Adjusting method of intelligent mattress

Abstract

The invention discloses an adjusting method of an intelligent mattress, and relates to the technical field of data identification, and the method comprises the steps: collecting feature data of a user in a sleeping process through a sensor deployed in the intelligent mattress, recognizing the feature data through a preset sleep state monitoring model, and determining a sleep stage of the user, according to the method, the recognition accuracy of the sleep stage is improved through fusion recognition of various data, then the parameters of the intelligent mattress are optimized by adopting the preset parameter adjustment model based on the sleep stage of the user, and the optimized parameters of the intelligent mattress are determined, so that the adjustment real-time performance of the intelligent mattress can be improved, and the user experience is improved. Finally, the intelligent mattress is adjusted on the basis of the optimized parameters of the intelligent mattress, and the adjustment strategy is optimized in a targeted mode, so that deep sleep is promoted to the maximum extent, and the overall sleep quality and health level of a user are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data recognition, in particular to a method for adjusting an intelligent mattress. BACKGROUND

[0002] Sleep is an important physiological process for human body to recover energy and maintain health. With the acceleration of life pace and the increase of work pressure, sleep quality problems are increasingly common, such as difficulty falling asleep, light sleep, and insufficient deep sleep time, which can seriously affect people's daytime mental state and long-term physical health. In order to improve sleep quality, intelligent mattresses have emerged as the times require. Existing intelligent mattresses usually have some basic functions, such as judging whether a user is on the bed through a simple pressure sensor, or adjusting the hardness and temperature of the mattress manually or by timing. However, these technologies have the following obvious shortcomings:

[0003] Lagging adjustment and non-intelligence: The adjustment of many intelligent mattresses relies on the user's feedback or preset fixed program, and cannot be adjusted in real time and dynamically during the user's sleep. For example, when the user tosses and turns due to improper mattress support, the system cannot immediately perceive and optimize, resulting in the user missing the best sleep opportunity or frequently waking up from deep sleep.

[0004] Insufficient sleep state recognition accuracy: Some intelligent mattresses claim to be able to monitor sleep, but their algorithms are relatively simple, usually only based on single data such as body movement or heart rate for rough judgment (such as "awake" or "sleep"), and cannot accurately identify fine sleep stages such as light sleep, deep sleep, and rapid eye movement sleep. Inaccurate sleep state recognition is the root cause of ineffective adjustment.

[0005] Lack of targeted adjustment strategy: Existing adjustment strategies are often universal and do not differentiate according to the specific sleep stage of the user. For example, the best support parameters and temperature of the mattress should be different when the user is trying to fall asleep (light sleep stage) and has entered deep sleep stage. Uniform or random adjustment not only may be ineffective, but also may interfere with the user's normal sleep process. SUMMARY

[0006] The present application provides a method for adjusting an intelligent mattress, aiming to solve the problems of lagging adjustment, inaccurate sleep recognition, lack of targeted adjustment strategy, and low efficiency of model training of existing intelligent mattresses, so as to realize precise and efficient personalized sleep intervention and significantly improve the user's deep sleep quality.

[0007] The present application provides a method for adjusting an intelligent mattress, comprising:

[0008] The feature data of the user during sleep is collected by sensors deployed in the smart mattress, and the feature data is identified by a pre-set sleep state monitoring model to determine the sleep stage of the user.

[0009] Based on the sleep stage of the user, a pre-set parameter adjustment model is used to optimize the parameters of the smart mattress, and the optimized smart mattress parameters are determined.

[0010] Based on the optimized smart mattress parameters, the smart mattress is adjusted to enable the user to quickly enter deep sleep or increase the deep sleep time.

[0011] In a possible implementation, the feature data of the user during sleep is collected by sensors deployed in the smart mattress, including:

[0012] During the user's sleep, the heart rate, respiration rate and body movement frequency at M sampling time points are continuously collected by sensors deployed in the smart mattress based on a pre-set data sampling frequency, and the feature data is obtained as time series data.

[0013] In a possible implementation, the method of the pre-set sleep state monitoring model includes:

[0014] The pre-stored sample feature data and the sample sleep stage label corresponding to the sample feature data are collected.

[0015] A convolutional neural network is used to construct a sleep state monitoring model, and the parameters of the sleep state monitoring model are initialized to generate a training population.

[0016] For any individual in the training population, the pre-stored sample feature data is used as input, and the sample sleep stage label corresponding to the sample feature data is used as expected output, the fitness of each individual is obtained, and the optimal individual is determined based on the fitness of each individual.

[0017] For any individual, the individual is subjected to group information learning by using a sinusoidal group information interaction method based on the optimal individual, and the individual after group information learning is obtained.

[0018] For any individual after group information learning, the individual is subjected to solution space collaborative exploration by using an individual collaborative information interaction method, and the individual after collaborative exploration is obtained.

[0019] For any individual after collaborative exploration, the individual is subjected to selective exploration by using an adaptive range regulation method, and the individual after selective exploration is obtained.

[0020] Determine whether the current number of training iterations is greater than or equal to the maximum number of training iterations. If so, redetermine the optimal individual based on the individuals after the selective exploration. Otherwise, return to the step of determining the optimal individual.

[0021] The parameters of the newly determined optimal individuals are used as the target parameters of the sleep state monitoring model, and the sleep state monitoring model is deployed based on the target parameters to obtain the pre-set sleep state monitoring model.

[0022] In one possible implementation, based on the optimal individual, a sinusoidal group information interaction method is used to perform group information learning on the individual to obtain an individual after group information learning, including:

[0023] Based on all individuals, the centroid of the training population is obtained as follows:

[0024]

[0025] in, This represents the k-th individual in the iter-th training process. Let f represent the centroid of the training population in the iterth training process, k = 1, 2, ..., NP, where NP represents the total number of individuals. k express fitness, f sum This represents the total fitness of all individuals;

[0026] Based on the centroid individual of the training population and the optimal individual, the sinusoidal interaction information is obtained as follows:

[0027]

[0028] in, express The corresponding sinusoidal interaction information is as follows: rand1 represents the first random number between [0, π], rand2 represents the random number between [0, 2π], π represents pi, and λ represents the information segmentation factor, which is set to...

[0029] Based on the sinusoidal interaction information, group information learning is performed on the individual to obtain the individual after group information learning:

[0030]

[0031] in, This represents an individual after learning the information from the k-th group.

[0032] In one possible implementation, for any individual after learning group information, an individual collaborative information interaction method is used to perform collaborative exploration of the solution space on the individual, resulting in an individual after collaborative exploration, including:

[0033] Determine the current training iteration and obtain the learning rate based on the current training iteration:

[0034]

[0035] Where, η iter Let η represent the learning rate for the iterth training iteration, with an initial value of 0.1; iter+1 Let t represent the learning rate for the (iter+1)th training iteration. max This represents the maximum number of training iterations, and α represents the learning rate adjustment factor, which is set to 10.

[0036] For any individual after learning information from a group, randomly match that individual with another individual after learning information from another group to obtain the collaborative exploration individual corresponding to each individual;

[0037] For any individual after learning from the group information, the collaborative exploration information of the solution space is obtained based on the individual's corresponding collaborative exploration partner and the learning rate:

[0038]

[0039] in, This represents the collaborative exploration information in the solution space corresponding to the individual after learning the group information of the i-th group during the (iter+1)-th training process. Let represent the individual after learning the group information from the i-th group during the iter-th training process. express Corresponding collaborative exploration individuals, This represents the optimal individual, rand3 represents the third random number between (0,1), and rand4 represents the fourth random number between (0,1). express and The Euclidean distance between them express and The Euclidean distance between them;

[0040] For any individual after learning the group information, based on the cooperative exploration information in the solution space, the individual is subjected to cooperative exploration in the solution space to obtain the individual after cooperative exploration:

[0041]

[0042] in, represents the individual after the ith cooperative exploration.

[0043] In a possible implementation, for any individual after the cooperative exploration, the adaptive range regulation method is used for selective exploration of the individual, to obtain an individual after selective exploration, comprising:

[0044] The adaptive range regulation factor is obtained based on the current training number, and is:

[0045]

[0046] wherein β iter represents the adaptive range regulation factor in the iter training process, and the initial value is set to 0.15; β iter+1 represents the adaptive range regulation factor in the iter+1 training process.

[0047] The exploration range is obtained based on the adaptive range regulation factor, and is:

[0048]

[0049] wherein, represents the d-dimensional parameter of the individual after the mth cooperative exploration in the iter training process, and d = 1, 2, …, D, D represents the total dimension of the parameter, represents the upper limit of the corresponding search range, represents the lower limit of the corresponding search range.

[0050] For any individual after the cooperative exploration, the individual is selectively explored according to the exploration range, to obtain an individual after selective exploration:

[0051]

[0052] wherein, represents the d-dimensional parameter of the individual after the mth selective exploration, and rand5 represents the fifth random number between 0 and 1.

[0053] In a possible implementation, before the parameter of the smart mattress is optimized by using the pre-set parameter adjustment model, the method further comprises:

[0054] Collecting basic data of the user, parameters of the smart mattress, entering time of deep sleep, and maintaining time of deep sleep;

[0055] Training the first parameter adjustment model by taking the basic data and the parameters of the smart mattress as inputs, and taking the entering time of deep sleep as the expected output.​​

[0056] training a second parameter adjustment model based on the base data and the smart mattress parameters as inputs and the time of maintaining deep sleep as the expected output.

[0057] In a possible implementation, based on the sleep stage of the user, a pre-set parameter adjustment model is used to optimize the parameters of the smart mattress, and the optimized smart mattress parameters are determined, including:

[0058] when the sleep stage of the user is non-deep sleep, the parameters of the smart mattress are optimized by a smart optimization algorithm, and the time of entering deep sleep is evaluated by the pre-set parameter adjustment model during the optimization process, the target parameters with the shortest time of entering deep sleep are output, and the optimized smart mattress parameters are obtained;

[0059] when the sleep stage of the user is deep sleep, the parameters of the smart mattress are optimized by a smart optimization algorithm, and the time of maintaining deep sleep is evaluated by the pre-set parameter adjustment model during the optimization process, the target parameters with the longest time of maintaining deep sleep are output, and the optimized smart mattress parameters are obtained.

[0060] In a possible implementation, the smart optimization algorithm is set as a particle swarm optimization algorithm or a pigeon swarm optimization algorithm.

[0061] In a possible implementation, the smart mattress parameters include one or more of temperature, waist support force, and shoulder support force.

[0062] Beneficial effects:

[0063] The application provides a smart mattress adjustment method. The feature data of a user during sleep is collected by a sensor deployed in a smart mattress, the feature data is identified by a pre-set sleep state monitoring model, the sleep stage of the user is determined, the identification accuracy of the sleep stage is improved through multiple data fusion identifications, then based on the sleep stage of the user, a pre-set parameter adjustment model is used to optimize the parameters of the smart mattress, the optimized smart mattress parameters are determined, the real-time adjustment of the smart mattress can be improved, finally the smart mattress is adjusted based on the optimized smart mattress parameters, the adjustment strategy is optimized in a targeted manner, and thus the deep sleep is maximally promoted, and the overall sleep quality and health level of the user are effectively improved. BRIEF DESCRIPTION OF DRAWINGS

[0064] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the description of the embodiments of the present application will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0065] Figure 1 is a flowchart of a method for adjusting an intelligent mattress according to an embodiment of the present application. DETAILED DESCRIPTION

[0066] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0067] As shown in Figure 1 , the present application provides a method for adjusting an intelligent mattress, comprising:

[0068] S101, collecting feature data of a user during sleep by a sensor deployed in the intelligent mattress, and identifying the feature data by a pre-set sleep state monitoring model to determine the sleep stage of the user.

[0069] The intelligent mattress can integrate various sensors, such as piezoelectric film sensors (for monitoring heart rate and respiration), pressure sensor arrays (for monitoring body movement and sleeping posture), temperature sensors, etc., to continuously collect physiological signals of the user at a pre-set frequency of 1 Hz. The continuously collected heart rate, respiration rate and body movement frequency data at M time points (e.g. M = 60, i.e. 1 minute of data) can be combined into a three-dimensional time series data vector as the feature data for this identification.

[0070] The collected feature data can then be input into a pre-deployed sleep state monitoring model. The model is a convolutional neural network that can automatically extract deep features from time series data and output a sleep stage classification result. The sleep stage can be divided into wakefulness, light sleep (N1, N2 stage), deep sleep (N3 stage), etc. The model output result is, for example, light sleep.

[0071] S102, based on the sleep stage of the user, optimizing the parameters of the intelligent mattress by a pre-set parameter adjustment model to determine the optimized intelligent mattress parameters.

[0072] Different user sleep stages need to use different parameters, so that the optimized smart mattress parameters can realize accurate adjustment of the smart mattress.

[0073] The parameters of the smart mattress can be physical parameters applied by the smart mattress, such as the overall temperature of the mattress, the waist support force (such as air bag pressure value, unit kPa), the shoulder support force (such as air bag pressure value, unit kPa), etc. The pre-set parameter adjustment model can be a pre-trained model for evaluating the parameters of the smart mattress. By evaluating the parameters of the smart mattress through the pre-set parameter adjustment model, the best evaluated parameters of the smart mattress can be found in the optimization process, and the optimized smart mattress parameters are obtained.

[0074] S103, based on the optimized smart mattress parameters, adjusting the smart mattress to make the user quickly enter deep sleep or improve the deep sleep time.

[0075] After the processor of the smart mattress obtains the optimized smart mattress parameters, it can send control instructions to the physical adjustment mechanism of the smart mattress.

[0076] For example, temperature adjustment: control the heating / cooling elements in the mattress to adjust the surface temperature of the mattress to the temperature value in the optimized smart mattress parameters.

[0077] Waist / shoulder support force adjustment: control the air bag or mechanical structure in the mattress to accurately adjust the support force of the waist and shoulder area to the value in the optimized smart mattress parameters. The adjustment process is smooth and quiet, avoiding disturbing the user. After the adjustment is completed, the system returns to step S101 to continue collecting the next round of data and performing the next round of monitoring-optimization-adjustment cycle until the user gets up.

[0078] Optionally, assuming that the smart mattress supports back pitch angle adjustment, when the user snores and lies on his back, the pitch angle can be adjusted to alleviate the user's snoring symptoms. At the same time, in order to avoid misjudgment caused by the user playing mobile phones or watching TV on the bed, a start sleeping button can be set in the application. After the user clicks the start sleeping button, the smart mattress receives the message and starts the sleep improvement work. If the user has not started sleeping, the smart mattress can run according to the user's manual adjustment parameters, so that the adjustment of the smart mattress is more accurate.

[0079] In one possible implementation, the feature data of the user during sleep is collected by the sensors deployed in the smart mattress, including:

[0080] During the user's sleep, the heart rate, respiration rate and body movement frequency at M sampling time points are continuously collected by the sensors deployed in the smart mattress based on a pre-set data sampling frequency, to obtain the feature data as time series data.

[0081] The characteristic data can be acquired through a piezoelectric signal or through a smart wearable device. This technology is a common prior art, and the technical solutions of the embodiments of the present application do not involve improvements to this part. The characteristic data can be acquired by using the prior art.

[0082] In a possible implementation, the method for pre-setting a sleep state monitoring model comprises:

[0083] The pre-stored sample characteristic data and sample sleep stage labels corresponding to the sample characteristic data are collected.

[0084] A convolutional neural network is used to construct a sleep state monitoring model, and the parameters of the sleep state monitoring model are initialized to generate a training population.

[0085] The parameters of the sleep state monitoring model (such as the connection weights between network layers) can be initialized by referring to the initialization methods of the particle swarm algorithm or the pigeon swarm algorithm, so as to generate the training population.

[0086] For any individual in the training population, the pre-stored sample characteristic data is taken as the input, and the sample sleep stage labels corresponding to the sample characteristic data are taken as the expected output. The fitness of each individual is acquired, and the optimal individual is determined based on the fitness of each individual.

[0087] For example, the pre-stored sample characteristic data is taken as the input, and the sample sleep stage labels corresponding to the sample characteristic data are taken as the expected output. The loss function value of the individual is acquired by using a cross-entropy loss function or a root mean square loss function. After the loss function value is added to a constant (such as 0.001) and the reciprocal is taken, the fitness of the individual can be obtained. The individual with the maximum fitness is the optimal individual.

[0088] For any individual, the individual is subjected to group information learning by using a sine group information interaction method based on the optimal individual, to obtain the individual after group information learning.

[0089] For any individual after group information learning, the individual is subjected to solution space cooperative exploration by using an individual cooperative information interaction method, to obtain the individual after cooperative exploration.

[0090] For any individual after cooperative exploration, the individual is subjected to selective exploration by using an adaptive range regulation method, to obtain the individual after selective exploration.

[0091] Determine whether the current number of training iterations is greater than or equal to the maximum number of training iterations. If so, redetermine the optimal individual based on the individuals after the selective exploration. Otherwise, return to the step of determining the optimal individual.

[0092] The parameters of the newly determined optimal individuals are used as the target parameters of the sleep state monitoring model, and the sleep state monitoring model is deployed based on the target parameters to obtain the pre-set sleep state monitoring model.

[0093] Optionally, after updating the individual, an over-limit handling operation can be performed to ensure the effectiveness of algorithm training.

[0094] This application's embodiments employ a convolutional neural network model trained based on an improved optimization algorithm, enabling high-precision identification of a user's fine sleep stages from multi-dimensional time-series data such as heart rate, respiration, and body movement. The model training method integrates sinusoidal group information interaction, individual collaborative information interaction, and adaptive range adjustment, effectively avoiding the problem of traditional algorithms easily getting trapped in local optima. This improves the model's convergence speed and recognition accuracy, laying a solid foundation for subsequent precise adjustments.

[0095] In one possible implementation, based on the optimal individual, a sinusoidal group information interaction method is used to perform group information learning on the individual to obtain an individual after group information learning, including:

[0096] Based on all individuals, the centroid of the training population is obtained as follows:

[0097]

[0098] in, This represents the k-th individual in the iter-th training process. Let f represent the centroid of the training population in the iterth training process, k = 1, 2, ..., NP, where NP represents the total number of individuals. k express fitness, f sum This represents the total fitness of all individuals;

[0099] Based on the centroid individual of the training population and the optimal individual, the sinusoidal interaction information is obtained as follows:

[0100]

[0101] in, express The corresponding sinusoidal interaction information is as follows: rand1 represents the first random number between [0, π], rand2 represents the random number between [0, 2π], π represents pi, and λ represents the information segmentation factor, which is set to...

[0102] According to the sinusoidal interaction information, population information learning is performed on the individual, and an individual after population information learning is obtained.

[0103]

[0104] wherein, represents the kth individual after population information learning.

[0105] By the sinusoidal population information interaction method, the performance of the training algorithm is significantly improved. The population barycenter individual guided search is used by fitness weighting, the global exploration and local development are balanced, the premature convergence problem of the traditional algorithm easily falling into local optimum is effectively avoided, and the ability to find the global optimal solution is enhanced. Secondly, the learning step and direction are dynamically adjusted by the introduced sinusoidal function, so that the parameter training process is smooth and gradual, the interference caused by parameter mutation on the training process is avoided, and the local exploration ability of the algorithm can be effectively improved.

[0106] In a possible implementation, for any individual after population information learning, an individual cooperation information interaction method is used to perform solution space cooperation exploration on the individual, and an individual after cooperation exploration is obtained, including:

[0107] The current training number is determined, and the learning rate is obtained based on the current training number:

[0108]

[0109] wherein, η iter represents the learning rate corresponding to the iter training process, and the initial value is set to 0.1; η iter+1 represents the learning rate corresponding to the iter+1 training process, t max represents the maximum training number, and a represents the learning rate control factor and is set to 10;

[0110] For any individual after population information learning, a random other individual after population information learning is matched to the individual, and a cooperation exploration individual corresponding to each individual is obtained.

[0111] For any individual after population information learning, the solution space cooperation exploration information is obtained according to the cooperation exploration individual corresponding to the individual and the learning rate.

[0112]

[0113] wherein, represents the solution space cooperation exploration information corresponding to the ith individual after population information learning in the iter+1 training process, represents an individual after the i-th population information learning in the iter-th training process, represents a corresponding cooperative exploration individual, represents an optimal individual, rand3 represents a third random number between (0, 1), and rand4 represents a fourth random number between (0, 1), represents the Euclidean distance between and represents the Euclidean distance between and

[0114] For any individual after population information learning, the individual is cooperatively explored in the solution space according to the cooperative exploration information of the solution space, and a cooperatively explored individual is obtained as follows:

[0115]

[0116] wherein, represents the i-th cooperatively explored individual.

[0117] Through the individual cooperative information interaction method, the distance information between individuals can be effectively utilized for cooperative information interaction, and the fine exploration ability of the algorithm in the solution space is significantly enhanced. First, the core advantage lies in the design of a learning rate that nonlinearly decays with the number of training times. This mechanism enables the algorithm to perform large-scale solution space exploration in the early stage (large learning rate) to quickly locate high-quality regions, and to shift to small-scale fine development in the later stage (small learning rate) to accurately approach the optimal solution, effectively balancing global exploration and local development. Second, by randomly matching cooperative individuals and introducing guidance information based on the Euclidean distance, individuals not only learn from the current optimal individual, but also exchange information with random individuals. This cooperative mode enriches population diversity, avoids stagnation of the algorithm, and effectively prevents premature convergence.

[0118] In a possible implementation, for any cooperatively explored individual, an adaptive range regulation method is used to selectively explore the individual, and a selectively explored individual is obtained, including:

[0119] Based on the current number of training times, an adaptive range regulation factor is obtained as follows:

[0120]

[0121] wherein, β iter represents the adaptive range regulation factor in the iter-th training process, and the initial value is set to 0.15; β iter+1 represents the adaptive range regulation factor in the iter+1-th training process.

[0122] Based on the adaptive range regulation factor, an exploration range is obtained:

[0123]

[0124] wherein, represents the d-dimensional parameter of the individual after the mth cooperative exploration in the iter training process, and d = 1, 2, …, D, D represents the total dimension of the parameter, represents the corresponding upper limit of the search range, represents the corresponding lower limit of the search range;

[0125] For any individual after cooperative exploration, according to the exploration range, the individual is selectively explored, and the individual after selective exploration is obtained:

[0126]

[0127] wherein, represents the d-dimensional parameter of the individual after the mth selective exploration, and rand5 represents the fifth random number between 0 and 1.

[0128] Through the adaptive range regulation method, the fine and intelligent management of the training process is realized, and the convergence performance and solution quality of the algorithm are significantly improved. The core advantage is that the exploration range can be dynamically and nonlinearly contracted with the increase of the training number. In the early stage of optimization, the larger exploration range encourages the individual to conduct global search in the vast solution space, effectively avoiding falling into local optimum. With the iteration, the exploration range gradually shrinks, and the search behavior of the algorithm smoothly transitions from large-scale exploration to fine local development, so that the population can more accurately gather to the global optimal solution. This coarse-to-fine search strategy not only ensures the discovery probability of the global optimal solution, but also speeds up the convergence speed of the algorithm in the later stage.

[0129] The embodiments of the present application can effectively improve the model training effect through the mutual cooperation of the above-mentioned several methods, so that the model after training can accurately identify the sleep stage.

[0130] In a possible implementation, before adjusting the parameters of the intelligent mattress by using the pre-set parameter adjustment model, the method further includes:

[0131] Collecting the basic data of the user, the parameters of the intelligent mattress, the entering time of deep sleep, and the maintaining time of deep sleep.

[0132] The first parameter adjustment model is trained by taking the basic data and the smart mattress parameters as inputs and taking the entering time of deep sleep as the expected output.

[0133] The second parameter adjustment model is trained by taking the basic data and the smart mattress parameters as inputs and taking the maintaining time of deep sleep as the expected output.

[0134] The basic data of the user can include age, gender, weight, average daily sleep time in the past week, height and other related data, and can also include pressure values measured by professional equipment and other related data.

[0135] In a possible implementation, the smart mattress parameters include one or more of temperature, waist support force and shoulder support force. The waist and shoulder of the user can be identified through pressure distribution, so that the support force is adjusted to ensure the sleep comfort of the user.

[0136] In a possible implementation, based on the sleep stage of the user, a pre-set parameter adjustment model is used to optimize the parameters of the smart mattress, and the optimized smart mattress parameters are determined, including:

[0137] When the sleep stage of the user is non-deep sleep, the parameters of the smart mattress are optimized through a smart optimization algorithm, and the entering time of deep sleep is evaluated through the pre-set parameter adjustment model during the optimization process, the target parameter with the shortest entering time is output, and the optimized smart mattress parameters are obtained.

[0138] When the sleep stage of the user is deep sleep, the parameters of the smart mattress are optimized through a smart optimization algorithm, and the maintaining time of deep sleep is evaluated through the pre-set parameter adjustment model during the optimization process, the target parameter with the longest maintaining time is output, and the optimized smart mattress parameters are obtained.

[0139] In a possible implementation, the smart optimization algorithm is set as a particle swarm optimization algorithm or a pigeon swarm optimization algorithm.

[0140] For example, the parameters of the smart mattress can be encoded as particles, and the basic data and the smart mattress parameters (i.e., the parameters contained in the particles) can be identified through the pre-set parameter adjustment model, and the entering time corresponding to each particle can be predicted, so that the target parameter with the shortest entering time can be obtained through the particle swarm optimization algorithm. The principle of outputting the target parameter with the longest maintaining time is the same, which will not be described here. The parameter can be quickly optimized through the particle swarm algorithm, and the real-time performance of parameter adjustment is improved.

[0141] The embodiment of the application creatively proposes a phased differentiated adjustment strategy. When the user is in a non-deep sleep stage, parameter optimization is performed with the goal of quickly entering deep sleep; when the user is in a deep sleep stage, parameter optimization is performed with the goal of prolonging deep sleep time. This targeted adjustment strategy makes each adjustment of the mattress serve the current most core sleep demand, avoiding ineffective or interfering adjustment, and achieving truly appropriate measures for the time.

[0142] The embodiment of the application constructs an intelligent closed-loop system of monitoring-analysis-decision-execution-remonitoring, which can monitor the user state in real time and continuously, and dynamically adjust the mattress parameters to adapt to the continuous changes of the user's body and sleep state during sleep, realizing the leap from passive adaptation to active service. By combining the user's basic data (age, gender, etc.) and the trained parameter adjustment model, the embodiment of the application can provide personalized mattress parameter recommendations for different users. Combined with the intelligent optimization algorithm for parameter optimization, the most suitable optimal parameter combination for the current user and current state can be found, thereby maximizing the promotion of deep sleep and effectively improving the overall sleep quality and health level of the user.

[0143] The application provides a kind of adjustment method of intelligent mattress, feature data of user in sleeping process is collected by sensor deployed in intelligent mattress, and the feature data is identified by pre-set sleep state monitoring model, determine the sleep stage of user, improve the identification accuracy of sleep stage by a variety of data fusion identification, then based on the sleep stage of the user, the parameter of the intelligent mattress is optimized using pre-set parameter adjustment model, determine the intelligent mattress parameter after optimization, can improve the real-time of adjustment intelligent mattress, finally with the intelligent mattress parameter after optimization as foundation, the intelligent mattress is adjusted, by optimizing the adjustment strategy, thereby maximizing the promotion of deep sleep, effectively improve the overall sleep quality and health level of user.

[0144] Each embodiment in the specification is described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same and similar parts of each embodiment can be referred to each other.

[0145] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, apparatuses, electronic device, and computer program products according to the embodiments of the present application. It is understood that each flow and / or block in the flowcharts and / or block diagrams, and a combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminals to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminals generate a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The flowcharts and / or block diagrams can be implemented by one flow or multiple flows and / or one block or multiple blocks. Figure 1 The flowcharts and / or block diagrams can be implemented by one flow or multiple flows and / or one block or multiple blocks.

[0146] These computer program instructions can also be stored in a computer-readable memory that can direct the computer or other programmable data processing terminals to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including instruction devices that implement the functions specified in the flowcharts and / or block diagrams. Figure 1 The flowcharts and / or block diagrams can be implemented by one flow or multiple flows and / or one block or multiple blocks. Figure 1 The flowcharts and / or block diagrams can be implemented by one flow or multiple flows and / or one block or multiple blocks.

[0147] These computer program instructions can also be loaded into a computer or other programmable data processing terminal, so that a series of operation steps are performed on the computer or other programmable terminal to produce a computer-implemented process, so that the instructions executed on the computer or other programmable terminal provide steps for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The flowcharts and / or block diagrams can be implemented by one flow or multiple flows and / or one block or multiple blocks. Figure 1 The flowcharts and / or block diagrams can be implemented by one flow or multiple flows and / or one block or multiple blocks.

[0148] Although the preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications to the embodiments once they know the basic inventive concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the embodiments of the present application.

[0149] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0150] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for adjusting a smart mattress, characterized in that, include: The user's sleep stage is determined by collecting characteristic data of the user during sleep through sensors deployed in the smart mattress and identifying the characteristic data through a pre-set sleep state monitoring model. Based on the user's sleep stage, the parameters of the smart mattress are optimized using a pre-set parameter adjustment model to determine the optimized smart mattress parameters. Based on the optimized smart mattress parameters, the smart mattress is adjusted to enable the user to quickly achieve deep sleep or increase the duration of deep sleep.

2. The adjustment method for the smart mattress according to claim 1, characterized in that, Sensors deployed in the smart mattress collect characteristic data of the user during sleep, including: During the user's sleep, based on a preset data sampling frequency, the heart rate, respiratory rate, and body movement frequency are continuously collected at M sampling time points by sensors deployed in the smart mattress to obtain characteristic data as time-series data.

3. The adjustment method for the smart mattress according to claim 1, characterized in that, The method for the pre-set sleep state monitoring model includes: Collect pre-stored sample feature data and the sample sleep stage labels corresponding to the sample feature data; A sleep state monitoring model is constructed using a convolutional neural network, and the parameters of the sleep state monitoring model are initialized to generate a training population. For any individual in the training population, the fitness of each individual is obtained by taking the pre-stored sample feature data as input and the sample sleep stage label corresponding to the sample feature data as the expected output. Based on the fitness of each individual, the optimal individual is determined. For any individual, based on the optimal individual, a sinusoidal group information interaction method is used to learn group information from the individual to obtain the individual after learning group information. For any individual after learning information from a group, the individual collaborative information interaction method is used to explore the solution space collaboratively to obtain the individual after collaborative exploration; For any individual after collaborative exploration, an adaptive range control method is used to selectively explore the individual to obtain the individual after selective exploration; Determine whether the current number of training iterations is greater than or equal to the maximum number of training iterations. If so, redetermine the optimal individual based on the individuals after the selective exploration. Otherwise, return to the step of determining the optimal individual. The parameters of the newly determined optimal individuals are used as the target parameters of the sleep state monitoring model, and the sleep state monitoring model is deployed based on the target parameters to obtain the pre-set sleep state monitoring model.

4. The adjustment method for the smart mattress according to claim 3, characterized in that, Based on the optimal individual, a sinusoidal group information interaction method is used to learn group information from the individual, resulting in individuals after group information learning, including: Based on all individuals, the centroid of the training population is obtained as follows: in, This represents the k-th individual in the iter-th training process. Let f represent the centroid of the training population in the iterth training process, k = 1, 2, ..., NP, where NP represents the total number of individuals. k express fitness, f sum This represents the total fitness of all individuals; Based on the centroid individual of the training population and the optimal individual, the sinusoidal interaction information is obtained as follows: in, express The corresponding sinusoidal interaction information is as follows: rand1 represents the first random number between [0, π], rand2 represents the random number between [0, 2π], π represents pi, and λ represents the information segmentation factor, which is set to... Based on the sinusoidal interaction information, group information learning is performed on the individual to obtain the individual after group information learning: in, This represents an individual after learning the information from the k-th group.

5. The adjustment method for the smart mattress according to claim 4, characterized in that, For any individual after learning information from a group, a collaborative information interaction method is used to explore the solution space collaboratively, resulting in individuals after collaborative exploration, including: Determine the current training iteration and obtain the learning rate based on the current training iteration: Where, η iter Let η represent the learning rate for the iterth training iteration, with an initial value of 0.1; iter+1 Let t represent the learning rate for the (iter+1)th training iteration. max This represents the maximum number of training iterations, and α represents the learning rate adjustment factor, which is set to 10. For any individual after learning information from a group, randomly match that individual with another individual after learning information from another group to obtain the collaborative exploration individual corresponding to each individual; For any individual after learning from the group information, the collaborative exploration information of the solution space is obtained based on the individual's corresponding collaborative exploration partner and the learning rate: in, This represents the collaborative exploration information in the solution space corresponding to the individual after learning the group information of the i-th group during the (iter+1)-th training process. Let represent the individual after learning the group information from the i-th group during the iter-th training process. express Corresponding collaborative exploration individuals, This represents the optimal individual, rand3 represents the third random number between (0,1), and rand4 represents the fourth random number between (0,1). express and The Euclidean distance between them express and The Euclidean distance between them; For any individual after learning the group information, based on the cooperative exploration information in the solution space, the individual is subjected to cooperative exploration in the solution space to obtain the individual after cooperative exploration: in, This represents the individual following the i-th collaborative exploration.

6. The adjustment method for the smart mattress according to claim 5, characterized in that, For any individual after collaborative exploration, an adaptive range control method is used to selectively explore that individual, resulting in individuals after selective exploration, including: Based on the current number of training iterations, the adaptive range adjustment factor is obtained as follows: Where, β iter β represents the adaptive range adjustment factor during the iterth training iteration, with an initial value set to 0.15; iter+1 This represents the adaptive range adjustment factor during the (iter+1)th training iteration; Based on the aforementioned adaptive range control factor, the exploration range is obtained as follows: in, Let represent the d-th dimension parameter of the individual after the m-th collaborative exploration in the iter-th training process, where d = 1, 2, ..., D, and D represents the total dimension of the parameters. express The corresponding upper limit of the search range, express The corresponding lower limit of the search range; For any individual after collaborative exploration, selective exploration is performed on that individual based on the exploration range, resulting in the following individuals after selective exploration: in, Let d represent the d-th dimension parameter of the individual after the m-th selective exploration, and rand5 represent the fifth random number between (0,1).

7. The adjustment method for the smart mattress according to claim 6, characterized in that, Before optimizing the parameters of the smart mattress using a pre-set parameter adjustment model, the following steps are also included: Collect basic user data, smart mattress parameters, time to enter deep sleep, and time to maintain deep sleep; Using the aforementioned basic data and smart mattress parameters as input, and the deep sleep entry time as the desired output, a first parameter adjustment model is trained. Using the aforementioned basic data and smart mattress parameters as input, and the desired output being the duration of deep sleep, a second parameter adjustment model is trained.

8. The adjustment method for the smart mattress according to claim 6, characterized in that, Based on the user's sleep stage, the parameters of the smart mattress are optimized using a pre-set parameter adjustment model to determine the optimized smart mattress parameters, including: When the user's sleep stage is non-deep sleep, the parameters of the smart mattress are optimized through an intelligent optimization algorithm. During the optimization process, the entry time of deep sleep is evaluated through a pre-set parameter adjustment model, and the target parameters with the shortest entry time are output to obtain the optimized smart mattress parameters. When the user's sleep stage is deep sleep, the parameters of the smart mattress are optimized through an intelligent optimization algorithm. During the optimization process, the retention time of deep sleep is evaluated through a pre-set parameter adjustment model, and the target parameter with the longest retention time is output to obtain the optimized smart mattress parameters.

9. The adjustment method for the smart mattress according to claim 8, characterized in that, The intelligent optimization algorithm is set to either particle swarm optimization algorithm or pigeon flock optimization algorithm.

10. The adjustment method for the smart mattress according to claim 9, characterized in that, The parameters of the smart mattress include one or more of the following: temperature, lumbar support, and shoulder support.

Images