Air conditioner control method, device and equipment in sleep application scene and medium

Abstract

The invention discloses an air conditioner control method, device and equipment in a sleep application scenario and a medium, and the air conditioner control method in the sleep application scenario comprises the steps that point cloud data of a user sleep area is continuously obtained through a millimeter wave radar, and the covering states of covering objects of all parts of a body are accurately recognized; the defects that a traditional infrared or temperature and humidity sensor cannot penetrate through the fabric, and local exposure is difficult to judge are overcome. Based on the recognition result, the system generates wind direction, wind speed and other wind feeling adjusting instructions in real time, air supply of the air conditioner is prevented from directly blowing exposed parts, and the thermal comfort is improved. The self-adaptive optimization of the strategy is realized by continuously monitoring the body movement frequency and the turn-over times of the user before and after adjustment, evaluating the adjustment effect and carrying out positive or negative weight adjustment on the control rule. The problems that in the prior art, due to the fact that the coverage state cannot be accurately perceived, air supply is uncomfortable, cold catching is likely to happen, and adjustment lags are solved, and the air conditioner use experience in the sleep scene is remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of air conditioner control, and in particular to an air conditioner control method and device in a sleep application scenario, an equipment and a medium. BACKGROUND

[0002] With the development of smart home technology, smart air conditioners have gradually evolved from single temperature regulation devices to comfort systems with environmental perception and adaptive adjustment capabilities. Especially in sleep application scenarios, users have higher requirements for the quietness, air softness and temperature and humidity stability of air conditioners. Therefore, some air conditioner products in the prior art introduce human presence detection, position tracking and even basic posture recognition functions, trying to achieve intelligent control such as avoiding direct blowing and zoned air supply.

[0003] However, in the actual sleep process, the body thermal comfort state of the user is not only affected by the environmental temperature and humidity, but also closely related to the bedding coverage state. For example, when the user turns over and causes the quilt to slide off, the limbs are exposed, and if the air conditioner still maintains the original air supply mode, it is easy to cause local cooling and cause discomfort and even health risks; on the contrary, if the user is completely covered and the environment is hot, but the air speed is not timely increased or the temperature is not timely reduced, it may cause overheating and sweating, affecting sleep quality. Therefore, accurate perception of the coverage state of the covering on the user's body is the key to improving the use experience of air conditioners in sleep scenarios.

[0004] The current mainstream technical solutions have obvious limitations in solving this problem. On the one hand, visual perception methods based on infrared thermal imaging or visible light cameras can roughly judge the position and outline of the human body, but their imaging effect is severely dependent on the environmental lighting conditions, and it is difficult to penetrate the fabric to distinguish between "skin exposed" and "quilt covered" states. Especially at night in low light or when the quilt material and body temperature are close, the recognition accuracy decreases significantly, and misjudgment and omission are easy to occur. On the other hand, some solutions use piezoelectric sensors in mattresses or wearable devices to indirectly infer the coverage state, which has a certain accuracy, but requires users to actively wear or modify the bed, which has problems such as inconvenience, poor comfort, high deployment cost, and is difficult to popularize.

[0005] In addition, most existing systems only make control decisions based on macro parameters such as room temperature, humidity or whether a human is present, lack fine recognition ability for the user's local coverage state, such as whether the feet, chest and abdomen are exposed, and thus cannot achieve targeted wind direction and speed adjustment. The control logic is static and the response is lagging, and it cannot adapt to the dynamic changes such as the user's frequent turning over and the quilt sliding during sleep. Even worse, due to misjudgment of the coverage state, improper air supply strategies are executed, which further aggravates the user's discomfort.

[0006] In summary, the prior art still has a technical gap in non-contact, high-precision, all-weather sensing of the covering state of the covering on the sleeping user. Due to the inability to accurately grasp the actual heat exposure of the user in real time, the air conditioning system is difficult to make wind feeling adjustment decisions that meet the real needs, ultimately leading to a decrease in thermal comfort during sleep and poor user experience. Therefore, there is an urgent need for an air conditioner control method that can overcome the above-mentioned defects, accurately identify the covering state, and dynamically optimize the air supply strategy, in order to truly improve the practicality and user satisfaction of intelligent air conditioners in sleep scenarios. SUMMARY

[0007] Embodiments of the present application provide an air conditioner control method, device, equipment and medium in a sleep application scenario, aiming to solve the technical problem of poor user experience of intelligent air conditioners in a sleep scenario due to the inability to accurately sense the covering state of the covering on the user during sleep.

[0008] In a first aspect, embodiments of the present application provide an air conditioner control method in a sleep application scenario, applied to an air conditioner control system, the air conditioner control system comprising a millimeter wave radar, the method comprising: continuously monitoring a target user's sleep area by the millimeter wave radar to obtain point cloud data of the target user; based on the point cloud data, identifying the covering state of the covering on the target user in a sleep state; based on the identified covering state of the covering, generating a wind feeling adjustment instruction according to a preset rule; sending the wind feeling adjustment instruction to an air conditioner, the air conditioner adjusting the air supply parameters based on the wind feeling adjustment instruction; continuously monitoring the behavior state changes of the target user, and adaptively optimizing the control strategy on which the wind feeling adjustment instruction is generated based on the behavior state changes.

[0009] In a second aspect, embodiments of the present application further provide an air conditioner control device in a sleep application scenario, for executing the air conditioner control method in a sleep application scenario as described above.

[0010] In a third aspect, embodiments of the present application further provide a computer device, comprising a memory and a processor connected to the memory; the memory is used to store a computer program; the processor is used to run the computer program stored in the memory to execute the steps of the air conditioner control method in a sleep application scenario described above.

[0011] In a fourth aspect, embodiments of the present application further provide a computer readable storage medium, the storage medium stores a computer program, the computer program comprises program instructions, the program instructions can realize the steps of the air conditioner control method in a sleep application scenario described above when executed by a processor.

[0012] Compared with the prior art, the present application has the beneficial effects that: In the technical scheme of the present application, the air conditioner control method in the sleep application scenario continuously acquires point cloud data of the user sleep area through the millimeter wave radar, accurately identifies the covering state of the covering of each part of the body, and overcomes the defects that the traditional infrared or temperature and humidity sensor cannot penetrate the fabric and is difficult to judge the local exposure. Based on the identification result, the system generates wind direction, wind speed and other wind feeling adjustment instructions in real time, so that the air supply of the air conditioner avoids direct blowing to the exposed part, and the thermal comfort is improved. By continuously monitoring the body movement frequency and the number of turning over of the user before and after adjustment, the adjustment effect is evaluated, and the positive or negative weight of the control rule is adjusted, so as to realize the adaptive optimization of the strategy. The problems of air supply discomfort, easy to be chilled and adjustment lag caused by the inability to accurately perceive the covering state in the prior art are solved, and the use experience of the air conditioner in the sleep scenario is significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0013] In order to more clearly illustrate the technical scheme in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0014] Figure 1 The flow chart of the air conditioner control method in the sleep application scenario provided by the present application is shown in the figure. Figure 2 The first sub-flow chart of the air conditioner control method in the sleep application scenario provided by the present application is shown in the figure. Figure 3 The second sub-flow chart of the air conditioner control method in the sleep application scenario provided by the present application is shown in the figure. Figure 4 The third sub-flow chart of the air conditioner control method in the sleep application scenario provided by the present application is shown in the figure. Figure 5 The fourth sub-flow chart of the air conditioner control method in the sleep application scenario provided by the present application is shown in the figure. Figure 6 The fifth sub-flow chart of the air conditioner control method in the sleep application scenario provided by the present application is shown in the figure. Figure 7 The sixth sub-flow chart of the air conditioner control method in the sleep application scenario provided by the present application is shown in the figure. Figure 8 The schematic block diagram of the unit of the air conditioner control device in the sleep application scenario provided by the present application is shown in the figure. Figure 9 The schematic block diagram of the computer device provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0015] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of the present application.

[0016] It should be understood that the terms "comprising" and "including" as used in the specification and the appended claims indicate the presence of the described features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0017] It should also be understood that the terms used in the present application specification are only for the purpose of describing the embodiments and are not intended to limit the present application. As used in the present application specification and the appended claims, the singular forms "a", "an" and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0018] It should be further understood that the term "and / or" as used in the present application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations thereof, and includes these combinations.

[0019] The present application discloses an air conditioner control method in a sleep application scenario to solve the technical problem of poor user experience of an air conditioner in a sleep state due to the inability to accurately perceive the covering state of the covering on the user's body in the sleep state in the prior art. The method is applied to an air conditioner control system, which is deployed in a home bedroom environment and includes a millimeter wave radar. The millimeter wave radar is integrated on the upper part of the air conditioner body or independently installed on the wall surface above the bedside. The transmitting antenna and receiving antenna array form a multiple-input multiple-output structure, the working frequency band is located in the 60GHz or 77GHz frequency band, and the millimeter wave radar has high spatial resolution and micro-motion detection capability.

[0020] With reference to Figures 1 to 7 , the air conditioner control method in the sleep application scenario includes the following steps: S110, continuously monitoring a target user's sleep area by the millimeter wave radar to obtain point cloud data of the target user; S120, identifying the covering state of the target user in the sleep state based on the point cloud data; S130, generating a wind feeling adjustment instruction according to a preset rule based on the identified covering state; 140. sending the air feeling adjustment instruction to an air conditioner, and adjusting a supply air parameter based on the air feeling adjustment instruction; 150. continuously monitoring a behavior state change of the target user, and adaptively optimizing a control strategy for generating the air feeling adjustment instruction based on the behavior state change.

[0021] After the system is started, the millimeter wave radar continuously scans the target user's sleep area at an update frequency of 10 frames per second, transmits a frequency-modulated continuous wave signal, and receives a reflected echo signal reflected by the human body and the cover. Through the range-Doppler processing and the angle of arrival estimation algorithm, the original echo data is parsed into point cloud data containing three-dimensional spatial coordinates, reflection intensity and radial velocity information. Each frame of point cloud data represents the spatial distribution state of all effective scattering points in the bed area at the current time. To ensure the accuracy of the monitoring range, the system performs initialization calibration when it is first run: in the absence of people, the background point cloud is collected, the height plane where the bed surface is located is identified by using the random sample consensus algorithm, and the region of interest (ROI) covering the entire bed surface is established in combination with the user's preset or default bed size, such as 1.5m x 2.0m. All subsequent point cloud processing is limited to this area to exclude the interference of non-related objects such as bedside cabinets and lamps.

[0022] Based on the obtained point cloud data, the system performs cover coverage state recognition. First, the density clustering algorithm is used for clustering analysis of the point cloud to identify the reflection cluster corresponding to the human body. By setting the neighborhood radius and the minimum number of points, and combining the length, width and point cloud number threshold of the bounding box, abnormal clusters with too small or too large size are removed. At the same time, according to the bed surface long axis direction, the spatial center constraint condition is set, and only the cluster group with its center located within a certain range of the bed body center line is reserved, so as to effectively exclude static interference targets such as bed bars and pillows, and to retain the effective human body area. Subsequently, the system automatically divides the human body area into four sub-areas of head, chest and abdomen, legs and feet according to the height distribution difference of the point cloud in the Z axis, in combination with the geometric shape features in the X-Y plane, such as point cloud density, bounding box proportion, and historical motion trajectory between consecutive frames, such as the periodic micro-motion of the chest and abdomen due to respiration. For each sub-area, the system extracts its radar reflection features, including reflection intensity, point cloud density, double-layer reflection structure in the depth direction, and micro-motion signal visibility. The above multi-dimensional features are input into a lightweight convolutional neural network, such as a classification model with MobileNet-V2 as the backbone, to output the coverage state label of each sub-area, and determine whether it is "covered", "exposed" or "uncertain"; and through the sliding window mechanism, time consistency verification is performed to ensure that the recognition result is stable and reliable.

[0023] After identifying the target user's covering state, the system enters the wind-sensation adjustment decision stage. The identification result is input to the wind-sensation strategy engine together with the current environmental parameters such as room temperature, humidity, and the user's pre-set preference information, such as high sensitivity to the feet and preference for low wind speed. The engine makes logical judgments based on the pre-set rule tree structure. For example, when "bare feet" and "room temperature below 24°C" are detected, the combined instruction of "wind direction up to 60°, wind speed reduced by one level, and set temperature increased by 1°C" is triggered. When "fully covered and in deep sleep stage", the low wind speed and constant temperature operation mode is maintained. The control strategy supports two modes. For continuous and stable covering state changes, the continuous control mode is adopted for gradual adjustment to avoid sudden changes causing discomfort; for sudden situations such as turning over causing the quilt to suddenly slide off, the instantaneous intervention mode is started to immediately execute rapid protective adjustment. The generated wind-sensation adjustment instructions are sent to the air conditioner main control unit via the communication module. Before issuance, the system will check the current operation mode of the air conditioner. If there is a conflict, the instruction priority will be adjusted or execution will be temporarily suspended to ensure effective control.

[0024] After executing the wind-sensation adjustment instructions, the system enters the feedback and optimization stage. The millimeter wave radar continues to monitor the target user's behavior state changes, focusing on analyzing the body movement frequency and turning over times per unit time before and after adjustment. If the body movement frequency significantly decreases and the posture tends to be stable after adjustment, it is determined that the adjustment is effective, and the system will adjust the rule weight corresponding to this successful strategy. Conversely, if the body movement frequency does not decrease but increases, it is considered as a failed adjustment, and the corresponding rule weight is adjusted downward, and the state determination threshold is adjusted appropriately in subsequent identification. This process realizes adaptive optimization of the control strategy based on the idea of reinforcement learning, enabling the system to gradually learn and converge to the best wind-sensation regulation path that meets the user's individual habits. In addition, sleep state information can also be integrated to determine the user's sleep stage through the respiratory rate and heart rate variability extracted by the millimeter wave radar. A more conservative adjustment strategy is adopted during deep sleep to avoid waking up, thereby realizing fine wind-sensation control driven by the coordination of covering state and physiological rhythm.

[0025] In an embodiment, the step of S110 further includes the following steps before the step of S110: S160, scanning the unmanned environment by the millimeter wave radar to establish a static background model; S170, identifying the bed surface range based on the static background model to determine the region of interest for the target user's sleep.

[0026] Before the target user's sleeping area is continuously monitored by the millimeter wave radar, an environment initialization and region of interest (ROI) construction step is also performed to ensure the accuracy of subsequent human body recognition and coverage state judgment. Specifically, after the air conditioner control system is first deployed or the user starts the corresponding air conditioner control mode for scene calibration, the system requires the sleeping area to be unoccupied and controls the millimeter wave radar to perform a static environment scan on the bedroom space for a period of time to continuously collect multiple frames of point cloud data. The system performs statistical processing on these point cloud frames without personnel interference in the time dimension, for example, calculates the average reflection intensity and position stability of each spatial position point, and constructs a stable static background model that records the reflection characteristics and spatial distribution of fixed objects in the room, such as bed frames, bedside tables, walls, and floors. On this basis, the system focuses on analyzing the point cloud distribution in the bed height interval, uses a plane fitting algorithm to search for a large-area, height-consistent, and rectangularly distributed flat reflection area, and identifies this area as the bed surface plane. To improve the identification robustness, the system compares the size of the fitted plane with the standard size of the bed body input by the user in advance or by default, and if the size deviation is within the preset acceptable range, the plane is confirmed as the effective bed surface. Subsequently, the system expands a certain preset range outward in the horizontal direction based on the boundary of the fitted plane to construct a three-dimensional space region slightly larger than the actual bed body as the region of interest for subsequent human body detection and coverage state recognition. The establishment of this region of interest enables the system to focus only on the bed surface and the space directly above it in subsequent operation, effectively shielding the interference caused by personnel walking around the bed, window curtains swinging, or other furniture reflection. In actual operation, if the room layout changes, such as the bed body moving, the user can re-trigger the initialization process, and the system will update the static background model and ROI to ensure the long-term stability and adaptability of the perception system. In addition, the initialization process can be guided and prompted through the user interaction interface to improve the operation convenience. By establishing a static background model and accurately delineating the region of interest, this embodiment provides a reliable environmental benchmark for subsequent coverage state recognition based on point cloud data, significantly improving the anti-interference ability and recognition accuracy of the system in complex home environments.

[0027] In an embodiment, the step S120 comprises the following steps: S121, performing clustering processing on the point cloud data to extract a point cloud cluster corresponding to the human body of the target user; S122, performing size filtering and spatial position constraint on the extracted point cloud cluster to exclude non-human body interference targets and obtain an effective human body region; S123, dividing the effective human body region into multiple body sub-regions based on the spatial rules of human anatomy structure according to the height distribution, geometric shape features, and historical motion trajectory of the target user in the space in the point cloud data.

[0028] The covering state of the target user in the sleep state is identified based on the point cloud data. First, the original point cloud data output by the millimeter wave radar is extracted and the human body area is confirmed. After obtaining each frame of point cloud data, the system limits it in the previously established range of the region of interest, and further screens the effective points located in the height band of +5cm to +60cm above the bed surface reference plane to exclude the interference of ground reflection and high-hanging objects. It should be noted that the so-called +5cm to +60cm above the bed surface reference plane is the height of the region where the human body usually sits or lies. Subsequently, the system processes the screened point cloud using a density clustering algorithm, preferably using a density-based spatial clustering application algorithm to group spatial points. The algorithm divides the points close in distance and densely distributed in space into the same cluster group by setting the neighborhood radius and the minimum number of points, thereby separating the possible human body reflection target from the background noise. After clustering is completed, the system analyzes each generated point cloud cluster, calculates the length, width, height of its three-dimensional bounding box, and the number of point clouds contained. On this basis, size filtering is performed, only clusters with a length of 0.6m to 2.0m, a width of 0.3m to 0.9m, a height of 0.1m to 0.6m, and a point cloud number exceeding a preset threshold, such as no less than 100 points, are retained, and abnormal clusters with too small size, such as clothes swinging, or too large size, such as multiple people overlapping, are excluded. At the same time, the system applies a spatial position constraint, defines a "core band" area along the long axis of the bed body with a width of ±40cm in the transverse direction of the bed body, and only retains point cloud clusters with their centers located in the core band; for clusters close to the edge of the bed, such as ≤10cm from the edge of the bed and with a small size, they are determined as static object reflections such as bedside cabinets, bed bars, etc. and are excluded. Through the above-mentioned double screening of size filtering and spatial position constraint, the system obtains one or more effective human body areas, ensuring that the subsequent processing object is the real user target. After confirming the effective human body area, the system further performs human body partitioning, i.e., dividing the overall human body area into multiple body sub-areas to support local covering state recognition. This process is based on the spatial rules of human body anatomy, combined with the height distribution, geometric shape features and historical motion trajectory in the point cloud data for comprehensive judgment. The system first projects the point cloud along the long axis direction of the bed surface (X axis), analyzes the density distribution in the projection curve, and identifies the chest and abdomen area with high point density and the head and foot end with low density. At the same time, according to the height difference in the Z axis direction, the head with the highest Z value, the chest and abdomen with the second highest Z value and accompanied by slight breathing movement, the legs with the longest length and moderate height, and the foot with the lowest Z value and sparse point cloud are distinguished. Combined with the geometric shape features in the X-Y plane, such as the chest and abdomen area showing large-area continuous distribution and the foot area showing narrow and long strip shape, the partitioning accuracy is further enhanced.

[0029] To improve the time continuity, the system also introduces historical motion trajectory analysis to track the position changes of the center of mass of each sub-region in consecutive frames. For example, the chest and abdomen region exhibits periodic up-and-down micro-motions due to breathing at a frequency of about 12-20 times per minute, the leg region exhibits overall displacement when turning over, and the head region is relatively stable. The system uses this dynamic feature to perform temporal smoothing on the partition boundaries to avoid jittering of the partition results caused by instantaneous body motion or noise. By fusing the height distribution, geometric shape, and historical motion trajectory, the system automatically divides the effective human body region into four body sub-regions, namely, the head, chest and abdomen, leg, and foot, providing a spatial basis for subsequent coverage state recognition and differentiated wind feeling adjustment for different parts. This partitioning process is completely automatic and based on the non-contact sensing capability of the millimeter wave radar, and is robust and practical for different body types and sleeping positions.

[0030] In an embodiment, the step of S120 further includes the following steps: S124, for each body sub-region, extracting the corresponding radar reflection features, including reflection intensity, point cloud density, occlusion depth relationship, and micro-motion features; S125, inputting the radar reflection features into a pre-trained state classification model to output the coverage state label of each body sub-region.

[0031] Based on the point cloud data, the coverage state of the target user in the sleep state is recognized. After the extraction of the effective human body region and the division of the body sub-regions, the system further performs fine recognition of the coverage state for each body sub-region. Specifically, the system extracts the corresponding radar reflection features of the body sub-regions, such as the head, chest and abdomen, leg, and foot, which are divided. These features are derived from the differences in electromagnetic wave responses of different materials and spatial structures to millimeter wave radar, and can effectively distinguish between the state of human skin exposure and the state of quilt coverage.

[0032] The extracted radar reflection features include the following four types: First, the reflection intensity, which refers to the average echo intensity of the point cloud in the sub-region. Since human tissues, especially water-containing skin and muscles, have strong reflection ability to millimeter waves, while common quilt materials such as cotton and polyester fibers have scattering and absorption effects on signals, resulting in attenuation of echo energy, the reflection intensity of the exposed area is usually 3-5 dB higher than that of the covered area, which constitutes the basis for state discrimination; Second, the point cloud density, which refers to the number of effective scattering points per unit space volume. The exposed human surface is continuous and smooth, which is manifested as a high-density, clear-edged cluster in the point cloud, while the quilt fiber structure is random and porous, resulting in sparse and scattered reflection points, so the point cloud density can effectively reflect the surface continuity and occlusion degree; Third, the occlusion depth relationship, that is, whether there is a "double-layer reflection" structure at the same horizontal position is analyzed to determine the occlusion relationship; when a body region is covered by a quilt, the radar signal will be reflected on the surface of the quilt, that is, the shallow layer, and the surface of the human body, that is, the deep layer, in turn, forming two separate reflection points in the distance dimension, and the distance between the two is usually a few centimeters, which depends on the thickness of the quilt. The system analyzes whether the double-layer structure exists through the distance-angle diagram as key evidence of the covering state; Fourth, the micro-motion feature, that is, whether the periodic micro-displacement caused by breathing or heartbeat can be detected in the sub-region; the micro-motion signal of the exposed region can be directly reflected in the echo phase change, showing a stable periodic jitter, while the micro-motion signal amplitude of the covered region is significantly weakened or even disappears due to the damping effect of the fabric. The system extracts the feature through Doppler spectrum analysis and uses it to verify the reliability of the covering state.

[0033] After the above multi-dimensional feature extraction is completed, the feature vectors of each body sub-region, such as the average reflection intensity, point cloud density, double-layer reflection existence flag, and micro-motion signal power spectrum peak value, are normalized and spliced into a fixed-dimensional input vector. The input vector is sent to a pre-trained lightweight state classification model for inference. The model preferably adopts a convolutional neural network structure and is supervised learned using a large amount of labeled millimeter wave point cloud data in the training stage, so that it can adaptively fuse multi-dimensional features and output classification results. The millimeter wave point cloud data contains "covered" and "exposed" samples under different body types, sleeping positions, and quilt materials. The model output is the covering state label of each body sub-region, including "covered", "exposed", or "uncertain". To improve the recognition stability, the system further uses a sliding window mechanism to judge the consistency of the classification results of continuous multiple frames, and only when the majority of frames output the same label, the state is regarded as the final recognition result and submitted to the subsequent control link. Through the fusion of reflection intensity, point cloud density, occlusion depth relationship, and micro-motion feature, and the intelligent classification of the lightweight deep learning model, the embodiment realizes high-precision, non-contact identification of the covering state of each part of the human body, providing a reliable sensing basis for personalized air feeling adjustment.

[0034] In an embodiment, the step of S120 further includes the following steps: S126, the covering state labels of continuous multiple frames are verified for consistency through a sliding window mechanism; S127, when the covering state label remains stable within a preset time length, the covering state recognition is confirmed as a valid recognition result.

[0035] The system performs consistency check on the cover state labels of continuous multiple frames of output, and smoothes the state change process by using a sliding window mechanism. The time length of the sliding window is set according to the typical behavior characteristics in the sleep scene, and is usually 2-3 seconds, corresponding to 20-30 frames of data collected by the millimeter wave radar at a frequency of 10 frames per second. The system independently maintains a sliding window buffer area for each body sub-region, which is used to store all state label sequences of the region in the latest window period. In each processing period, the system performs statistical analysis on the label sequences in the window, calculates the occurrence frequencies of the "covered", "exposed" and "uncertain" three types of labels, and determines the current stable state if a certain state label occupies the absolute majority in the window. If there is no obvious dominant state or frequent switching, the last effective state is maintained or marked as "uncertain". Only when a certain state label continuously maintains the dominant position in the sliding window and the duration reaches the preset length, that is, it completely passes through a sliding window period, the system confirms that the cover state is an effective recognition result, and submits it to the wind feeling strategy engine for control decision. For example, when the user turns over and causes the quilt to slide temporarily, the leg area may appear "exposed" label in a few frames, but if the state does not exist continuously for 2-3 seconds, it is not considered as an effective state change, and the system does not trigger the air conditioner adjustment, thereby avoiding frequent miscontrol caused by instantaneous action. On the contrary, if the user keeps the feet exposed for a long time, the system identifies "exposed" as the dominant label in multiple sliding windows, and confirms the effective state and starts the corresponding wind feeling adjustment. This mechanism can also combine with the minimum holding time constraint of state switching, for example, it is required that the switching between "covered" and "exposed" should satisfy at least 1.5 seconds of continuous duration before changing, further suppressing state jitter. By using the sliding window mechanism for time consistency check, the embodiment significantly improves the robustness and reliability of cover state recognition, ensures that the air conditioner control instruction is triggered only when the user's real and stable cover state changes, effectively prevents misoperation caused by point cloud fluctuation or temporary shielding, and improves the applicability and user experience of the system in actual complex sleep environment.

[0036] In an embodiment, the step S130 comprises the following steps: S131, inputting the cover state of the cover, the environmental parameters of the target user's sleep area, and the preset preferences of the target user into a preset wind feeling strategy engine; S132, generating a wind feeling adjustment instruction by the wind feeling strategy engine based on a preset rule tree; S133, generating a gradual adjustment instruction of a continuous control mode when the cover state is continuously stable; S134, generating a one-time adjustment instruction of an instantaneous intervention mode when the cover state changes suddenly.

[0037] After confirming that the coverage state of each body sub-region is a valid recognition result, the system inputs the coverage state of the cover as a core input condition, together with environmental parameters of the target user's sleep area, such as room temperature, humidity obtained by an integrated sensor or an external device, and preset preferences of the target user, such as "sensitive to feet", "prefer low wind speed", "night temperature not lower than 24℃", and other personalized configurations set through a mobile terminal application or an air conditioner panel, into a preset wind feeling strategy engine. The engine organizes control logic in a rule tree structure, and the nodes of the tree include condition judgment nodes, such as "whether the feet are exposed", "whether the room temperature is lower than 24℃", and "whether the user is in a deep sleep stage", and execution action nodes, such as "wind direction up to 60°", "wind speed reduced by one level", and "set temperature increased by 1℃", and the rule priority is set according to the health protection principle, for example, the comfort needs of exposed limbs, the elderly or children are given priority. The system traverses the rule tree, matches the current input condition, and outputs one or more sets of candidate wind feeling adjustment instructions. For example, when "exposed feet" and "room temperature of 22℃" are identified, the rule "if the user's preset preference allows, execute wind direction up + wind speed reduced to 1 level + set temperature + 1℃ for 15 minutes" is triggered. If it is detected that the user is a child at the same time, the protection strategy is automatically enhanced, such as increasing the temperature increase amplitude to 1.5℃.

[0038] To adapt to different changing paces of coverage states, the wind feeling strategy engine supports two control modes: when the system determines that a certain coverage state, such as the chest and abdomen being continuously exposed for more than a preset time, is a slow-changing process through a sliding window mechanism, it starts a continuous control mode and generates progressive adjustment instructions, such as reducing the wind speed in stages or slowly increasing the temperature, to avoid sudden parameter changes causing user discomfort, while also considering energy efficiency; when a sudden change in the coverage state is detected, such as the quilt instantly sliding off due to turning over, the feet changing from "covered" to "exposed", and the change amplitude exceeding a threshold, it is determined to be an emergency, and the instantaneous intervention mode is immediately started to generate one-time rapid adjustment instructions, such as "immediately close the horizontal swing of the air deflector and vertically lift the blades to the maximum angle", to avoid cold wind blowing directly at the fastest response speed. The instruction is executed once and then completed, and whether to continue adjusting in the next period is determined by the recognition result. The switching between the two modes is dynamically determined by the strategy engine according to the state change rate and persistence, ensuring that the system can respond to sudden exposure risks and achieve long-term comfortable smooth regulation.

[0039] In an embodiment, the step of S150 includes: S151, comparing the body movement frequency and the number of turning over before and after the execution of the wind feeling adjustment instruction; S152, according to the change trend of body movement frequency or the number of turning over, evaluating the execution effect of the air feeling adjustment instruction; S153, if the evaluation result shows that the body movement frequency or the number of turning over is reduced, a positive weight adjustment is given to the control rule on which the air feeling adjustment instruction is generated; S154, if the evaluation result shows that the body movement frequency or the number of turning over is increased, a negative weight adjustment is given to the control rule.

[0040] The process of continuously monitoring the behavior state changes of the target user and adaptively optimizing the control strategy based on the behavior state changes constructs a closed-loop learning mechanism driven by user physiological feedback, aiming to make the air conditioner wind feeling adjustment strategy gradually approach the user's personalized comfort preference. Specifically, after the wind feeling adjustment instruction is sent and executed, the system enters the effect evaluation stage, continuously collects point cloud data through the millimeter wave radar, extracts the body motion behavior characteristics of the target user before and after adjustment for a period of time, and focuses on analyzing the changes of body motion frequency and the number of turning over. The body motion frequency is obtained by calculating the number of displacement fluctuations of the whole body or local area in space per unit time, and the number of turning over is identified and counted by analyzing the significant displacement of the body center of mass on the bed surface in the transverse (Y-axis) direction and the change of body posture, such as turning from supine to lateral. The system compares the baseline body motion frequency and the number of turning over before adjustment with the actual values after adjustment, and evaluates the execution effect of the wind feeling adjustment instruction according to the change trend. If the body motion frequency significantly decreases and the number of turning over decreases after adjustment, and the posture tends to be stable, it is determined that the user has obtained better thermal comfort due to the air supply adjustment, and the evaluation result is "effective". On the contrary, if the body motion frequency does not decrease but increases or the number of turning over is more frequent, it indicates that the current adjustment may cause discomfort, and the evaluation result is "ineffective". Based on the evaluation result, the system adjusts the dynamic weight of the control rule on which the wind feeling adjustment instruction of this time is generated. Specifically, each control rule is associated with a variable execution priority weight in the wind feeling strategy engine; when the evaluation result shows that the adjustment is effective, the system gives positive weight adjustment to the rule, for example, increases the weight value by a certain proportion, so that it is preferentially called in subsequent similar scenarios; when the evaluation result shows that the adjustment is ineffective, negative weight adjustment is given, for example, the weight value is reduced by a certain proportion, reducing the triggering probability. This weight adjustment process simulates the reward and punishment mechanism in reinforcement learning, enabling the system to continuously learn and converge to the optimal control path suitable for the user in the long-term operation. To further improve the learning stability, the system can set a minimum weight threshold and a maximum weight upper limit to prevent the rule from being completely disabled or excessively prioritized. At the same time, the weight update can use a moving average method to avoid the impact of a single misjudgment on the strategy. In addition, the system can also record the environmental context of each adjustment, such as room temperature, season, and sleep stage, to realize differentiated strategy optimization based on scenarios. By using body motion frequency and the number of turning over as objective feedback indicators and adaptively adjusting the weight of the control rule based on this, the embodiment realizes intelligent strategy evolution without user intervention, significantly improving the personalized service level and sleep environment adaptation ability of the air conditioning system in the long-term use.

[0041] In an embodiment, the air conditioner control system also supports differentiated coverage state recognition and collaborative wind feel control in the multi-user simultaneous sleep scenario. When the millimeter wave radar detects two or more valid human body regions meeting the size and position constraints in the bed surface area of interest, the system automatically switches to the multi-user recognition mode, and independently establishes a point cloud cluster tracking and body sub-region division model for each user. The coverage state of each user is respectively extracted and classified to form an independent state label set. The wind feel strategy engine adopts a multi-channel decision logic, first generates candidate wind feel adjustment instructions for each user, and then merges them into a global unified instruction through a priority arbitration mechanism. The priority can be set based on user attributes, such as giving priority to the elderly or children over adults, or according to user preset preferences. For example, when user A is a child with full body coverage and user B is an elderly person with bare chest, the system prioritizes the protective strategy of "reducing wind speed and moderately increasing temperature", and limits the upper temperature limit to avoid overheating of user A. In addition, the system can integrate sleep state information to optimize the control rhythm: through the respiratory rhythm and body movement characteristics extracted by the millimeter wave radar, combined with sleep stage data provided by external devices such as Bluetooth-connected smart bands, the system determines whether the user is currently in deep sleep, light sleep or rapid eye movement. In the deep sleep stage, a more conservative adjustment strategy such as adjusting the wind direction only without changing the temperature is executed to avoid interfering with sleep continuity. The air conditioner control instructions are sent through multiple communication interfaces such as Wi-Fi, infrared or Bluetooth, compatible with mainstream brand protocols, and verified before sending to check whether the air conditioner is currently in energy-saving lock, to avoid instruction conflicts. When the coverage state is identified as "uncertain" for a long time or there is a conflict between multiple sub-region states, the system enters a conservative mode, executes the default protection strategy of "lifting the wind direction, reducing the wind speed, and moderately increasing the temperature", and increases the subsequent state confirmation threshold to ensure safe operation.

[0042] In summary, the air conditioner control method in the sleep application scenario provided by the present application integrates high-precision perception of millimeter wave radar, multi-region coverage state recognition, closed-loop feedback optimization, and intelligent decision control, breaks through the technical bottleneck of traditional intelligent air conditioners in the sleep scenario, and has outstanding practicality and innovation. This method does not require contact with the user and does not depend on light conditions, and has high privacy and strong environmental adaptability, and can be widely used in home bedrooms, elderly care, infant care, intelligent mattress matching, and high-end hotel rooms. Especially in the elderly and infant groups, it can effectively prevent health risks caused by night cooling and improve the level of active health management. With the continuous decline in the cost of millimeter wave radar and the improvement of edge computing capabilities, this scheme is easy to integrate into existing air conditioning systems or deployed as an independent module for upgrading. Its characteristics of supporting multi-user identification, sleep state fusion, and adaptive learning are in line with the trend of smart homes towards "unconscious, personalized, and scenario-based" development, and are expected to become the core control logic of the next generation of intelligent sleep environment. Combined with the Internet of Things platform, it can also be extended to link with curtains, lights, humidifiers, and other devices to build a whole-house intelligent sleep ecosystem. Therefore, the present application not only has strong technical feasibility, but also has clear market positioning, and has broad application prospects and industrial value.

[0043] Figure 8 is a schematic block diagram of an air conditioner control device 600 in a sleep application scenario provided by an embodiment of the present application. As shown in Figure 8 Corresponding to the above-mentioned air conditioner control method in the sleep application scenario, the present application also provides an air conditioner control device 600 in the sleep application scenario. The air conditioner control device 600 in the sleep application scenario includes units for executing the above-mentioned air conditioner control method in the sleep application scenario, and the device can be configured in a desktop computer, a tablet computer, a smart phone, and the like. Terminal.

[0044] Specifically, please refer to Figure 8 The air conditioner control device 600 in the sleep application scenario includes: A data acquisition unit 610 is configured to continuously monitor the target user's sleep area by the millimeter wave radar, and acquire point cloud data of the target user; A state judgment unit 620 is configured to identify the coverage state of the target user in the sleep state based on the point cloud data; An instruction generation unit 630 is configured to generate a wind feeling adjustment instruction according to a preset rule based on the identified coverage state of the coverage; An instruction control unit 640 is configured to send the wind feeling adjustment instruction to the air conditioner, and the air conditioner adjusts the air supply parameter based on the wind feeling adjustment instruction; The feedback optimization unit 650 is used to continuously monitor changes in the target user's behavior state and adaptively optimize the control strategy on which the wind adjustment command is generated based on the changes in the behavior state.

[0045] In one embodiment, the data acquisition unit 610 further includes: Background model building unit is used to scan the unmanned environment with millimeter-wave radar and establish a static background model; The ROI determination unit is used to identify the bed surface range based on the static background model and determine the region of interest for the target user's sleep.

[0046] Furthermore, the state determination unit 620 includes: The clustering processing unit is used to perform clustering processing on the point cloud data and extract the point cloud clusters corresponding to the human body of the target user; The interference elimination unit is used to perform size filtering and spatial position constraints on the extracted point cloud clusters to eliminate non-human interference targets and obtain effective human areas. The sub-region division unit is used to divide the effective human body region into multiple body sub-regions based on the spatial patterns of human anatomy, according to the height distribution, geometric features, and historical movement trajectories of the target user in the point cloud data.

[0047] Furthermore, the state determination unit 620 also includes: The reflection feature extraction unit is used to extract the corresponding radar reflection features for each of the body sub-regions. The radar reflection features include reflection intensity, point cloud density, occlusion depth relationship, and micro-motion features. The coverage state label generation unit is used to input the radar reflection features into a pre-trained state classification model and output the coverage state labels for each body sub-region.

[0048] Furthermore, the state determination unit 620 also includes: A consistency verification unit is used to perform consistency verification on the coverage status labels of multiple consecutive frames using a sliding window mechanism. The validity confirmation unit is used to confirm that the coverage status identification is a valid identification result when the coverage status label remains stable within a preset time period.

[0049] In one embodiment, the instruction generation unit 630 includes: The strategy engine input unit is used to input the coverage status of the covering, the environmental parameters of the target user's sleeping area, and the target user's preset preferences into the preset wind-sensing strategy engine. The strategy engine judgment unit is used to generate wind adjustment instructions based on a preset rule tree through the wind-sensing strategy engine; The first judgment unit is used to generate a progressive adjustment command for the continuous control mode when the coverage state remains stable. The second judgment unit is used to generate a one-time adjustment command for the instantaneous intervention mode when the coverage state changes abruptly.

[0050] In one embodiment, the feedback optimization unit 650 includes: The limb state comparison unit is used to compare the body movement frequency and the number of times the wind-feeling adjustment command is executed before and after the execution of the command. The execution effect evaluation unit is used to evaluate the execution effect of wind adjustment commands based on the changing trend of body movement frequency or number of rollovers; A positive feedback unit is used to assign a positive weight adjustment to the control rule on which the wind adjustment command is based if the evaluation result shows a decrease in body movement frequency or number of times of turning over. A negative feedback unit is used to apply a negative weight adjustment to the control rule if the evaluation results show an increase in body movement frequency or number of rollovers.

[0051] The air conditioner control device 600 in the aforementioned sleep application scenario can be implemented as a computer program, which can, for example... Figure 9 It runs on the computer device shown.

[0052] Please see Figure 9 , Figure 9 This is a schematic block diagram of a computer device 500 provided in an embodiment of this application. The computer device 500 can be a terminal or a server. The terminal can be an electronic device with communication functions, such as a desktop computer, tablet computer, or smartphone. The server can be a standalone server or a server cluster composed of multiple servers.

[0053] See Figure 9 The computer device 500 includes a processor 502, a memory, and a network interface 505 connected via a system bus 501. The memory may include a non-volatile storage medium 503 and internal memory 504.

[0054] The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions that, when executed, cause the processor 502 to perform an air conditioner control method for a sleep application scenario.

[0055] The processor 502 provides computing and control capabilities to support the operation of the entire computer device 500.

[0056] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute an air conditioner control method in a sleep application scenario.

[0057] This network interface 505 is used for network communication with other devices. Those skilled in the art will understand that... Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device 500 to which the present application is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0058] The processor 502 is used to run a computer program 5032 stored in a memory to implement the steps of the above method.

[0059] It should be understood that in the embodiments of this application, the processor 502 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0060] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program includes program instructions and can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.

[0061] Therefore, the present invention also provides a storage medium. This storage medium can be a computer-readable storage medium. The storage medium stores a computer program, wherein the computer program includes program instructions. When executed by a processor, the program instructions cause the processor to perform the steps of the above-described method.

[0062] The storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.

[0063] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0064] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of each unit is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0065] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the device of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0066] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0067] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered 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. An air conditioner control method for a sleep application scenario, characterized in that, Applied to an air conditioning control system, the air conditioning control system including millimeter-wave radar, the method includes: The millimeter-wave radar continuously monitors the target user's sleep area to obtain point cloud data of the target user; Based on the point cloud data, identify the coverage status of the cover when the target user is asleep; Based on the identified coverage status of the covering, a wind adjustment command is generated according to a preset rule; The air conditioning unit sends the air conditioning command to the air conditioning unit, and the air conditioning unit adjusts the air supply parameters based on the air conditioning command. Continuously monitor changes in the target user's behavior and adaptively optimize the control strategy used to generate wind adjustment commands based on these changes.

2. The air conditioner control method for a sleep application scenario according to claim 1, characterized in that, Before the step of continuously monitoring the target user's sleep area using the millimeter-wave radar to obtain the target user's point cloud data, the method further includes: A static background model is established by scanning the unmanned environment using millimeter-wave radar. Based on the static background model, the range of the bed surface is identified, and the region of interest for the target user's sleep is determined.

3. The air conditioner control method for a sleep application scenario according to claim 1, characterized in that, The step of identifying the covering status of a target user in a sleep state based on the point cloud data includes: Clustering is performed on the point cloud data to extract the point cloud clusters corresponding to the target user's human body; The extracted point cloud clusters are subjected to size filtering and spatial position constraints to eliminate non-human interference targets and obtain effective human areas. Based on the height distribution, geometric features, and historical movement trajectories of the target user in the point cloud data, the effective human body region is divided into multiple body sub-regions based on the spatial patterns of human anatomy.

4. The air conditioner control method for a sleep application scenario according to claim 3, characterized in that, The step of identifying the cover status of the target user in a sleep state based on the point cloud data further includes: For each of the body sub-regions, the corresponding radar reflection features are extracted, including reflection intensity, point cloud density, occlusion depth relationship, and micro-motion features. The radar reflection features are input into a pre-trained state classification model, which outputs the coverage state labels for each body sub-region.

5. The air conditioner control method for a sleep application scenario according to claim 4, characterized in that, The step of identifying the cover status of the target user in a sleep state based on the point cloud data further includes: The coverage status labels of multiple consecutive frames are checked for consistency using a sliding window mechanism; When the coverage status label remains stable within a preset time period, the coverage status recognition is confirmed as a valid recognition result.

6. The air conditioner control method for a sleep application scenario according to claim 1, characterized in that, The step of generating wind adjustment commands based on a preset rule tree using the wind-sensing strategy engine includes: The coverage status of the covering, the environmental parameters of the target user's sleeping area, and the target user's preset preferences are input into the preset wind-sensing strategy engine; The wind-sensing strategy engine generates wind-sensing adjustment commands based on a preset rule tree. When the coverage state remains stable, a progressive adjustment command for continuous control mode is generated; When the coverage state changes abruptly, a one-time adjustment command for the instantaneous intervention mode is generated.

7. The air conditioner control method for a sleep application scenario according to claim 1, characterized in that, The step of continuously monitoring changes in the target user's behavioral state and adaptively optimizing the control strategy upon which the wind adjustment command is generated based on these changes includes: The body movement frequency and the number of times the body rolls over are compared before and after the wind adjustment command is executed; The effectiveness of wind-feeling adjustment commands is evaluated based on the trend of changes in body movement frequency or number of rollovers. If the evaluation results show a decrease in body movement frequency or number of times you roll over, then the control rules on which the wind adjustment command is based are given a positive weight adjustment. If the evaluation results show an increase in body movement frequency or number of times the body rolls over, then the control rule is adjusted with a negative weight.

8. An air conditioner control device for a sleep application scenario, characterized in that, Used to perform the air conditioner control method in a sleep application scenario as described in any one of claims 1 to 7.

9. A computer device, characterized in that, The computer device includes a memory and a processor connected to the memory; the memory is used to store a computer program; the processor is used to run the computer program stored in the memory to perform the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which includes program instructions that, when executed by a processor, can implement the steps of the method as described in any one of claims 1 to 7.

Images