An environmental linkage intelligent wake-up method and system for a smart bed

Abstract

The application relates to the field of intelligent bed wake-up work, and discloses an environment linkage intelligent wake-up method and system of an intelligent bed, which comprises the following steps: monitoring and predicting environment information in a target room through arrangement of environment sensors and physiological information sensors, simultaneously predicting a sleep state of a user, and constructing a personal sleep baseline of the user; combining the personal sleep baseline of the user, linkage of a sleep state prediction model and environment prediction data, constructing an intelligent wake-up decision of the user, and finally generating a sleep quality report of the user. The application can construct a bedroom into a collaborative work network for sleep and health services, greatly improves the added value and competitiveness of a single product, ensures long-term cultivation of a healthy sleep rhythm of a user, and improves the life quality.

Description

Technical Field

[0001] This invention relates to the field of smart bed wake-up, and in particular to a smart bed environmental linkage smart wake-up method and system. Background Technology

[0002] Traditional alarm clocks' "noise alarms" are abrupt interruptions that violate physiology. They ignore the user's sleep stage and environmental conditions, easily leading to increased "sleep inertia." In contrast, waking up through environmental linkage is a slow, natural physiological transition process influenced by both internal and external environments. Our biological clock is primarily regulated by light and temperature. Environmental linkage, by simulating "sunrise" and "temperature changes," directly affects the user's suprachiasmatic nucleus, gently adjusting cortisol and melatonin levels, guiding the body naturally from sleep to wakefulness. This is a "natural" wake-up. Awakening is not just about the brain being awake, but also about the body being ready to greet a new day. A warm, bright environment filled with pleasant sounds and smells is vastly different from a cold, dark environment startled by a jarring alarm, resulting in a drastically different wake-up experience and subsequent energy levels. Environmental linkage unifies sensory experiences, creating optimal conditions for wakefulness. A user's sleep depth and wakefulness are closely related to environmental factors. For example, users may sleep restlessly and be more easily awakened in excessively cold or hot environments; in noisy environments, a stronger arousal is required. Environmental data provides crucial "context" for intelligent decision-making, enabling systems to make more accurate and considerate decisions.

[0003] Smart beds, by integrating with the environment for natural wake-up, can reduce sleep inertia and significantly alleviate grogginess and drowsiness upon waking. They also improve users' mood upon waking, allowing them to achieve peak attention, memory, and decision-making abilities more quickly. Furthermore, by linking lights, air conditioning, curtains, and speakers, smart beds create a collaborative network within the bedroom that supports sleep and health, greatly enhancing the added value and competitiveness of the individual product. This ensures users cultivate healthy sleep rhythms and improve their quality of life in the long term. Summary of the Invention

[0004] This invention overcomes the shortcomings of the prior art and provides a smart bed with an environment-linked intelligent wake-up method and system.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] The first aspect of this invention provides an environmental linkage intelligent wake-up method for a smart bed, comprising the following steps:

[0007] Environmental sensors are deployed in the user's room to collect environmental data, and the environmental data is preprocessed to obtain preprocessed sensor data.

[0008] By using the PINN neural network model and combining preprocessed sensor data, the environmental change data of the target room is predicted and calibrated as real-time room environment prediction data.

[0009] By using deep learning models and real-time environmental prediction data, the sleep status of users in the target room is predicted, and an individual sleep baseline is built based on the sleep status.

[0010] Within the target smart bed, by combining the user's personal sleep baseline with sleep state prediction models and real-time room environment prediction data, a smart wake-up decision for the user within the target room is constructed.

[0011] Generate sleep quality reports for users and, based on these reports, propose sleep adjustment decisions for them.

[0012] Furthermore, in a preferred embodiment of the present invention, the step of deploying environmental sensors in the user's room, collecting environmental data, and preprocessing the environmental data to obtain preprocessed sensor data specifically involves:

[0013] Identify the room where the smart bed is placed and where the user rests, mark it as the target room, and mark the smart bed in the target room as the target smart bed;

[0014] An environmental correlation sensor is deployed in the target room, wherein the environmental correlation sensor includes sensors that monitor the temperature, light intensity, and sound intensity in the target room;

[0015] All environmental sensors are connected in a star topology, and different environmental sensor nodes are connected to the main controller. Based on the main controller, clock response signals are sent to different environmental sensors to control them to respond simultaneously and operate in the target room, thereby obtaining the corresponding sensor data.

[0016] Missing values ​​are filled in the sensor data, and a moving average filter window is introduced to smooth the sensor data for noise with a frequency greater than a preset value, thus obtaining preprocessed sensor data.

[0017] Furthermore, in a preferred embodiment of the present invention, the step of predicting environmental change data of the target room using the PINN neural network model, combined with preprocessed sensor data, and labeling it as real-time room environment prediction data, specifically involves:

[0018] The PINN neural network model is introduced, a preset historical time step is set, and environmental data of the target room within the historical time step is obtained. At the same time, the environmental data of the target room within the historical time step is imported into the PINN neural network model.

[0019] Among them, the environmental data of the target room within the historical time step is divided into several sample data and imported into the PINN neural network model;

[0020] The encoder within the PINN neural network model learns the environmental data of the target room within a historical time step, outputs a dynamic feature vector of the environmental context, and calculates the mean square error of the dynamic feature vector of the environmental context by combining a fully connected feedforward neural network with future time steps in the PINN neural network model.

[0021] During the mean square error calculation process, the dynamic feature vector of the environment context is updated in real time by combining the gradient descent method. If the mean square error of the dynamic feature vector of the environment context converges to the predetermined value, it is determined that the PINN neural network model is in a state of perfect fit, and the trained PINN neural network model is obtained and labeled as the target neural network model.

[0022] The target neural network model presets future time steps and imports preprocessed sensor data, outputting the environmental data of the target room within the future time steps in real time, which is labeled as real-time room environment prediction data.

[0023] Furthermore, in a preferred embodiment of the present invention, the step of predicting the sleep state of a user in a target room using a deep learning model combined with real-time environmental prediction data, and constructing a personal sleep baseline based on the sleep state, specifically involves:

[0024] Sensors are deployed on the target smart bed to monitor the user's physiological information and monitor the user's physiological information during sleep in real time, which is then labeled as the user's sleep physiological information.

[0025] The user's sleep physiological information includes the user's heart rate, respiratory rate, and body movement signals during sleep, and the user's sleep physiological information is extracted in terms of time domain features and frequency domain features.

[0026] The body movement signals include the number of times the user turns over during sleep and the total range of motion of the turning over;

[0027] All sleep physiological information is synchronized, and the sleep physiological information, real-time room environment prediction data, and body movement signals are converted into feature vectors to obtain sleep-related feature vectors.

[0028] A bidirectional LSTM model incorporating an attention mechanism is introduced, wherein the architecture of the bidirectional LSTM model incorporating an attention mechanism includes an input layer, a feature learning layer, and an attention layer.

[0029] Sleep-related feature vectors are imported into the input layer for training to form a feature vector sequence. Deep dependency learning is performed on the feature sequence vectors in the feature learning layer. Finally, the attention layer controls the bidirectional LSTM model with attention mechanism to label the transition moments in the sleep process, resulting in the trained bidirectional LSTM model with attention mechanism, which is labeled as the sleep state prediction model.

[0030] The sleep state prediction model outputs the user's real-time sleep state on the smart bed, and combines the user's real-time sleep state on the smart bed to output the user's personal sleep baseline.

[0031] Among them, the indicators within a user's personal sleep baseline include sleep structure, physiological indicators, and indicators of changes in environmental interaction.

[0032] Furthermore, in a preferred embodiment of the present invention, the step of combining the user's personal sleep baseline with a sleep state prediction model and real-time room environment prediction data to construct an intelligent wake-up decision for the user within the target smart bed specifically involves:

[0033] A data fusion framework is constructed, which links the user's personal sleep baseline, sleep state prediction model and real-time room environment prediction data into the data integration layer of the data fusion framework and performs time-series alignment processing.

[0034] Among them, the data fusion framework generates a decision model for linking different data after input data, which is labeled as a data joint decision model;

[0035] In the data joint decision-making model, a multi-objective constraint optimization algorithm is designed. Within the multi-objective constraint optimization algorithm, a constraint condition is set, wherein the constraint condition is the latest wake-up time of the user.

[0036] In the data joint decision-making model, an initial decision-making stage and a decision-learning stage are preset. In the initial decision-making stage, the data joint decision-making model is linked to the target smart bed to control the target smart bed to formulate the intelligent wake-up decision corresponding to the initial decision-making stage based on the user's personal sleep baseline.

[0037] Among them, the intelligent wake-up decision corresponding to the initial stage of the decision is that, within the workday time set by the user, when the user's sleep time reaches the total average sleep time included in the user's personal sleep baseline, the target smart bed is controlled to wake up the user, and when the user's sleep time reaches the user's latest wake-up time, the target smart bed is directly controlled to wake up the user.

[0038] The smart bed wakes up the user by connecting to the device in the target room that is used to wake up the user. It automatically performs the function of waking up the user. If the user does not actively turn off the device in the target room, it will stop the operation of the device in the target room when the physiological indicators of the user's personal sleep baseline reach the wake-up state physiological indicators.

[0039] By combining data-driven joint decision-making models, an intelligent wake-up decision is constructed for the decision learning stage.

[0040] Furthermore, in a preferred embodiment of the present invention, the step of constructing intelligent wake-up decision in the decision learning stage by combining the data joint decision model specifically includes:

[0041] During the decision learning phase, a data-driven joint decision model is used to analyze the relationship between real-time room environment prediction data and the user's personal sleep baseline environmental interaction change indicators and sleep state.

[0042] Among them, the environmental interaction change index and sleep state jointly recorded the user's corresponding sleep state under different combinations of light intensity, ambient temperature and sound intensity, and determined the combination of light intensity, ambient temperature and sound intensity when the user is in a wakeful state, which was then labeled as a wakeable environmental data combination.

[0043] The wakeable environmental data combination is applied in real time to the data joint decision model in the decision learning stage for model iterative learning. When the combination of different environmental data identified by the environmental correlation sensor in the target room is a wakeable environmental data combination, the data joint decision model controls the target smart bed to connect to the device in the target room used to wake up the user and automatically executes the function of waking up the user.

[0044] In the decision learning phase, when the user's sleep time reaches the user's latest wake-up time, but the current combination of environmental data is not a wake-up-friendly combination of environmental data, the environmental adjustment device in the target room is connected to the target smart bed to adjust the combination of environmental data in the target room to be equal to the wake-up-friendly combination of environmental data, and the target smart bed is controlled to wake up the user.

[0045] After waking the user, analyze the user's physiological information at the time of wake-up and preset the physiological information health range. If the user's physiological information at the time of wake-up is not within the physiological information health range, it is determined that the user's comfort level is not qualified when the current target smart bed wakes up the user.

[0046] If the user's comfort level is not up to standard, the data-driven joint decision-making model in the decision-making learning phase will be updated and learned a second time until the output intelligent wake-up decision meets the user's comfort level.

[0047] Furthermore, in a preferred embodiment of the present invention, when the user comfort level is unsatisfactory, the data joint decision model in the decision learning stage is subjected to a second model update learning until the output intelligent wake-up decision satisfies the user comfort level. Specifically, this involves:

[0048] When the user's comfort level is not up to standard, the real-time room environment prediction data is analyzed, and the time point when the real-time room environment prediction data is output to obtain the wake-up environment data combination is marked as the wake-up time point, and the environment adjustment time point is preset.

[0049] During the environmental control time point, based on the target smart bed connecting to the environmental control equipment in the target room, the combination of environmental data in the target room is adjusted to equal the combination of wakeable environmental data, and the combination of wakeable environmental data is maintained until the wake-up time point;

[0050] If the user's physiological information remains within the physiological information monitoring range when they wake up, then the combination of environmental data in the target room will be adjusted in advance to generate a corresponding intelligent wake-up decision.

[0051] If the user's comfort level is still not up to standard, the user's sleep state is determined to be abnormal. The system will continue to adjust the combination of environmental data in the target room in advance. During the advance adjustment process, the target smart bed will control the music playback device in the target room to play music for waking up until the user wakes up.

[0052] Furthermore, in a preferred embodiment of the present invention, the step of generating a user's sleep quality report and proposing sleep adjustment decisions based on the user's sleep quality report specifically includes:

[0053] By combining the user's personal sleep baseline with sleep state prediction models and real-time room environment prediction data, a monthly sleep quality report is generated after the target smart bed wakes the user.

[0054] The monthly sleep quality report records the user's average sleep status during the month and real-time room environment data corresponding to different sleep statuses, as well as the user's physiological information after waking up.

[0055] At the same time, the monthly sleep quality report will indicate the time points when the user's sleep state is abnormal and the user's physiological information is abnormal after waking up;

[0056] A big data network is introduced, and a sleep health knowledge graph is obtained based on the big data network. The sleep health knowledge graph records the adjustment decisions corresponding to the user in different sleep states and when different physiological information is generated after awakening. The sleep health knowledge graph is synchronized to the target smart bed, and the corresponding adjustment decisions are output in the monthly sleep quality report.

[0057] A second aspect of this invention also provides an environment-linked intelligent wake-up system for a smart bed. The environment-linked intelligent wake-up system integrates a high-performance computing architecture and a data storage module, including a non-volatile memory composed of a DDR4 RDIMM memory module with ECC verification and an NVMe solid-state storage array using 3D NAND flash memory, and a multi-core processor based on the Zen4 microarchitecture. The memory contains a program for an environment-linked intelligent wake-up method with an intelligent engine. When the program is executed in parallel through a superscalar pipeline execution unit within the processor, the following steps are implemented:

[0058] Environmental sensors are deployed in the user's room to collect environmental data, and the environmental data is preprocessed to obtain preprocessed sensor data.

[0059] By using the PINN neural network model and combining preprocessed sensor data, the environmental change data of the target room is predicted and calibrated as real-time room environment prediction data.

[0060] By using deep learning models and real-time environmental prediction data, the sleep status of users in the target room is predicted, and an individual sleep baseline is built based on the sleep status.

[0061] Within the target smart bed, by combining the user's personal sleep baseline with sleep state prediction models and real-time room environment prediction data, a smart wake-up decision for the user within the target room is constructed.

[0062] Generate sleep quality reports for users and, based on these reports, propose sleep adjustment decisions for them.

[0063] This invention addresses the technical deficiencies in the prior art and offers the following advantages: By deploying environmental and physiological sensors, it monitors and predicts environmental information within a target room, simultaneously predicting the user's sleep state and establishing a personal sleep baseline. Combining this personal sleep baseline with a sleep state prediction model and environmental prediction data, it constructs an intelligent wake-up decision for the user, ultimately generating a sleep quality report. This invention transforms the bedroom into a collaborative network serving sleep and health, significantly enhancing the added value and competitiveness of individual products, ensuring users cultivate healthy sleep rhythms in the long term, and improving their quality of life. Attached Figure Description

[0064] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0065] Figure 1 A flowchart of an environmental linkage intelligent wake-up method for a smart bed is shown;

[0066] Figure 2 A flowchart illustrating the method for constructing intelligent wake-up decisions for users within a target room is shown.

[0067] Figure 3 A program view of an environment-linked intelligent wake-up system for a smart bed is shown. Detailed Implementation

[0068] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0069] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0070] Figure 1 A flowchart illustrating an environmental-linked intelligent wake-up method for a smart bed is shown, including the following steps:

[0071] S102: Deploy environmental sensors in the user's room, collect environmental data, and preprocess the environmental data to obtain preprocessed sensor data.

[0072] S104: Using the PINN neural network model and combined with preprocessed sensor data, predict the environmental change data of the target room and label it as real-time room environment prediction data;

[0073] S106: By using a deep learning model and combining real-time environmental prediction data, predict the sleep status of users in the target room and build an individual sleep baseline based on the sleep status.

[0074] S108: Within the target smart bed, the system combines the user's personal sleep baseline with a sleep state prediction model and real-time room environment prediction data to construct an intelligent wake-up decision for the user within the target room.

[0075] S110: Generate sleep quality reports for users and, based on these reports, propose sleep adjustment decisions for users.

[0076] Furthermore, in a preferred embodiment of the present invention, the step of deploying environmental sensors in the user's room, collecting environmental data, and preprocessing the environmental data to obtain preprocessed sensor data specifically involves:

[0077] Identify the room where the smart bed is placed and where the user rests, mark it as the target room, and mark the smart bed in the target room as the target smart bed;

[0078] An environmental correlation sensor is deployed in the target room, wherein the environmental correlation sensor includes sensors that monitor the temperature, light intensity, and sound intensity in the target room;

[0079] All environmental sensors are connected in a star topology, and different environmental sensor nodes are connected to the main controller. Based on the main controller, clock response signals are sent to different environmental sensors to control them to respond simultaneously and operate in the target room, thereby obtaining the corresponding sensor data.

[0080] Missing values ​​are filled in the sensor data, and a moving average filter window is introduced to smooth the sensor data for noise with a frequency greater than a preset value, thus obtaining preprocessed sensor data.

[0081] It's important to note that acquiring room environmental information, including temperature, light intensity, and sound intensity, is crucial because people are easily affected by their environment during sleep. Therefore, monitoring environmental data is necessary to adjust it and trigger a wake-up call for the user. After monitoring, the data needs preprocessing, including missing value imputation and sliding filter noise reduction. Simultaneous monitoring during the process is also required to ensure environmental data accuracy and facilitate subsequent calculations.

[0082] Furthermore, in a preferred embodiment of the present invention, the step of predicting environmental change data of the target room using the PINN neural network model, combined with preprocessed sensor data, and labeling it as real-time room environment prediction data, specifically involves:

[0083] The PINN neural network model is introduced, a preset historical time step is set, and environmental data of the target room within the historical time step is obtained. At the same time, the environmental data of the target room within the historical time step is imported into the PINN neural network model.

[0084] Among them, the environmental data of the target room within the historical time step is divided into several sample data and imported into the PINN neural network model;

[0085] The encoder within the PINN neural network model learns the environmental data of the target room within a historical time step, outputs a dynamic feature vector of the environmental context, and calculates the mean square error of the dynamic feature vector of the environmental context by combining a fully connected feedforward neural network with future time steps in the PINN neural network model.

[0086] During the mean square error calculation process, the dynamic feature vector of the environment context is updated in real time by combining the gradient descent method. If the mean square error of the dynamic feature vector of the environment context converges to the predetermined value, it is determined that the PINN neural network model is in a state of perfect fit, and the trained PINN neural network model is obtained and labeled as the target neural network model.

[0087] The target neural network model presets future time steps and imports preprocessed sensor data, outputting the environmental data of the target room within the future time steps in real time, which is labeled as real-time room environment prediction data.

[0088] It's important to note that PINN uses partial differential equations describing physical processes as constraints, embedding them into the loss function of the neural network to guide its learning within the framework of physical laws. This is particularly suitable for processes with clear physical laws, such as temperature and airflow, hence the choice of the PINN neural network model. Using the PINN neural network model to predict environmental data involves calculations including mean squared error convergence until a perfect fit is achieved, combined with historical data analysis to assess environmental changes and generate real-time room environment prediction data.

[0089] Furthermore, in a preferred embodiment of the present invention, the step of predicting the sleep state of a user in a target room using a deep learning model combined with real-time environmental prediction data, and constructing a personal sleep baseline based on the sleep state, specifically involves:

[0090] Sensors are deployed on the target smart bed to monitor the user's physiological information and monitor the user's physiological information during sleep in real time, which is then labeled as the user's sleep physiological information.

[0091] The user's sleep physiological information includes the user's heart rate, respiratory rate, and body movement signals during sleep, and the user's sleep physiological information is extracted in terms of time domain features and frequency domain features.

[0092] The body movement signals include the number of times the user turns over during sleep and the total range of motion of the turning over;

[0093] All sleep physiological information is synchronized, and the sleep physiological information, real-time room environment prediction data, and body movement signals are converted into feature vectors to obtain sleep-related feature vectors.

[0094] A bidirectional LSTM model incorporating an attention mechanism is introduced, wherein the architecture of the bidirectional LSTM model incorporating an attention mechanism includes an input layer, a feature learning layer, and an attention layer.

[0095] Sleep-related feature vectors are imported into the input layer for training to form a feature vector sequence. Deep dependency learning is performed on the feature sequence vectors in the feature learning layer. Finally, the attention layer controls the bidirectional LSTM model with attention mechanism to label the transition moments in the sleep process, resulting in the trained bidirectional LSTM model with attention mechanism, which is labeled as the sleep state prediction model.

[0096] The sleep state prediction model outputs the user's real-time sleep state on the smart bed, and combines the user's real-time sleep state on the smart bed to output the user's personal sleep baseline.

[0097] Among them, the indicators within a user's personal sleep baseline include sleep structure, physiological indicators, and indicators of changes in environmental interaction.

[0098] It's important to note that deep learning models are used to predict a user's sleep state in a target room by combining real-time environmental prediction data. Sleep state includes the user's heart rate, respiratory rate, and body movement signals during sleep, and is categorized into light sleep and deep sleep. Different sleep states result in different health outcomes. A bidirectional LSTM model using an attention mechanism can predict whether a user is in light or deep sleep given current heart and respiratory rate signals. Key indicators of a user's personal sleep baseline include sleep structure, physiological indicators, and environmental interaction change indicators, such as total sleep time, wakefulness after falling asleep, deep sleep duration, average nighttime heart and respiratory rate, and the average temperature and humidity range during sleep. A user's personal sleep baseline reflects their average sleep indicators, which can be used to make sleep-wake decisions.

[0099] Furthermore, in a preferred embodiment of the present invention, the step of generating a user's sleep quality report and proposing sleep adjustment decisions based on the user's sleep quality report specifically includes:

[0100] By combining the user's personal sleep baseline with sleep state prediction models and real-time room environment prediction data, a monthly sleep quality report is generated after the target smart bed wakes the user.

[0101] The monthly sleep quality report records the user's average sleep status during the month and real-time room environment data corresponding to different sleep statuses, as well as the user's physiological information after waking up.

[0102] At the same time, the monthly sleep quality report will indicate the time points when the user's sleep state is abnormal and the user's physiological information is abnormal after waking up;

[0103] A big data network is introduced, and a sleep health knowledge graph is obtained based on the big data network. The sleep health knowledge graph records the adjustment decisions corresponding to the user in different sleep states and when different physiological information is generated after awakening. The sleep health knowledge graph is synchronized to the target smart bed, and the corresponding adjustment decisions are output in the monthly sleep quality report.

[0104] It's important to note that the monthly sleep quality report reflects a user's sleep patterns. If abnormal sleep patterns occur that impact health, adjustments need to be made based on these patterns. Therefore, a sleep health knowledge graph is used to provide adjustment decisions for users in different sleep states and when different physiological information is generated after awakening.

[0105] Figure 2 The flowchart illustrates a method for constructing intelligent wake-up decisions for users within a target room, including the following steps:

[0106] S202: Within the target smart bed, combine the user's personal sleep baseline, link the sleep state prediction model and real-time room environment prediction data, and construct an intelligent wake-up decision for the user within the target room;

[0107] S204: Combining data-driven joint decision-making models to construct intelligent wake-up decisions during the decision-making learning phase;

[0108] S206: If the user comfort level is not up to standard, the data joint decision model in the decision learning stage will be updated and learned a second time until the output intelligent wake-up decision makes the user comfort level up to standard.

[0109] Furthermore, in a preferred embodiment of the present invention, S202 specifically includes:

[0110] A data fusion framework is constructed, which links the user's personal sleep baseline, sleep state prediction model and real-time room environment prediction data into the data integration layer of the data fusion framework and performs time-series alignment processing.

[0111] Among them, the data fusion framework generates a decision model for linking different data after input data, which is labeled as a data joint decision model;

[0112] In the data joint decision-making model, a multi-objective constraint optimization algorithm is designed. Within the multi-objective constraint optimization algorithm, a constraint condition is set, wherein the constraint condition is the latest wake-up time of the user.

[0113] In the data joint decision-making model, an initial decision-making stage and a decision-learning stage are preset. In the initial decision-making stage, the data joint decision-making model is linked to the target smart bed to control the target smart bed to formulate the intelligent wake-up decision corresponding to the initial decision-making stage based on the user's personal sleep baseline.

[0114] Among them, the intelligent wake-up decision corresponding to the initial stage of the decision is that, within the workday time set by the user, when the user's sleep time reaches the total average sleep time included in the user's personal sleep baseline, the target smart bed is controlled to wake up the user, and when the user's sleep time reaches the user's latest wake-up time, the target smart bed is directly controlled to wake up the user.

[0115] The smart bed wakes up the user by connecting to the device in the target room that is used to wake up the user. It automatically performs the function of waking up the user. If the user does not actively turn off the device in the target room, it will stop the operation of the device in the target room when the physiological indicators of the user's personal sleep baseline reach the wake-up state physiological indicators.

[0116] By combining data-driven joint decision-making models, an intelligent wake-up decision is constructed for the decision learning stage.

[0117] It should be noted that by combining the user's personal sleep baseline, sleep state prediction model, and real-time room environment prediction data, a model can be constructed to make wake-up decisions based on the aforementioned data. The model construction needs to include constraints, namely, the latest wake-up time for the user; exceeding this time may result in the user being late for work, etc. In the data-driven joint decision-making model, an initial decision-making phase and a decision-learning phase are preset. In the initial phase, the user's sleep state is not learned; therefore, the wake-up decision is generated based on the currently monitored user's personal sleep baseline, that is, judging the user's usual wake-up time and sleep duration to wake the user at an appropriate time, and waking the user at the latest required wake-up time. Furthermore, it is necessary to control the target smart bed to activate similar alarm clock and vibrating bed mechanisms in the target room to wake the user. During the wake-up process, physiological information, including heart rate and respiratory rate, is analyzed in real time. When the preset threshold is reached—that is, the heart rate and respiratory rate during the wake-up process—the user is considered fully awake.

[0118] Furthermore, in a preferred embodiment of the present invention, S204 specifically includes:

[0119] During the decision learning phase, a data-driven joint decision model is used to analyze the relationship between real-time room environment prediction data and the user's personal sleep baseline environmental interaction change indicators and sleep state.

[0120] Among them, the environmental interaction change index and sleep state jointly recorded the user's corresponding sleep state under different combinations of light intensity, ambient temperature and sound intensity, and determined the combination of light intensity, ambient temperature and sound intensity when the user is in a wakeful state, which was then labeled as a wakeable environmental data combination.

[0121] The wakeable environmental data combination is applied in real time to the data joint decision model in the decision learning stage for model iterative learning. When the combination of different environmental data identified by the environmental correlation sensor in the target room is a wakeable environmental data combination, the data joint decision model controls the target smart bed to connect to the device in the target room used to wake up the user and automatically executes the function of waking up the user.

[0122] In the decision learning phase, when the user's sleep time reaches the user's latest wake-up time, but the current combination of environmental data is not a wake-up-friendly combination of environmental data, the environmental adjustment device in the target room is connected to the target smart bed to adjust the combination of environmental data in the target room to be equal to the wake-up-friendly combination of environmental data, and the target smart bed is controlled to wake up the user.

[0123] After waking the user, analyze the user's physiological information at the time of wake-up and preset the physiological information health range. If the user's physiological information at the time of wake-up is not within the physiological information health range, it is determined that the user's comfort level is not qualified when the current target smart bed wakes up the user.

[0124] If the user's comfort level is not up to standard, the data-driven joint decision-making model in the decision-making learning phase will be updated and learned a second time until the output intelligent wake-up decision meets the user's comfort level.

[0125] It's important to note that during the learning phase, environmental data is used to process user wake-up. Just as there are suitable sleep environments, there are also suitable wake-up environments, such as appropriate ambient music, temperature, and lighting. By learning the user's preferred environmental data combinations, a suitable wake-up environment can be created. Under this wake-up environment combination, the user's sleep will transition from deep sleep to light sleep, and devices like alarm clocks can wake the user in a healthy way. At this stage, the environmental data is used to wake the user based on natural environmental changes under normal conditions, combined with physiological information such as sleep duration. If the user's sleep time coincides with their latest wake-up time, but the current environmental data combination is not a wake-up suitable combination, directly waking the user could potentially harm their health. In this case, the environment is adjusted to a wake-up suitable sleep environment combination before waking, ensuring a healthy and comfortable wake-up.

[0126] Furthermore, in a preferred embodiment of the present invention, S206 specifically includes:

[0127] When the user's comfort level is not up to standard, the real-time room environment prediction data is analyzed, and the time point when the real-time room environment prediction data is output to obtain the wake-up environment data combination is marked as the wake-up time point, and the environment adjustment time point is preset.

[0128] During the environmental control time point, based on the target smart bed connecting to the environmental control equipment in the target room, the combination of environmental data in the target room is adjusted to equal the combination of wakeable environmental data, and the combination of wakeable environmental data is maintained until the wake-up time point;

[0129] If the user's physiological information remains within the physiological information monitoring range when they wake up, then the combination of environmental data in the target room will be adjusted in advance to generate a corresponding intelligent wake-up decision.

[0130] If the user's comfort level is still not up to standard, the user's sleep state is determined to be abnormal. The system will continue to adjust the combination of environmental data in the target room in advance. During the advance adjustment process, the target smart bed will control the music playback device in the target room to play music for waking up until the user wakes up.

[0131] It should be noted that if the user's comfort level is unsatisfactory, the room environment data can be adjusted in advance, such as by adjusting it 15 minutes earlier, to allow the user to adapt to the environment beforehand. This will help maintain normal physiological information levels at wake-up time. If the user's comfort level is still unsatisfactory, it indicates an abnormal sleep state, such as insomnia, excessive sleep, or overexertion. In this case, suitable relaxing music should be played to help the user relax until they wake up.

[0132] like Figure 3 As shown, a second aspect of the present invention also provides an environment-linked intelligent wake-up system for a smart bed. The environment-linked intelligent wake-up system integrates a high-performance computing architecture and a data storage module, including a non-volatile memory composed of a DDR4 RDIMM memory module with ECC verification and an NVMe solid-state storage array using 3D NAND flash memory, and a multi-core processor based on the Zen4 microarchitecture. The memory contains a program for an environment-linked intelligent wake-up method with an intelligent engine. When the program is executed in parallel through a superscalar pipeline execution unit within the processor, the following steps are implemented:

[0133] Environmental sensors are deployed in the user's room to collect environmental data, and the environmental data is preprocessed to obtain preprocessed sensor data.

[0134] By using the PINN neural network model and combining preprocessed sensor data, the environmental change data of the target room is predicted and calibrated as real-time room environment prediction data.

[0135] By using deep learning models and real-time environmental prediction data, the sleep status of users in the target room is predicted, and an individual sleep baseline is built based on the sleep status.

[0136] Within the target smart bed, by combining the user's personal sleep baseline with sleep state prediction models and real-time room environment prediction data, a smart wake-up decision for the user within the target room is constructed.

[0137] Generate sleep quality reports for users and, based on these reports, propose sleep adjustment decisions for them.

[0138] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A smart bed's environment-linked intelligent wake-up method, characterized in that, Includes the following steps: Environmental sensors are deployed in the user's room to collect environmental data, and the environmental data is preprocessed to obtain preprocessed sensor data. By using the PINN neural network model and combining preprocessed sensor data, the environmental change data of the target room is predicted and calibrated as real-time room environment prediction data. By using deep learning models and real-time environmental prediction data, the sleep status of users in the target room is predicted, and an individual sleep baseline is built based on the sleep status. Within the target smart bed, by combining the user's personal sleep baseline with sleep state prediction models and real-time room environment prediction data, a smart wake-up decision for the user within the target room is constructed. Generate sleep quality reports for users and, based on these reports, propose sleep adjustment decisions for them. Specifically, the step of using the PINN neural network model, combined with preprocessed sensor data, to predict environmental changes in the target room and calibrate this as real-time room environment prediction data involves: The PINN neural network model is introduced, a preset historical time step is set, and environmental data of the target room within the historical time step is obtained. At the same time, the environmental data of the target room within the historical time step is imported into the PINN neural network model. Among them, the environmental data of the target room within the historical time step is divided into several sample data and imported into the PINN neural network model; The encoder within the PINN neural network model learns the environmental data of the target room within a historical time step, outputs a dynamic feature vector of the environmental context, and calculates the mean square error of the dynamic feature vector of the environmental context by combining a fully connected feedforward neural network with future time steps in the PINN neural network model. During the mean square error calculation process, the dynamic feature vector of the environment context is updated in real time by combining the gradient descent method. If the mean square error of the dynamic feature vector of the environment context converges to the predetermined value, it is determined that the PINN neural network model is in a perfectly fitted state, and the trained PINN neural network model is obtained and labeled as the target neural network model. The target neural network model presets a future time step and imports preprocessed sensor data to output the environmental data of the target room within the future time step in real time, which is labeled as real-time room environment prediction data. Specifically, the method of using a deep learning model, combined with real-time environmental prediction data, to predict the sleep state of users in the target room, and constructing an individual sleep baseline based on the sleep state, involves: Sensors are deployed on the target smart bed to monitor the user's physiological information and monitor the user's physiological information during sleep in real time, which is then labeled as the user's sleep physiological information. The user's sleep physiological information includes the user's heart rate, respiratory rate, and body movement signals during sleep, and the user's sleep physiological information is extracted in terms of time domain features and frequency domain features. The body movement signals include the number of times the user turns over during sleep and the total range of motion of the turning over; All sleep physiological information is synchronized, and the sleep physiological information and real-time room environment prediction data are converted into feature vectors to obtain sleep-related feature vectors; A bidirectional LSTM model incorporating an attention mechanism is introduced, wherein the architecture of the bidirectional LSTM model incorporating an attention mechanism includes an input layer, a feature learning layer, and an attention layer. Sleep-related feature vectors are imported into the input layer for training to form a feature vector sequence. Deep dependency learning is performed on the feature sequence vectors in the feature learning layer. Finally, the attention layer controls the bidirectional LSTM model with attention mechanism to label the transition moments in the sleep process, resulting in the trained bidirectional LSTM model with attention mechanism, which is labeled as the sleep state prediction model. The sleep state prediction model outputs the user's real-time sleep state on the smart bed, and combines the user's real-time sleep state on the smart bed to output the user's personal sleep baseline. Among them, the indicators within a user's personal sleep baseline include sleep structure, physiological indicators, and indicators of changes in environmental interaction.

2. The intelligent wake-up method for an intelligent bed based on environmental linkage as described in claim 1, characterized in that, The process of deploying environmental sensors in the user's room, collecting environmental data, and preprocessing the environmental data to obtain preprocessed sensor data specifically involves: Identify the room where the smart bed is placed and where the user rests, mark it as the target room, and mark the smart bed in the target room as the target smart bed; An environmental correlation sensor is deployed in the target room, wherein the environmental correlation sensor includes sensors that monitor the temperature, light intensity, and sound intensity in the target room; All environmental sensors are connected in a star topology, and different environmental sensor nodes are connected to the main controller. Based on the main controller, clock response signals are sent to different environmental sensors to control them to respond simultaneously and operate in the target room, thereby obtaining the corresponding sensor data. Missing values ​​are filled in the sensor data, and a moving average filter window is introduced to smooth the sensor data for noise with a frequency greater than a preset value, thus obtaining preprocessed sensor data.

3. The intelligent wake-up method for an intelligent bed based on environmental linkage as described in claim 1, characterized in that, The process of constructing an intelligent wake-up decision system for the user within the target smart bed, by combining the user's personal sleep baseline with a sleep state prediction model and real-time room environment prediction data, specifically involves: A data fusion framework is constructed, which links the user's personal sleep baseline, sleep state prediction model and real-time room environment prediction data into the data integration layer of the data fusion framework and performs time-series alignment processing. Among them, the data fusion framework generates a decision model for linking different data after input data, which is labeled as a data joint decision model; In the data joint decision-making model, a multi-objective constraint optimization algorithm is designed. Within the multi-objective constraint optimization algorithm, a constraint condition is set, wherein the constraint condition is the latest wake-up time of the user. In the data joint decision-making model, an initial decision-making stage and a decision-learning stage are preset. In the initial decision-making stage, the data joint decision-making model is linked to the target smart bed to control the target smart bed to formulate the intelligent wake-up decision corresponding to the initial decision-making stage based on the user's personal sleep baseline. Among them, the intelligent wake-up decision corresponding to the initial stage of the decision is that, within the workday time set by the user, when the user's sleep time reaches the total average sleep time included in the user's personal sleep baseline, the target smart bed is controlled to wake up the user, and when the user's sleep time reaches the user's latest wake-up time, the target smart bed is directly controlled to wake up the user. The smart bed wakes up the user by connecting to the device in the target room that is used to wake up the user. It automatically performs the function of waking up the user. If the user does not actively turn off the device in the target room, it will stop the operation of the device in the target room when the physiological indicators of the user's personal sleep baseline reach the wake-up state physiological indicators. By combining data-driven joint decision-making models, an intelligent wake-up decision is constructed for the decision learning stage.

4. The intelligent wake-up method for an intelligent bed based on environmental linkage as described in claim 3, characterized in that, The intelligent wake-up decision-making process for the decision learning stage, which combines data-driven joint decision-making models, specifically involves: During the decision learning phase, a data-driven joint decision model is used to analyze the relationship between real-time room environment prediction data and the user's personal sleep baseline, including environmental interaction change indicators and sleep state. Among them, the environmental interaction change index and sleep state jointly recorded the user's corresponding sleep state under different combinations of light intensity, ambient temperature and sound intensity, and determined the combination of light intensity, ambient temperature and sound intensity when the user is in a wakeful state, which was then labeled as a wakeable environmental data combination. The wakeable environmental data combination is applied in real time to the data joint decision model in the decision learning stage for model iterative learning. When the combination of different environmental data identified by the environmental correlation sensor in the target room is a wakeable environmental data combination, the data joint decision model controls the target smart bed to connect to the device in the target room used to wake up the user and automatically executes the function of waking up the user. In the decision learning phase, when the user's sleep time reaches the user's latest wake-up time, but the current combination of environmental data is not a wake-up-friendly combination of environmental data, the environmental adjustment device in the target room is connected to the target smart bed to adjust the combination of environmental data in the target room to be equal to the wake-up-friendly combination of environmental data, and the target smart bed is controlled to wake up the user. After waking the user, analyze the user's physiological information at the time of wake-up and preset the physiological information health range. If the user's physiological information at the time of wake-up is not within the physiological information health range, it is determined that the user's comfort level is not qualified when the current target smart bed wakes up the user. If the user's comfort level is not satisfactory, the data-driven joint decision-making model in the decision-making learning phase will be updated and learned a second time until the output intelligent wake-up decision satisfies the user's comfort level.

5. The intelligent wake-up method for an intelligent bed based on environmental linkage according to claim 4, characterized in that, When user comfort is not satisfactory, the data-driven joint decision-making model in the decision-learning phase undergoes a second update until the output intelligent wake-up decision satisfies user comfort. Specifically: When the user's comfort level is not up to standard, the real-time room environment prediction data is analyzed, and the time point when the real-time room environment prediction data is output to obtain the wake-up environment data combination is marked as the wake-up time point, and the environment adjustment time point is preset. During the environmental control time point, based on the target smart bed connecting to the environmental control equipment in the target room, the combination of environmental data in the target room is adjusted to equal the combination of wakeable environmental data, and the combination of wakeable environmental data is maintained until the wake-up time point; If the user's physiological information remains within the physiological information monitoring range when they wake up, then the combination of environmental data in the target room will be adjusted in advance to generate a corresponding intelligent wake-up decision. If the user's comfort level is still not up to standard, the user's sleep state is determined to be abnormal. The system will continue to adjust the combination of environmental data in the target room in advance. During the advance adjustment process, the target smart bed will control the music playback device in the target room to play music for waking up until the user wakes up.

6. The intelligent wake-up method for an intelligent bed based on environmental linkage as described in claim 1, characterized in that, The process of generating a sleep quality report for the user and proposing sleep adjustment decisions based on the report includes: By combining the user's personal sleep baseline with sleep state prediction models and real-time room environment prediction data, a monthly sleep quality report is generated after the target smart bed wakes the user. The monthly sleep quality report records the user's average sleep status during the month and real-time room environment data corresponding to different sleep statuses, as well as the user's physiological information after waking up. At the same time, the monthly sleep quality report will indicate the time points when the user's sleep state is abnormal and the user's physiological information is abnormal after waking up; A big data network is introduced, and a sleep health knowledge graph is obtained based on the big data network. The sleep health knowledge graph records the adjustment decisions corresponding to the user in different sleep states and when different physiological information is generated after awakening. The sleep health knowledge graph is synchronized to the target smart bed, and the corresponding adjustment decisions are output in the monthly sleep quality report.

7. An environmental linkage intelligent wake-up system for a smart bed, characterized in that, The environment-linked intelligent wake-up system integrates a high-performance computing architecture and a data storage module, including a non-volatile memory consisting of a DDR4 RDIMM memory module with ECC verification and an NVMe solid-state storage array using 3D NAND flash memory, and a multi-core processor based on the Zen4 microarchitecture; the memory contains a program for an environment-linked intelligent wake-up method with an intelligent engine, and when the program is decoded and executed in parallel by the superscalar pipeline execution unit in the processor, the steps of the environment-linked intelligent wake-up method as described in any one of claims 1-6 are implemented.

Images