deep sleep state, record the deep sleep duration of the user as K 2 , compare K 2 with O 2 . If K 2 < O 2 , then generate a deep sleep optimization instruction. If K 2 > O 2 , then do not generate a deep sleep optimization instruction;

[0065] If the sleep state is a dream state, record the dream duration of the user as K 3 , compare K 3 with O 3 . If K 3 < O 3 , then generate a dream optimization instruction. If K 3 > O 3, no dream optimization instruction is generated;

[0066] If the sleep state is the waking state, record the waking duration of the user as K 4 , and take K 4 Compare it with O 4 . If K 4 <O 4 , a waking optimization instruction is generated. If K 4 >O 4 , no waking optimization instruction is generated.

[0067] Further, a method for optimizing the initial sound playback data according to the comparison result to obtain a sound control strategy includes:

[0068] Step d1: Obtain the sleep state of the current user. Take the initial sound volume in the initial sound playback data as a fixed quantity, and the initial playback duration as a variable, and mark it as the current adjustment value E;

[0069] Step d2: Let E = E + W, and record the physiological evaluation coefficient under the current adjustment value E, 0 < W < 3 hours;

[0070] Step d3: Repeat step d2 in a loop. When the adjustment value E is equal to the preset playback duration threshold, obtain J physiological evaluation coefficients, and jump to step d4. J is an integer greater than zero;

[0071] Step d4: Take the initial playback duration in the initial sound playback data as a fixed quantity, and the initial sound volume as a variable, and mark it as the current adjustment value F;

[0072] Step d5: Let F = F + M, and record the physiological evaluation coefficient under the current adjustment value F, 0 < M < 70 dB;

[0073] Step d6: Repeat step d5 in a loop. When the adjustment value F is equal to the preset sound volume threshold, obtain L physiological evaluation coefficients. L is an integer greater than zero;

[0074] Step d7: Accumulate the J physiological evaluation coefficients and the L physiological evaluation coefficients to obtain R physiological evaluation coefficients, and sort the R physiological evaluation coefficients from smallest to largest in value;

[0075] Step d8: Take the sound volume adjustment value and the playback duration adjustment value corresponding to the physiological evaluation coefficient with the smallest value as the sound control strategy.

[0076] In a second aspect, the present invention provides a sleep assistance system for dynamic natural sound simulation, including:

[0077] A data analysis module, configured to obtain historical preference data of a user and external environment data, and generate a first reference coefficient based on the historical preference data of the user and the external environment data, where the historical preference data includes sound type, volume value, and playing duration;

[0078] An initial setting module, configured to obtain historical sleep cycle data of a user, and input the first reference coefficient and the historical sleep cycle data into a pre-constructed first sound training model to obtain initial sound playing data of the user;

[0079] A state determination module, configured to obtain a physiological evaluation coefficient of a user within a T time period, and determine a sleep state of the user within a T + a time period based on the physiological evaluation coefficient;

[0080] An optimization and regulation module, based on the sleep state of the user within a T + a time period, obtains a corresponding sleep duration, compares the corresponding sleep duration with a preset sleep duration, and optimizes the initial sound playing data according to the comparison result to obtain a sound regulation strategy;

[0081] The technical effects and advantages of the present invention:

[0082] 1. By deeply analyzing the historical preference data of the user, covering sound type, volume value, and playing duration, the present invention can accurately match personal preferences, greatly improving the user's comfort and compliance. At the same time, by combining external environment data, such as temperature difference, humidity difference, and sound intensity, fully considering the impact of environmental factors on sleep, the generated first reference coefficient provides a scientific and quantitative basis for subsequent strategies, making sleep assistance more in line with the actual situation. Using historical sleep cycle data to further optimize the initial sound playing data can make targeted adjustments according to the user's past sleep patterns, comprehensively realizing personalized sleep assistance customization and meeting the diverse needs of different users.

[0083] 2. The present invention determines the sleep state based on the real-time obtained physiological evaluation coefficient and compares it with the preset sleep duration to optimize the sound playing strategy, realizing dynamic and intelligent regulation of sleep assistance. During sleep, it can timely adjust sound parameters, such as sound volume and playing duration, according to the actual duration of the user's light sleep, deep sleep, dream, or waking state. By continuously cycling and testing the physiological evaluation coefficients under different adjustment values and sorting and screening, the optimal sleep regulation strategy is determined, effectively improving sleep quality, helping the user better enter the appropriate sleep stage, reducing sleep disorders, and enhancing the overall sleep health level, creating an intelligent and scientific sleep improvement plan for the user. Description of the Drawings

[0084] Figure 1 It is a flowchart of a sleep assistance method for dynamic natural sound simulation in Embodiment 1;

[0085] Figure 2 Flowchart of the method for obtaining historical sleep cycle data of the user in Embodiment 1;

[0086] Figure 3 Flowchart of obtaining the sound regulation strategy in Embodiment 1;

[0087] Figure 4 Structural schematic diagram of the sleep assistance system for dynamic natural sound simulation in Embodiment 2; Detailed implementation manners

[0088] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0089] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and thus their repeated description will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0090] It should be understood that although the terms "first", "second", etc. may be used here to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit can be called the second unit, and similarly the second unit can be called the first unit. The term "and / or" used here includes any and all combinations of one or more of the listed related items.

[0091] Embodiment 1

[0092] Please refer to Figure 1 As shown, this embodiment discloses and provides a sleep assistance method for dynamic natural sound simulation, and this method is applied to a sleep assistance device; the method includes:

[0093] Step 1: Obtain the historical preference data of the user and obtain the external environment data, and generate a first reference coefficient based on the historical preference data of the user and the external environment data, where the external environment data includes the temperature difference, humidity difference, sound intensity, etc. that the user is in per unit time;

[0094] It should be understood that: an environmental data sensor and a sound effect preference setting function are provided in the sleep aid device. The environmental data sensor includes a temperature sensor, a humidity sensor, an optical fiber sensor, and a sound sensor. The temperature sensor is used to sense the environmental temperature and optimize comfort. The humidity sensor is used to monitor the air humidity and maintain a suitable sleep environment. The optical fiber sensor is used to detect the room light intensity and adjust the sound effect or wake-up function. The sound sensor is used to sense the environmental noise and dynamically adjust the sleep aid sound effect.

[0095] It should be noted that: the historical preference data includes the sound type, volume value, and playing duration; it is set by the user through the sleep aid device according to personal preferences and recorded in the database. Natural sound effects such as the sound of gurgling streams, gentle wind sounds, and soothing piano melodies are also pre-stored in the database.

[0096] In implementation, the method for generating the first reference coefficient based on the user's historical preference data and external environmental data includes:

[0097] Step s1: Analyze the historical preference data to obtain the historical preference coefficient HPC;

[0098] Step s2: Analyze the external environmental data to obtain the environmental evaluation coefficient EAC;

[0099] Step s3: Let the current time be t, the historical preference coefficient sequence be {HPC 1 , HPC 2 ,..., HPC t}, and HPC t be the historical preference coefficient at the current time t;

[0100] The environmental evaluation coefficient sequence is {EAC 1 , EAC 2 ,..., EAC t}, and EAC t be the environmental evaluation coefficient at the current time t; the attenuation factor is

[0101] Step s4: Calculate the weighted sum of the historical preference coefficients r = 1, 2, ……, t;

[0102] And calculate the weighted sum of the environmental evaluation coefficients

[0103] Step s5: Comprehensively process the historical preference coefficient HPC and the environmental evaluation coefficient EAC to obtain the first reference coefficient, and its calculation formula is:

[0104]

[0105] In the formula, FRC represents the first reference coefficient, and σ 1 and σ 2 are weighting factors, and σ 1 +σ 2 = 1.

[0106] Among them, the method for obtaining the historical preference coefficient includes:

[0107] Step a1: Assume there are n different sound types, and each sound type corresponds to a technical score S i , where i = 1, 2,..., n;

[0108] Step a2: Perform weighted adjustment according to the selection frequency F i of the sound type by the user. The F i is the proportion of the number of times the i-th sound type is selected to the total number of selections;

[0109] Step a3: The volume value V is non-linearly mapped according to a logarithmic relationship. Assume the minimum volume value is V min , and the maximum volume value is V max . Its mapping formula is:

[0110]

[0111] In the formula, V m represents the volume value obtained after non-linearly mapping the volume value V.

[0112] Step a4: The sound duration C is processed using a piecewise function. Assume C 1 , C 2 , C 3 are different duration thresholds; when C < C 1 , C p = 0; when C 1 < C < C 2 , When C 2 < C < C 3 , When C > C 3 , C p = 1, and C p is the playing duration.

[0113] Step a5: Perform formula-based calculations on the sound type S i , the selection frequency F i , the volume value V m after non-linear mapping, and the playing duration C p to obtain the historical preference coefficient;

[0114]

[0115] In the formula, HPC represents the historical preference coefficient.

[0116] Among them, the method for obtaining the environmental evaluation coefficient includes:

[0117] Step b1: The temperature standard deviation ΔT is processed using a Gaussian function. Let the standard temperature range be [T min , T max , and the real-time temperature be T s . The calculation formula is:

[0118]

[0119] Step b1: The humidity standard deviation ΔH is processed using a Gaussian function. Let the standard humidity range be [H min , H max , and the real-time humidity be H s . The calculation formula is:

[0120]

[0121] Step b3: The sound intensity standard deviation ΔQ is processed using a power function. Let the standard sound intensity range be [Q min , Q max , and the real-time sound intensity be Q s . The calculation formula is:

[0122]

[0123] Step b4: The calculation formula for the environmental evaluation coefficient EAC is:

[0124] EAC = μ 1 ×ΔT + μ 2 ×ΔH + μ 3 ×ΔQ

[0125] In the formula, EAC represents the environmental evaluation coefficient, μ 1 , μ 2 and μ 3 are weighting factors, and μ 1 + μ 2 + μ 3 = 1.

[0126] The requirements of different users for the sleep environment vary greatly. This step considers external environmental data (such as indoor temperature, humidity, and sound intensity) and historical preference data to generate the first reference coefficient, which helps to adjust the sleep aid strategy according to the specific situation of each user.

[0127] Step 2: Obtain the user's historical sleep cycle data, and input the first reference coefficient and the historical sleep cycle data into a pre-constructed first sound training model to obtain the user's initial sound playback data;

[0128] It should be noted that: the historical sleep cycle data includes each sleep state and its corresponding sleep duration, and the sleep states include light sleep state, deep sleep state, dream state, and wake state; the initial sound playback data includes the initial music type, the initial sound volume, and the initial playback duration.

[0129] It is worth noting that: each sleep state includes a corresponding set of initial sound playback data. Specifically, the initial sound playback data corresponding to the light sleep state includes the first initial music type, the first initial sound volume, and the first initial playback duration, the initial sound playback data corresponding to the deep sleep state includes the second initial music type, the second initial sound volume, and the second initial playback duration, the initial sound playback data corresponding to the dream state includes the third initial music type, the third initial sound volume, and the third initial playback duration, and the initial sound playback data corresponding to the wake state includes the fourth initial music type, the fourth initial sound volume, and the fourth initial playback duration.

[0130] Please refer to Figure 2 As shown, in implementation, the method for obtaining the user's historical sleep cycle data includes:

[0131] Step c1: Obtain historical electroencephalogram data, use the time in the historical electroencephalogram data as the horizontal axis, and use the electroencephalogram in the historical electroencephalogram data as the vertical axis to construct an electroencephalogram time domain graph, and the electroencephalogram includes theta wave, delta wave, beta wave, and alpha wave;

[0132] It should be noted that: the electroencephalogram theta wave, delta wave, beta wave, and alpha wave are collected by a sleep monitoring device during the user's sleep.

[0133] Step c2: Divide the electroencephalogram time domain graph according to the electroencephalogram to obtain a set of historical electroencephalograms, and the set of historical electroencephalograms includes Z historical electroencephalogram waveforms, and the historical electroencephalogram waveforms include theta wave waveforms, delta wave waveforms, beta wave waveforms, and alpha wave waveforms.

[0134] Step c3: Extract the z-th historical electroencephalogram waveform in the set of historical electroencephalograms, where z is an integer greater than zero, and the initial value of z is 1;

[0135] Step c4: Extract the corresponding standard electroencephalogram waveform, calculate the similarity between the historical electroencephalogram waveform and the standard electroencephalogram waveform. If the similarity between the historical electroencephalogram waveform and the standard electroencephalogram waveform is greater than or equal to the preset electroencephalogram similarity threshold, extract the corresponding sleep duration and jump to step c5; if the similarity between the historical electroencephalogram waveform and the standard electroencephalogram waveform is less than the preset electroencephalogram similarity threshold, directly jump to step c5;

[0136] It should be noted that: The standard electroencephalogram waveform includes electroencephalogram waveforms of different sleep stages. Exemplarily, if the standard electroencephalogram waveform shows irregular theta waves, it indicates that the electroencephalogram activity is relatively low-frequency, which is the light sleep state; if the standard electroencephalogram waveform shows delta waves (slow waves), it indicates that the user's body is almost inactive and the heart rate is low, which is the deep sleep state; if the historical electroencephalogram waveform shows relatively active beta waves or alpha waves, accompanied by rapid eye movement and irregular breathing, showing the state of light sleep accompanied by muscle paralysis, it indicates the dream state; if the historical electroencephalogram waveform is relatively chaotic and frequent, accompanied by high-frequency brain waves (beta waves), it is the wake state.

[0137] It should be further noted that: The similarity algorithm includes the cosine similarity algorithm or the Euclidean distance algorithm, etc.; before calculating the similarity between the historical electroencephalogram waveform and the standard electroencephalogram waveform, preprocessing is also required, and the preprocessing includes image enhancement, image denoising, and image segmentation, etc.

[0138] Step c5: Let z = z + 1 and jump back to step c3;

[0139] Step c6: Repeat steps c3 to c5 until z = Z, then end the loop and obtain the corresponding sleep duration.

[0140] In implementation, the construction method of the first sound training model includes:

[0141] Divide the historical voice training data into a voice training set and a voice test set, and construct a regression network model. The historical voice training data includes voice feature data and corresponding initial voice playback data. The voice feature data includes a first reference coefficient and historical sleep cycle data. Use the voice feature data in the voice training set as the input of the regression network model, and use the corresponding initial voice playback data in the voice training set as the output of the regression network model to train the regression network model to obtain an initial first regression network model. With minimizing the sum of prediction accuracies as the training objective, use the voice test set to evaluate the initial first regression network model. If the sum of prediction accuracies is less than the threshold of the preset sum of accuracies, then use the corresponding initial first regression network model as the first voice training model. If the sum of prediction accuracies is greater than or equal to the threshold of the preset sum of accuracies, then use the original voice training set to retrain the initial first regression network model until the test result meets the set threshold. Among them, the calculation formula for the sum of prediction accuracies is: In the formula: MSE represents the sum of prediction accuracies, A w represents the predicted value of the w-th group of initial voice playback data in the voice test set, G w represents the actual value of the w-th group of initial voice playback data in the voice test set, and W represents the number of groups.

[0142] It should be noted that: the initial first regression network model includes a decision tree regression network model, a random forest regression network model, a support vector machine regression network model, a linear regression network model or a neural network model.

[0143] Step 3: Obtain the physiological evaluation coefficient of the user within the T time period, and determine the sleep state of the user within the T + a time period based on the physiological evaluation coefficient;

[0144] In implementation, the method for obtaining the physiological evaluation coefficient of the user within the T time period includes:

[0145] Obtain the real-time physiological characterization data of the user. The real-time physiological characterization data includes the absolute value of the heart rate difference, the absolute value of the respiratory rate difference, the type of brain wave, and the absolute value of the muscle electrical activity frequency difference;

[0146] It should be noted that: the absolute value of the heart rate difference, the absolute value of the respiratory rate difference, and the absolute value of the muscle electrical activity frequency difference are all calculated by taking the absolute value of the difference between the real-time collected heart rate value, respiratory rate, and muscle electrical activity frequency and the corresponding preset values. The smaller the corresponding absolute value of the difference, the more stable the physiological evaluation coefficient of the user and the better the sleep state.

[0147] Normalize the real-time physiological characterization data to obtain the physiological evaluation coefficient. Its calculation formula is:

[0148] SLX = γ 1 × XL + γ 2 × HX + γ 3 × Nb + γ 4 × HL

[0149] Wherein, SLX represents the physiological evaluation coefficient, XL represents the absolute value of the heart rate difference, HX represents the absolute value of the respiratory rate difference, Nb represents the electroencephalogram type, HL represents the absolute value of the muscle electrical activity frequency difference, γ 1 、γ 2 、γ 3 and γ 4 are weighting factors, and γ 1 + γ 2 + γ 3 + γ 4 = 1.

[0150] It should be noted that: before calculating the physiological evaluation coefficient, preprocess the electroencephalogram type Nb. Let Nb 1 、Nb 2 、Nb 3 and Nb 4 be different frequency band power thresholds; when Nb ∈ Nb 1 , Nb = 1; when Nb ∈ Nb 2 , Nb = 2; when Nb ∈ Nb 3 , Nb = 3; when Nb ∈ Nb 4 , Nb = 4; Nb 1 is the frequency band power of the δ wave (0.5 - 3 Hz), Nb 2 is the frequency band power of the θ wave (4 - 7 Hz), Nb 3 is the frequency band power of the α wave (8 - 13 Hz), Nb 4 is the frequency band power of the β wave (14 - 30 Hz).

[0151] It should be noted that the sleep aid device is also connected to various physiological monitoring devices in a wired or wireless manner. The physiological monitoring devices include a heart rate sensor, a respiratory sensor, an electroencephalogram device, and an electromyogram sensor. The models and types of various sensors are not specifically limited. The electroencephalogram device is used to detect the brain electrical activity of the user. Brain waves mainly include alpha waves, beta waves, theta waves, and delta waves. A heart rate monitoring device (such as a smart bracelet or a chest-worn heart rate monitor) is used to record the heart rate. During sleep, the heart rate gradually decreases. During deep sleep, the heart rate is relatively stable and low, while during the dream state, there may be slight fluctuations, and during wakefulness or light sleep, the heart rate is relatively high. A respiratory monitoring device (such as a nasal clip respiratory sensor or a pressure sensor under the mattress) is used to obtain the respiratory rate. Generally, the respiratory rate slows down during sleep, and the respiration is more regular and slow during deep sleep. The stability of the respiratory rate can be calculated, for example, the standard deviation of the respiratory rate within a T time period, as a component of the physiological evaluation coefficient. An electromyogram sensor (usually attached near the facial or limb muscles) is used to monitor muscle activity. During deep sleep and some light sleep stages, muscle activity is less, while during the rapid eye movement sleep (dream) stage, except for the eye muscles, the other muscles of the body are usually in a paralyzed state.

[0152] In implementation, the method for determining the sleep state of the user within the T+a time period based on the physiological evaluation coefficient includes:

[0153] Input the physiological evaluation coefficient into a pre-constructed state prediction model to obtain the sleep state of the user within the T+a time period.

[0154] Specifically, the specific training process of the state prediction model includes:

[0155] Pre-collect P groups of physiological evaluation coefficients, where P is an integer greater than 1. Set corresponding prediction results for the physiological evaluation coefficients. The prediction results include light sleep state, deep sleep state, dream state, and wake state. Different digital labels are set for the light sleep state, deep sleep state, dream state, and wake state. Exemplarily, set the digital label for the light sleep state as 1, the digital label for the deep sleep state as 2, the digital label for the dream state as 3; set the digital label for the wake state as 4; the prediction results corresponding to the physiological evaluation coefficients are collected by those skilled in the art during the diagnosis of historical physiological evaluation coefficients. Those skilled in the art collect P groups of different physiological evaluation coefficients and, according to actual experience, sequentially determine the corresponding prediction results for the P groups of different physiological evaluation coefficients in turn;

[0156] Mark the digital label of the prediction result as the prediction label, and convert the physiological evaluation coefficient and the corresponding prediction label into a corresponding set of feature vectors;

[0157] Use each group of feature vectors as the input of the state prediction model. The state prediction model outputs a group of prediction labels corresponding to each group of physiological evaluation coefficients, and uses the actual prediction labels corresponding to each group of physiological evaluation coefficients as the prediction target. The actual prediction labels are the prediction labels preset corresponding to the physiological evaluation coefficients. Use minimizing the sum of the prediction errors of the prediction labels corresponding to all physiological evaluation coefficients as the training target. Train the state prediction model until the sum of the prediction errors converges and then stop training. The state prediction model is specifically a deep neural network model.

[0158] Step 4: Based on the sleep state of the user during the T + a time period, obtain the corresponding sleep duration, compare the corresponding sleep duration with the preset sleep duration, and optimize the initial sound playback data according to the comparison result to obtain a sound regulation strategy.

[0159] It should be noted that: the comparison results include a light sleep optimization instruction, a deep sleep optimization instruction, a dream optimization instruction, and a wakefulness optimization instruction. The preset sleep duration includes a first sleep duration O 1 , a second sleep duration O 2 , a third sleep duration O 3 and a fourth sleep duration O 4 ; the preset sleep duration is set according to the experience of those skilled in the art, and the preset sleep duration is a number greater than zero.

[0160] In implementation, the method for obtaining the corresponding sleep duration based on the sleep state of the user during the T + a time period includes:

[0161] Record the sleep state output by the state prediction model.

[0162] If the sleep state is a light sleep state, record the user's light sleep duration as K 1 , compare K 1 with O 1 . If K 1 < O 1 , then do not generate a light sleep optimization instruction. If K 1 > O 1 , then generate a light sleep optimization instruction.

[0163] If the sleep state is a deep sleep state, record the user's deep sleep duration as K 2 , compare K 2 with O 2 . If K 2 < O 2 , then generate a deep sleep optimization instruction. If K 2 > O 2 , then do not generate a deep sleep optimization instruction.

[0164] If the sleep state is a dream state, record the dream duration of the user as K 3 , compare K 3 with O 3 . If K 3 < O 3 , generate a dream optimization instruction. If K 3 > O 3 , do not generate a dream optimization instruction;

[0165] If the sleep state is a wake state, record the wake duration of the user as K 4 , compare K 4 with O 4 . If K 4 < O 4 , generate a wake optimization instruction. If K 4 > O 4 , do not generate a wake optimization instruction;

[0166] It should be noted that: setting K 1 > O 1 , the reason for generating a light sleep optimization instruction is that if the user exceeds the preset first sleep duration, it means that the user has been in a light sleep state. Staying in a light sleep state for a long time will lead to insufficient secretion of growth hormone, affect the repair of various organs and tissues of the body, make it difficult to eliminate the body's fatigue, and is not conducive to the physical and mental health of the user.

[0167] Please refer to Figure 3 as shown. In implementation, the method for optimizing the initial sound playback data according to the comparison result to obtain a sound control strategy includes:

[0168] Step d1: Obtain the sleep state of the current user. Take the initial sound volume in the initial sound playback data as a fixed quantity, and the initial playback duration as a variable, and mark it as the current adjustment value E;

[0169] Step d2: Let E = E + W, and record the physiological evaluation coefficient under the current adjustment value E, 0 < W < 3 hours;

[0170] Step d3: Repeat step d2 in a loop. When the adjustment value E is equal to the preset playback duration threshold, obtain J physiological evaluation coefficients, and jump to step d4. J is an integer greater than zero;

[0171] Step d4: Take the initial playback duration in the initial sound playback data as a fixed quantity, and the initial sound volume as a variable, and mark it as the current adjustment value F;

[0172] Step d5: Let F = F + M, and record the physiological evaluation coefficient under the current adjustment value F, 0 < M < 70 dB;

[0173] Step d6: Repeat step d5, and when the adjustment value F is equal to the preset sound volume threshold, L physiological evaluation coefficients are obtained, where L is an integer greater than zero;

[0174] Step d7: accumulating the J physiological evaluation coefficients and the L physiological evaluation coefficients to obtain R physiological evaluation coefficients, and sorting the R physiological evaluation coefficients from small to large values;

[0175] Step d8: The sound volume adjustment value and the playback duration adjustment value corresponding to the physiological evaluation coefficient with the smallest value are used as the sound control strategy.

[0176] This embodiment can accurately match personal preferences by deeply analyzing the user's historical preference data, covering sound type, volume value and playback time, and greatly improve the user's comfort and compliance. At the same time, combined with external environmental data, such as temperature difference, humidity difference and sound intensity, the impact of environmental factors on sleep is fully considered. The generated first reference coefficient provides a scientific quantitative basis for subsequent strategies, making sleep assistance more in line with actual situations. The initial sound playback data is further optimized using historical sleep cycle data, and targeted adjustments can be made based on the user's past sleep patterns, realizing all-round personalized sleep assistance customization to meet the diverse needs of different users.

[0177] This embodiment determines the sleep state based on the physiological assessment coefficient obtained in real time, and optimizes the sound playback strategy by comparing it with the preset sleep duration, thereby realizing dynamic intelligent regulation of sleep assistance. During sleep, the sound parameters, such as sound volume and playback duration, can be adjusted in time according to the actual duration of the user's light sleep, deep sleep, dream or wake-up state. By continuously testing the physiological assessment coefficients under different adjustment values ​​and sorting and screening, the optimal sleep regulation strategy is determined, the sleep quality is effectively improved, and users are helped to better enter the appropriate sleep stage, reduce sleep disorders, and improve the overall sleep health level, creating an intelligent and scientific sleep improvement plan for users.

[0178] Example 2

[0179] See also Figure 4 As shown, this embodiment provides a dynamic natural sound simulation sleep aid system, the system includes: a data analysis module, an initial setting module, a state determination module and an optimization and control module; each module is connected by wire and / or wireless means to achieve data transmission between modules;

[0180] A data analysis module, used to obtain historical preference data of the user and external environment data, and generate a first reference coefficient based on the historical preference data of the user and the external environment data, wherein the historical preference data includes a sound type, a volume value, and a playback duration;

[0181] An initial setting module, configured to obtain historical sleep cycle data of a user, and input a first reference coefficient and the historical sleep cycle data into a pre-constructed first sound training model to obtain initial sound playback data of the user;

[0182] A state determination module, configured to obtain a physiological evaluation coefficient of the user within a T time period, and determine a sleep state of the user within a T+a time period based on the physiological evaluation coefficient;

[0183] An optimization and regulation module, based on the sleep state of the user within the T+a time period, obtains a corresponding sleep duration, compares the corresponding sleep duration with a preset sleep duration, and optimizes the initial sound playback data according to the comparison result to obtain a sound regulation strategy.

[0184] Embodiment 3

[0185] This embodiment provides an electronic device, including: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps in the above-mentioned embodiment of the sleep assistance method for dynamic natural sound simulation are implemented, such as Figure 1 The flowchart shown. Alternatively, when the processor executes the computer program, the functions of each module / unit in the above-mentioned system embodiments are implemented, such as Figure 4 The system structure diagram shown.

[0186] Exemplarily, the computer program can be divided into one or more modules / units. The one or more modules / units are stored in the memory and executed by the processor to complete this application. The one or more modules / units can be a series of computer program instruction segments capable of completing specific functions, and this instruction segment is used to describe the execution process of the computer program in the terminal device.

[0187] The memory is an internal storage unit of the terminal device, such as the hard disk or memory of the terminal device. The memory can also be an external storage device of the terminal device, such as a plug-in hard disk equipped on the terminal device, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. Further, the memory can also include both the internal storage unit and the external storage device of the terminal device. The memory is used to store the computer program and other programs and data required by the terminal device. The memory can also be used to temporarily store data that has been output or will be output.

[0188] Embodiment 4

[0189] This embodiment provides a computer-readable storage medium storing instructions that, when run on a computer, cause the computer to execute the sleep assistance method for dynamic natural sound simulation in Embodiment 1.

[0190] All the formulas involved above are calculated by removing the dimension and taking their numerical values. It is a formula obtained by software simulation of a large amount of collected data to be closest to the actual situation. The weight factors in the formula and each preset threshold in the analysis process are set by those skilled in the art according to the actual situation or obtained by simulation of a large amount of data; the magnitude of the weight factor is a specific numerical value obtained by quantifying each parameter for subsequent comparison. Regarding the magnitude of the weight factor, it depends on the amount of sample data and the processing coefficients initially set by those skilled in the art for each group of sample data; as long as it does not affect the proportional relationship between the parameters and the quantified values.

[0191] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, and all of them should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0192] Finally: The above is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A sleep aid method using dynamic natural sound simulation, characterized in that: include: Acquire historical preference data of the user and acquire external environment data, and generate a first reference coefficient based on the historical preference data of the user and the external environment data, wherein the historical preference data includes a sound type, a volume value, and a playback duration; Acquire historical sleep cycle data of the user, and input the first reference coefficient and the historical sleep cycle data into a pre-built first sound training model to obtain initial sound playback data of the user; Obtaining a physiological evaluation coefficient of the user in a time period T, and determining the sleep state of the user in a time period T+a based on the physiological evaluation coefficient; Based on the user's sleep state in the T+a time period, the corresponding sleep duration is obtained, the corresponding sleep duration is compared with the preset sleep duration, and the initial sound playback data is optimized according to the comparison result to obtain a sound control strategy.

2. The sleep aid method of dynamic natural sound simulation according to claim 1, characterized in that: The external environment data includes the temperature difference, humidity difference and sound intensity of the user within a unit time; The method for generating a first reference coefficient based on the user's historical preference data and external environment data includes: Step s1: Analyze the historical preference data to obtain the historical preference coefficient HPC; Step s2: Analyze the external environmental data to obtain the environmental assessment coefficient EAC; Step s3: Let the current time be t, and the historical preference coefficient sequence be {HPC1, HPC2,...,HPC t }, HPC t is the historical preference coefficient at the current time t; The sequence of environmental assessment coefficients is {EAC1,EAC2,...,EAC t }, EAC t is the environmental assessment coefficient at the current time t; the attenuation factor is Step s4: Calculate the weighted sum of historical preference coefficients and calculate the weighted sum of environmental assessment coefficients Step s5: Comprehensively process the historical preference coefficient HPC and the environmental assessment coefficient EAC to obtain the first reference coefficient, which is calculated as follows: Wherein, FRC represents the first reference coefficient, σ1 and σ2 are weight factors, and σ1+σ2=1.

3. The sleep aid method of dynamic natural sound simulation according to claim 2, characterized in that: The method for obtaining the historical preference coefficient includes: Step a1: Suppose there are n different sound types, each sound type corresponds to a technical score S i , i=1,2,……,n; Step a2: Frequency F according to the user's choice of sound type i Weighted adjustment is performed, the F i is the ratio of the number of times the i-th sound type is selected to the total number of times selected; Step a3: The volume value V is nonlinearly mapped according to the logarithmic relationship, and the minimum volume value is set to V min , the maximum volume value is V max , and its mapping formula is: Where V m Represents the volume value after nonlinear mapping of the volume value V; Step a4: The sound duration C is processed using a piecewise function, and C1, C2, and C3 are different duration thresholds; when C<C1, C p =0; when C1<C<C2, When C2<C<C3, When C>C3, C p =1, C p is the playback duration; Step a5: Type the sound S i , select frequency F i , the volume value V after nonlinear mapping m and playback duration C p Perform formulaic calculations to obtain historical preference coefficients; Where HPC represents the historical preference coefficient.

4. The sleep aid method of dynamic natural sound simulation according to claim 3, characterized in that: The method for obtaining the environmental assessment coefficient includes: Step b1: The temperature standard deviation ΔT is processed using a Gaussian function, assuming that the standard temperature range is [T min ,T max ], the real-time temperature is T s , the calculation formula is: Step b1: The humidity standard deviation ΔH is processed using a Gaussian function, assuming that the standard humidity range is [H min ,H max ], real-time humidity is H s , the calculation formula is: Step b3: The sound intensity standard deviation ΔQ is processed using a power function, assuming that the standard sound intensity range is [Q min ,Q max ], the real-time sound intensity is Q s , the calculation formula is: Step b4: The calculation formula of the environmental assessment coefficient EAC is: EAC=μ1×ΔT+μ2×ΔH+μ3×ΔQ Where EAC represents the environmental assessment coefficient, μ1, μ2 and μ3 are weight factors, and μ1+μ2+μ3=1.

5. The sleep aid method of dynamic natural sound simulation according to claim 4, characterized in that: The historical sleep cycle data includes each sleep state and its corresponding sleep duration, and the sleep state includes a light sleep state, a deep sleep state, a dream state, and a wake-up state; The initial sound playing data includes the initial music type, the initial sound volume and the initial playing time; Methods for obtaining the user's historical sleep cycle data include: Step c1: Acquire historical brain wave data, and construct a brain wave time domain diagram with the time in the historical brain wave data as the horizontal axis and the brain wave in the historical brain wave data as the vertical axis. The brain wave includes theta wave, delta wave, beta wave and alpha wave; Step c2: dividing the EEG time domain graph according to the EEG waves to obtain a historical EEG wave set, wherein the historical EEG wave set includes Z historical EEG wave waveforms, and the historical EEG wave waveforms include a θ wave waveform, a δ wave waveform, a β wave waveform, and an α wave waveform; Step c3: extract the zth historical brain wave waveform in the historical brain wave set, where z is an integer greater than zero and the initial value of z is 1; Step c4: extract the corresponding standard EEG waveform, calculate the similarity between the historical EEG waveform and the standard EEG waveform, if the similarity between the historical EEG waveform and the standard EEG waveform is greater than or equal to the preset EEG similarity threshold, extract the corresponding sleep duration and jump to step c5; if the similarity between the historical EEG waveform and the standard EEG waveform is less than the preset EEG similarity threshold, directly jump to step c5; Step c5: set z=z+1 and jump back to step c3; Step c6: Repeat steps c3 to c5 until z=Z, then end the loop and obtain the corresponding sleep duration.

6. The sleep aid method of dynamic natural sound simulation according to claim 5, characterized in that: The method for constructing the first sound training model includes: The historical sound training data is divided into a sound training set and a sound test set, and a regression network model is constructed. The historical sound training data includes sound feature data and corresponding initial sound playback data; the sound feature data includes a first reference coefficient and historical sleep cycle data; the sound feature data in the sound training set is used as the input of the regression network model, and the initial sound playback data corresponding to the sound feature data in the sound training set is used as the output of the regression network model, and the regression network model is trained to obtain an initial first regression network model. The training goal is to minimize the sum of prediction accuracies, and the initial first regression network model is evaluated using the sound test set. If the sum of prediction accuracies is less than a preset threshold of the sum of accuracies, the corresponding initial first regression network model is used as the first sound training model. If the sum of prediction accuracies is greater than or equal to a preset threshold of the sum of accuracies, the initial first regression network model is trained again using the original sound training set until the test result meets the set threshold; the initial first regression network model is a decision tree regression network model, a random forest regression network model, a support vector machine regression network model, a linear regression network model or a neural network model.

7. The sleep aid method of dynamic natural sound simulation according to claim 6, characterized in that: The method for obtaining the physiological evaluation coefficient of the user in the T time period includes: Acquire real-time physiological characterization data of the user, wherein the real-time physiological characterization data includes an absolute value of a heart rate difference, an absolute value of a breathing rate difference, an EEG type, and an absolute value of a muscle electrical activity frequency difference; Normalizing the real-time physiological characterization data to obtain physiological evaluation coefficients; The method for determining the sleep state of the user in the T+a time period based on the physiological evaluation coefficient includes: The physiological assessment coefficient is input into a pre-built state prediction model to obtain the sleep state of the user in the T+a time period, wherein the specific training process of the state prediction model includes: Pre-collect P groups of physiological evaluation coefficients, where P is an integer greater than 1, and set corresponding prediction results for the physiological evaluation coefficients, the prediction results including a light sleep state, a deep sleep state, a dream state, and a wake-up state, and set different digital labels for the light sleep state, the deep sleep state, the dream state, and the wake-up state, and mark the digital labels of the prediction results as prediction labels, and convert the physiological evaluation coefficients and the corresponding prediction labels into a corresponding set of feature vectors; Each group of feature vectors is used as the input of a state prediction model. The state prediction model takes a group of prediction labels corresponding to each group of physiological evaluation coefficients as output, and takes the actual prediction labels corresponding to each group of physiological evaluation coefficients as prediction targets, where the actual prediction labels are pre-set prediction labels corresponding to the physiological evaluation coefficients; minimizing the sum of prediction errors of prediction labels corresponding to all physiological evaluation coefficients is used as a training target; the state prediction model is trained until the sum of prediction errors converges, and the training is stopped. The state prediction model is a deep neural network model.

8. The sleep aid method of dynamic natural sound simulation according to claim 7, characterized in that: The comparison result includes a light sleep optimization instruction, a deep sleep optimization instruction, a dream optimization instruction, and a wakefulness optimization instruction; the preset sleep duration includes a first sleep duration O1, a second sleep duration O2, a third sleep duration O3, and a fourth sleep duration O4; Based on the sleep state of the user in the T+a time period, the method for obtaining the corresponding sleep duration includes: Record the sleep state output by the state prediction model; If the sleep state is a light sleep state, the light sleep duration of the user is recorded as K1, and K1 is compared with O1. If K1<O1, no light sleep optimization instruction is generated; if K1>O1, a light sleep optimization instruction is generated; If the sleep state is a deep sleep state, the user's deep sleep duration is recorded as K2, and K2 is compared with O2. If K2 < O2, a deep sleep optimization instruction is generated; if K2 > O2, no deep sleep optimization instruction is generated; If the sleep state is a dream state, the dream duration of the user is recorded as K3, and K3 is compared with O3. If K3<O3, a dream optimization instruction is generated; if K3>O3, no dream optimization instruction is generated; If the sleeping state is the awakening state, the user's awakening time is recorded as K4, and K4 is compared with O4. If K4<O4, a wake-up optimization instruction is generated; if K4>O4, no wake-up optimization instruction is generated.

9. The sleep aid method of dynamic natural sound simulation according to claim 8, characterized in that: The method of optimizing the initial sound playback data according to the comparison result and obtaining the sound control strategy includes: Step d1: obtaining the current user's sleep state, taking the initial sound volume in the initial sound playback data as a fixed amount, and the initial playback duration as a variable, and marking them as the current adjustment value E; Step d2: Let E=E+W, and record the physiological evaluation coefficient under the current adjustment value E, 0<W<3 hours; Step d3: Repeat step d2, when the adjustment value E is equal to the preset playback time threshold, obtain J physiological evaluation coefficients, and jump to step d4, where J is an integer greater than zero; Step d4: taking the initial playing time in the initial sound playing data as a fixed value and the initial sound volume as a variable, and marking them as the current adjustment value F; Step d5: Let F = F + M, and record the physiological evaluation coefficient under the current adjustment value F, 0 < M < 70 decibels; Step d6: Repeat step d5, and when the adjustment value F is equal to the preset sound volume threshold, L physiological evaluation coefficients are obtained, where L is an integer greater than zero; Step d7: accumulating the J physiological evaluation coefficients and the L physiological evaluation coefficients to obtain R physiological evaluation coefficients, and sorting the R physiological evaluation coefficients from small to large values; Step d8: The sound volume adjustment value and the playback duration adjustment value corresponding to the physiological evaluation coefficient with the smallest value are used as the sound control strategy.

10. A dynamic natural sound simulation sleep assistance system, used for implementing the dynamic natural sound simulation sleep assistance method according to any one of claims 1 to 9, characterized in that: include: A data analysis module, used to obtain historical preference data of the user and external environment data, and generate a first reference coefficient based on the historical preference data of the user and the external environment data, wherein the historical preference data includes a sound type, a volume value, and a playback duration; An initial setting module, used to obtain historical sleep cycle data of the user, and input the first reference coefficient and the historical sleep cycle data into a pre-built first sound training model to obtain initial sound playback data of the user; A state determination module, used to obtain a physiological evaluation coefficient of the user in a time period T, and determine the sleep state of the user in a time period T+a based on the physiological evaluation coefficient; The optimization and control module obtains the corresponding sleep duration based on the user's sleep state in the T+a time period, compares the corresponding sleep duration with the preset sleep duration, optimizes the initial sound playback data according to the comparison result, and obtains the sound control strategy.

Images