Bathroom exhaust control method and device, home host, system and storage medium

Abstract

The invention relates to a bathroom exhaust control method and device, a home host, a system and a storage medium, and the method comprises the steps: obtaining multi-modal data which comprises point cloud data, pressure distribution data, heat radiation distribution data and environment time sequence data in a bathroom space; the point cloud data, the pressure distribution data, the heat radiation distribution data and the environment time sequence data are input into a pre-trained deep learning model, exhaust control parameters are obtained, and the exhaust control parameters comprise exhaust frequency and exhaust power; and based on the air exhaust frequency and the air exhaust power, an air exhaust fan is controlled to exhaust air in the bathroom space. In this way, the point cloud data, the pressure distribution data, the heat radiation distribution data and the environment time sequence data in the bathroom space can be analyzed based on the pre-trained deep learning model, so that the exhaust control parameters are accurately determined, accurate exhaust control is achieved, and the purposes of considering energy consumption and demisting efficiency are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of deep learning, and particularly relates to a bathroom exhaust control method, device, home server, system and storage medium. BACKGROUND

[0002] A bathroom generates a large amount of water vapor due to bathing, and a bathroom exhaust system can reduce humidity and effectively remove mist in the bathroom through exhaust, so the bathroom exhaust system is widely used.

[0003] However, the existing bathroom exhaust system usually controls exhaust based on a fixed humidity threshold or a timing mode, and cannot accurately control exhaust according to actual needs, so there is a problem of energy waste or low mist removal efficiency. Therefore, how to improve the accuracy of bathroom exhaust control has become a technical problem to be solved. SUMMARY

[0004] The present application provides a bathroom exhaust control method, device, home server, system and storage medium to solve the problem that the existing bathroom exhaust system usually controls exhaust based on a fixed humidity threshold or a timing mode, and cannot accurately control exhaust according to actual needs.

[0005] In a first aspect, the present application provides a bathroom exhaust control method, which comprises:

[0006] obtaining multi-modal data, wherein the multi-modal data comprises point cloud data, pressure distribution data, thermal radiation distribution data and environment time series data in a bathroom space;

[0007] inputting the point cloud data, the pressure distribution data, the thermal radiation distribution data and the environment time series data into a pre-trained deep learning model to obtain exhaust control parameters, wherein the exhaust control parameters comprise an exhaust frequency and an exhaust power;

[0008] controlling an exhaust fan to exhaust the bathroom space based on the exhaust frequency and the exhaust power.

[0009] Optionally, the deep learning model comprises a first feature extraction branch, a second feature extraction branch, a feature fusion layer and a detection head.

[0010] The inputting the point cloud data, the pressure distribution data, the thermal radiation distribution data and the environment time series data into a pre-trained deep learning model to obtain exhaust control parameters comprises:

[0011] inputting the point cloud data and the thermal radiation distribution data into the first feature extraction branch for feature extraction to obtain spatial features;

[0012] inputting the point cloud data and the thermal radiation distribution data into the first feature extraction branch for feature extraction to obtain spatial features;

[0013] inputting the spatial features and the time-series-pressure joint features into the feature fusion layer for feature fusion to obtain fusion features;

[0014] inputting the fusion features into the detection head for detection to obtain the exhaust control parameter.

[0015] Optionally, the first feature extraction branch comprises an input layer, a farthest point sampling layer, a multi-layer perception layer, a max-pooling layer and a first full connection layer;

[0016] The inputting the point cloud data and the thermal radiation distribution data into the first feature extraction branch for feature extraction to obtain spatial features comprises:

[0017] inputting the point cloud data and the thermal radiation distribution data into the input layer for splicing to obtain first intermediate data;

[0018] inputting the first intermediate data into the farthest point sampling layer for down-sampling to obtain second intermediate data;

[0019] inputting the second intermediate data into the multi-layer perception layer for point-by-point extraction of local features and dimensionality increase to obtain third intermediate data;

[0020] inputting the third intermediate data into the max-pooling layer for local feature aggregation to obtain fourth intermediate data;

[0021] inputting the fourth intermediate data into the first full connection layer for feature mapping to obtain the spatial features.

[0022] Optionally, the second feature extraction branch comprises a hidden layer, a second full connection layer and a splicing layer;

[0023] The inputting the pressure distribution data and the environment time-series data into the second feature extraction branch for feature extraction to obtain time-series-pressure joint features comprises:

[0024] inputting the environment time-series data into the hidden layer for embedding representation to obtain time-series features;

[0025] inputting the pressure distribution data into the second full connection layer for feature mapping to obtain pressure features;

[0026] inputting the time-series features and the pressure features into the splicing layer for splicing to obtain the time-series-pressure joint features.

[0027] Optionally, after the point cloud data, the pressure distribution data, the thermal radiation distribution data and the environmental time sequence data are input into the pre-trained deep learning model to obtain the exhaust control parameter, the method further comprises:

[0028] desensitizing the fusion features to obtain desensitized features;

[0029] sending the desensitized features, the environmental time sequence data and the exhaust control parameter to a cloud platform, so that the cloud platform stores the desensitized features, the environmental time sequence data and the exhaust control parameter in a preset database, and updates its own deep learning model by federated learning using the desensitized features, the environmental time sequence data and the exhaust control parameter.

[0030] Optionally, after the desensitized features, the environmental time sequence data and the exhaust control parameter are sent to the cloud platform, the method further comprises:

[0031] In the case that the bathroom space needs to be re-controlled, the desensitized features and the current environmental time sequence data of the current user in the bathroom space are obtained, and the desensitized features and the current environmental time sequence data of the current user are sent to the cloud platform, so that the cloud platform matches the desensitized features of the current user with the desensitized features stored in the preset database, and in the case that the desensitized features of the current user match the desensitized features stored in the preset database successfully, the exhaust control parameter most matched with the current environmental time sequence data is determined and returned;

[0032] receiving the exhaust control parameter returned by the cloud platform, and controlling the exhaust according to the exhaust control parameter returned by the cloud platform.

[0033] Optionally, after the exhaust frequency and the exhaust power are controlled based on the exhaust frequency and the exhaust power, the method further comprises:

[0034] real-time acquisition of the humidity value in the bathroom space, and calculation of the difference between the humidity value in the bathroom space and a preset humidity value to obtain a humidity deviation;

[0035] based on the humidity deviation, dynamically correcting the exhaust frequency and the exhaust power.

[0036] In a second aspect, the embodiments of the present application also provide a bathroom exhaust control device, which comprises:

[0037] a first acquisition module for acquiring multi-modal data, wherein the multi-modal data comprises point cloud data, pressure distribution data, thermal radiation distribution data and environmental time sequence data in a bathroom space;

[0038] an input module, configured to input the point cloud data, the pressure distribution data, the thermal radiation distribution data, and the environment time sequence data into a pre-trained deep learning model to obtain an exhaust control parameter, wherein the exhaust control parameter comprises an exhaust frequency and an exhaust power;

[0039] a control module, configured to control an exhaust fan to perform exhaust on the bathroom space based on the exhaust frequency and the exhaust power.

[0040] In a third aspect, the embodiments of the present application further provide a home server, comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory complete mutual communication through the communication bus;

[0041] the memory, configured to store a computer program;

[0042] the processor, configured to execute the program stored on the memory, to implement the bathroom exhaust control method in the first aspect.

[0043] In a fourth aspect, the embodiments of the present application further provide a bathroom exhaust control system, comprising a cloud platform and the home server in the third aspect.

[0044] The cloud platform is in communication connection with the home server.

[0045] In a fifth aspect, the embodiments of the present application further provide a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the bathroom exhaust control method in the first aspect.

[0046] The above technical solution provided by the embodiments of the present application has the following advantages compared with the prior art: the method provided by the embodiments of the present application acquires multi-modal data, wherein the multi-modal data comprises point cloud data, pressure distribution data, thermal radiation distribution data, and environment time sequence data in a bathroom space; the point cloud data, the pressure distribution data, the thermal radiation distribution data, and the environment time sequence data are input into a pre-trained deep learning model to obtain an exhaust control parameter, wherein the exhaust control parameter comprises an exhaust frequency and an exhaust power; and the exhaust fan is controlled to perform exhaust on the bathroom space based on the exhaust frequency and the exhaust power. In this way, the point cloud data, the pressure distribution data, the thermal radiation distribution data, and the environment time sequence data in the bathroom space can be analyzed based on the pre-trained deep learning model, so as to accurately determine the exhaust control parameter, thereby realizing accurate exhaust control and achieving the purpose of balancing energy consumption and defogging efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0047] The accompanying drawings, which are incorporated herein and constitute part of the specification, illustrate embodiments consistent with the application and, together with the description, further serve to explain the principles of the application.

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the accompanying drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows. Obviously, for those of ordinary skill in the art, other drawings can be obtained based on these drawings without any creative effort.

[0049] One or more embodiments are illustrated by way of example in the drawings that are not intended to be limiting of the application, and which will be described in greater detail below. In the drawings, like reference numerals refer to like elements, unless otherwise indicated. The drawings are not necessarily to scale, the emphasis instead being placed upon illustrating the principles of the embodiments.

[0050] Figure 1 A flowchart of a bathroom exhaust control method provided by an embodiment of the present application;

[0051] Figure 2 A structural diagram of a deep learning model provided by an embodiment of the present application;

[0052] Figure 3 A structural diagram of a first feature extraction branch provided by an embodiment of the present application;

[0053] Figure 4 A structural diagram of a second feature extraction branch provided by an embodiment of the present application;

[0054] Figure 5 A structural diagram of a bathroom exhaust control device provided by an embodiment of the present application;

[0055] Figure 6 A structural diagram of a home server provided by an embodiment of the present application;

[0056] Figure 7 A structural diagram of a bathroom exhaust control system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0057] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without any creative effort fall within the scope of protection of the present application.

[0058] The disclosure below provides many different embodiments or examples for implementing various configurations of the application. For the purpose of simplicity, the description below of the specific examples refers only to the described embodiments. It is clearly understood that the description of the specific examples is included in order to fully enable a person of ordinary skill in the art to make and use the application, and is not intended to limit the application in any way. Furthermore, the described embodiments are to be considered as illustrative only and the scope of the application to be indicated by the appended claims rather than the foregoing description. In addition, the reference numbers and / or letters in different examples can be repeated. Such repetition is for the purpose of simplicity and clarity and does not indicate a relationship between the various embodiments and / or arrangements discussed.

[0059] To solve the problem that the existing bathroom exhaust system usually controls exhaust based on a fixed humidity threshold or a timing mode, and cannot accurately control according to actual needs, the application provides a bathroom exhaust control method, device, home hub, system and storage medium, which can improve the accuracy of bathroom exhaust control.

[0060] Referring to Figure 1 , Figure 1 A flowchart of a bathroom exhaust control method provided by an embodiment of the application is shown in FIG. 1. As shown in FIG. 1, the bathroom exhaust control method can include the following steps: Figure 1

[0061] Step S101, acquiring multi-modal data, wherein the multi-modal data includes point cloud data, pressure distribution data, thermal radiation distribution data and environmental time series data in the bathroom space.

[0062] Specifically, the multi-modal data described above can be multi-modal non-visual sensing data. Compared with visual sensing data, non-visual sensing data can avoid the risk of user privacy leakage and improve user privacy security. The multi-modal data can include point cloud data, pressure distribution data, thermal radiation distribution data and environmental time series data in the bathroom space. The point cloud data here can be three-dimensional (3D) point cloud data generated by a millimeter wave radar. As an optional implementation, after acquiring the point cloud data, a dynamic background filtering algorithm can be used to remove the point cloud corresponding to stationary objects in the point cloud data, and then a density-based clustering algorithm (Density-Based Spatial Clustering of Applications with Noise, DBSCAN) is used to cluster and segment the point cloud corresponding to the human body, so that the subsequent model can extract the user contour based on the processed point cloud data to obtain the height, shoulder width, gait cycle and other related features of the human body. The pressure distribution data here can be trajectory standard deviation and movement frequency data detected by a distributed pressure sensor array. The trajectory standard deviation can represent the amplitude of the human body's center of gravity swing, and the movement frequency can represent the speed of the human body movement. As an optional implementation, the trajectory standard deviation and the movement frequency can be determined based on a pressure center trajectory, and the calculation formula of the pressure center trajectory is as follows:​

[0063]

[0064] Where C(t) represents the pressure center at time t, p i,j (t) represents the pressure value of the pressure sensor at grid position (i, j) at time t, x i Represents the horizontal physical coordinate of the pressure sensor in the i-th row, y j This represents the vertical physical coordinates of the j-th column pressure sensor. The thermal radiation distribution data here can be data collected using an infrared pyroelectric sensor, facilitating subsequent model extraction of heat map gradient features based on this data to identify the location and number of human bodies. The environmental time-series data can include water temperature time-series data, humidity time-series data, and room temperature time-series data. The water temperature time-series data can be time-series data formed by collecting water temperature data at different times using a water temperature sensor; the humidity time-series data can be time-series data formed by collecting humidity data at different times within the bathroom area using a humidity sensor; and the room temperature time-series data can be time-series data formed by collecting room temperature data at different times within the bathroom area using a room temperature sensor. As an optional implementation, after obtaining the environmental time-series data, it can also be standardized using a Z-score for easier subsequent model analysis and processing.

[0065] Step S102: Input point cloud data, pressure distribution data, thermal radiation distribution data and environmental time series data into a pre-trained deep learning model to obtain exhaust control parameters, including exhaust frequency and exhaust power.

[0066] Specifically, the aforementioned pre-trained deep learning model can be any pre-trained deep learning model, such as a Convolutional Neural Network (CNN), a Diffusion Model, or a custom network model, etc., and this application does not impose any specific limitations.

[0067] Step S103: Based on the exhaust frequency and exhaust power, control the exhaust fan to exhaust air from the bathroom space.

[0068] After predicting the exhaust frequency and exhaust power, the exhaust fan can be controlled to exhaust air from the bathroom space based on the exhaust frequency and exhaust power, so that the humidity level in the bathroom space meets the actual needs.

[0069] In this embodiment, point cloud data, pressure distribution data, thermal radiation distribution data, and environmental time series data in the bathroom space can be analyzed based on a pre-trained deep learning model to accurately determine the exhaust control parameters, thereby achieving precise exhaust control and balancing energy consumption and defogging efficiency.

[0070] In an optional embodiment, step S102 above, inputting point cloud data, pressure distribution data, thermal radiation distribution data, and environmental time series data into a pre-trained deep learning model to obtain exhaust control parameters, includes:

[0071] Point cloud data and thermal radiation distribution data are input into the first feature extraction branch for feature extraction to obtain spatial features;

[0072] The pressure distribution data and environmental time-series data are input into the second feature extraction branch for feature extraction to obtain time-series-pressure joint features.

[0073] Spatial features and temporal-stress joint features are input into a feature fusion layer for feature fusion to obtain fused features;

[0074] The fused features are input into the detection head for detection to obtain the exhaust control parameters.

[0075] Specifically, such as Figure 2 As shown, the deep learning model may include a first feature extraction branch, a second feature extraction branch, a feature fusion layer, and a detection head. The first feature extraction branch is used to extract features from point cloud data and thermal radiation distribution data; the second feature extraction branch is used to extract features from pressure distribution data and environmental time-series data; the feature fusion layer is used to fuse the features extracted by the first and second feature extraction branches; and the detection head is used to detect the fused features obtained by the feature fusion layer.

[0076] When using a deep learning model to predict exhaust control parameters, point cloud data and thermal radiation distribution data can be input into the first feature extraction branch for feature extraction to obtain spatial features. Pressure distribution data and environmental time series data can be input into the second feature extraction branch for feature extraction to obtain time series-pressure joint features. Then, the spatial features and time series-pressure joint features are input into the feature fusion layer for feature fusion to obtain fused features. Finally, the fused features are input into the detection head for detection to obtain exhaust control parameters.

[0077] In this way, exhaust control parameters can be accurately determined based on pre-trained deep learning models, thereby achieving precise exhaust control and balancing energy consumption and defogging efficiency.

[0078] In an optional embodiment, the above steps, including inputting point cloud data and thermal radiation distribution data into the first feature extraction branch for feature extraction to obtain spatial features, include:

[0079] Point cloud data and thermal radiation distribution data are input into the input layer and spliced ​​together to obtain the first intermediate data;

[0080] The first intermediate data is input into the farthest sampling layer for downsampling to obtain the second intermediate data;

[0081] The second intermediate data is input into the multilayer perceptron layer to extract local features point by point and increase the dimensionality to obtain the third intermediate data.

[0082] The third intermediate data is input into the max pooling layer for local feature aggregation to obtain the fourth intermediate data;

[0083] The fourth intermediate data is input into the first fully connected layer for feature mapping to obtain spatial features.

[0084] Specifically, such as Figure 3 As shown, the first feature extraction branch may include an input layer, a farthest point sampling layer, a multilayer perceptron layer, a max pooling layer, and a first fully connected layer. The input layer concatenates the point cloud data and thermal radiation distribution data, serving as the input for subsequent layers. The first intermediate data output by this input layer can be N×D dimensional data, where N represents the number of points (e.g., 1024 points), and D represents the initial feature dimension of each point (usually 3D coordinates (x, y, z), possibly including other features such as velocity, normal vectors, etc.). The farthest point sampling (FPS) layer downsamples the input data, resulting in more uniform coverage of the original shape. The second intermediate data output by this farthest point sampling layer can be M×D dimensional data, where M < N. The Multilayer Perceptron (MLP) layer is used to extract local features point by point and increase their dimensionality. For example, MLP(64) can map the features of each point from D dimensions to 64 dimensions, outputting M×64, and MLP(128) can further map to 128 dimensions, outputting M×128 (i.e., the third intermediate data). The Max Pooling layer is used to aggregate the local features of all points to generate a global feature vector. The fourth intermediate data output by the Max Pooling layer can be 1×128 dimensional data. The first fully connected layer is used to perform feature mapping, mapping to high-dimensional spatial features, further fusing global information, enhancing the discriminative ability of features, and providing richer feature representations for downstream tasks (such as classification and segmentation). The spatial features output by the first fully connected layer can be 1×256 dimensional data.

[0085] In this way, point cloud data and thermal radiation distribution data can be sequentially input into the input layer, the farthest point sampling layer, the multilayer perceptron layer, the max pooling layer, and the first fully connected layer for feature extraction, ultimately obtaining the spatial features related to the bathroom and the human body for subsequent use.

[0086] In an optional embodiment, the above steps, including inputting pressure distribution data and environmental time-series data into the second feature extraction branch for feature extraction to obtain time-series-pressure joint features, include:

[0087] Environmental time-series data is input into the hidden layer for embedding representation to obtain time-series features;

[0088] The pressure distribution data is input into the second fully connected layer for feature mapping to obtain pressure features;

[0089] The temporal and stress features are input into the splicing layer for splicing to obtain the temporal-stress joint features.

[0090] Specifically, see Figure 4 The second feature extraction branch includes a hidden layer, a second fully connected layer, and a splicing layer. The hidden layer can be a Long Short-Term Memory (LSTM) layer, used to process water temperature time-series data, humidity time-series data, and room temperature time-series data. Its processing can be represented by the following formula:

[0091] h t =LSTM(x t h t-1 );

[0092] Where, x t This represents standardized time-series data (such as water temperature time-series data, humidity time-series data, and room temperature time-series data), h t h represents the environmental state at time t. t-1 This represents the environmental state at time t-1.

[0093] Assuming that the water temperature time series data, humidity time series data, and room temperature time series data are selected using a 60-second sliding window, then h t For a 60×3 matrix, the temporal attention weights and temporal features can be represented by the following formula:

[0094] α t =Softmax(W a ·h t );

[0095]

[0096] Where, α t h represents the attention weight at time t. t W represents the environmental state at time t. a W represents the preset weighting coefficient. a ·h t f represents the global query vector. timeThis represents temporal characteristics. The second fully connected layer can represent the trajectory standard deviation σ. pressure and mobile frequency f move After concatenating them into a 2D vector, they are mapped to a 32-dimensional pressure feature f. pressure This splicing layer is used to combine temporal features f time and pressure characteristics f pressure By concatenating the data, we obtain the time-stress joint feature f. joint =[f time ;f pressure (160 dimensions).

[0097] After obtaining the time-pressure joint features f joint Then, the spatial features f can be... space Combined time-pressure characteristics f joint The input is fed into the feature fusion layer for feature fusion to obtain the fused feature, which can be represented by the following formula:

[0098] f fusion =ReLU(W f [f space ;f joint ]+b f );

[0099] Among them, f fusion f represents the fusion feature. space Representing spatial characteristics, f joint Indicates the time-stress joint feature, b f W represents the bias term of each network layer. f This represents the preset weight coefficients. Then, the fused features f are... fusion The data is input into the detection head for detection to obtain the exhaust control parameters, including the exhaust frequency f. fan And exhaust power P power It can be expressed using the following formula:

[0100] f fan =Sigmoid(W1*f fusion );

[0101] P power =W2*f fusion ;

[0102] Among them, f fan P represents the exhaust fan speed ratio, i.e., the exhaust frequency. power This indicates the exhaust power, W1 and W2 represent the preset weighting coefficients, and f fusion This indicates the fusion feature.

[0103] In this way, pressure distribution data and environmental time-series data can be sequentially input into the hidden layer, the second fully connected layer, and the splicing layer for feature extraction, ultimately obtaining time-series-pressure joint features for subsequent use.

[0104] In an optional embodiment, after step S102 above, where point cloud data, pressure distribution data, thermal radiation distribution data, and environmental time-series data are input into a pre-trained deep learning model to obtain exhaust control parameters, the method further includes:

[0105] Desensitize the fusion features to obtain desensitized features;

[0106] The desensitization features, environmental time-series data, and ventilation control parameters are sent to the cloud platform, so that the cloud platform can store the desensitization features, environmental time-series data, and ventilation control parameters in a preset database, and use the desensitization features, environmental time-series data, and ventilation control parameters to perform federated learning to update its own deep learning model.

[0107] Specifically, data anonymization methods can include data encryption, adding Gaussian noise ∈ ~N(0,σ2), and other methods. The aforementioned preset database can be a database on a cloud platform or another database connected to the cloud platform. As an optional implementation, historical user water temperature settings and corresponding exhaust parameters can be recorded to construct key-value pairs Key:(T water ,T room (season), Value:(f fan ,p power The data is stored in a pre-defined database. Federated learning refers to the cloud platform collecting anonymized data (de-identified features + control parameters) sent by each home host computer and updating its own deep learning model. Specifically, the FedAvg algorithm can be used to aggregate model parameters.

[0108]

[0109] Where, θ global This represents the global model parameters, where N represents the number of clients participating in the aggregation. k n represents the local data sample size of the k-th client. total θ represents the total number of data samples from all clients. k This represents the local model parameters for the k-th client.

[0110] After obtaining the exhaust control parameters, the fused features can be desensitized to obtain desensitized features. Then, the desensitized features, environmental time series data, and exhaust control parameters are sent to the cloud platform. The cloud platform can then store the desensitized features, environmental time series data, and exhaust control parameters in a preset database, and use the desensitized features, environmental time series data, and exhaust control parameters to perform federated learning to update its own deep learning model.

[0111] In this way, the cloud platform can continuously optimize its own model using federated learning, and each home host can connect to the network to download the latest model from the cloud platform for use.

[0112] In an optional embodiment, after the above steps of sending the desensitization features, environmental time-series data, and exhaust control parameters to the cloud platform, the method further includes:

[0113] When the ventilation control in the bathroom needs to be re-controlled, the de-identified features of the current user and the current environmental time series data in the bathroom are obtained, and the de-identified features of the current user and the current environmental time series data are sent to the cloud platform. The cloud platform then matches the de-identified features of the current user with the de-identified features stored in the preset database. If the de-identified features of the current user and the de-identified features stored in the preset database are successfully matched, the ventilation control parameters that best match the current environmental time series data are determined and returned.

[0114] Receive the exhaust control parameters returned by the cloud platform and perform exhaust control according to the exhaust control parameters returned by the cloud platform.

[0115] Specifically, after sending the anonymized features, environmental time-series data, and ventilation control parameters to the cloud platform, if the bathroom space needs to be re-controlled for ventilation, the system can obtain the current user's anonymized features and current environmental time-series data, and send these to the cloud platform. The cloud platform can then match the current user's anonymized features with those stored in a pre-defined database. If a match is found, the system calculates similarity and determines the ventilation control parameter that best matches the current environmental time-series data from multiple ventilation control parameters corresponding to different environments for that user. This parameter is then returned to the home control unit. The home control unit can then receive the ventilation control parameters returned by the cloud platform and control the ventilation accordingly.

[0116] This avoids the need for the same user to repeatedly use a deep learning model to predict ventilation control parameters while showering in the bathroom, thus preventing unnecessary performance waste.

[0117] In an optional embodiment, after step S103 above, where the exhaust fan is controlled to exhaust air from the bathroom space based on the exhaust frequency and exhaust power, the method further includes:

[0118] The humidity value in the bathroom is acquired in real time, and the difference between the humidity value in the bathroom and the preset humidity value is calculated to obtain the humidity deviation.

[0119] Based on humidity deviation, the exhaust frequency and exhaust power are dynamically corrected.

[0120] Specifically, when controlling the exhaust fan to ventilate the bathroom space, the humidity level in the bathroom space can be acquired in real time, and the difference between the current humidity level and the preset humidity level can be calculated to obtain the humidity deviation. Then, based on the humidity deviation, the exhaust frequency and exhaust power are dynamically adjusted. The adjustment process for the exhaust frequency can be expressed by the following formula:

[0121]

[0122] in, This represents the exhaust frequency predicted by the model, and ΔH represents the humidity deviation. H represents the current humidity level in the bathroom. target Indicates the preset humidity value, K p K i and K d The weights can be optimized online using reinforcement learning. The process for correcting exhaust power is similar to that for correcting exhaust frequency, and will not be described in detail here.

[0123] In this way, the exhaust frequency and exhaust power can be dynamically adjusted based on the measured humidity deviation to suppress fog fluctuations.

[0124] The bathroom ventilation control method provided in this embodiment has the following beneficial effects:

[0125] 1. Achieve precise airflow control: Predict the optimal exhaust strategy based on user characteristics and environmental conditions, improve fog removal efficiency, and enhance energy-saving effect by dynamically adjusting exhaust power compared to the fixed mode.

[0126] 2. User experience optimization: Reduced manual operation, adapted to different groups such as the elderly with cold protection needs and children with scald protection needs, making the user experience more comfortable.

[0127] 3. Completely avoid the risk of privacy leakage: It can automatically upload anonymized data to the cloud platform, realizing personalized, high-efficiency and intelligent bathroom ventilation while solving privacy issues.

[0128] 4. Personalized ventilation control: Predicts fog generation trends based on user identity and water temperature preferences to achieve personalized ventilation control.

[0129] See Figure 5 , Figure 5 This is a schematic diagram of a bathroom ventilation control device provided in an embodiment of this application. Figure 5 As shown, the bathroom exhaust control device 500 includes:

[0130] The first acquisition module 501 is used to acquire multimodal data, including point cloud data, pressure distribution data, thermal radiation distribution data and environmental time series data within the bathroom space.

[0131] The input module 502 is used to input point cloud data, pressure distribution data, thermal radiation distribution data and environmental time series data into a pre-trained deep learning model to obtain exhaust control parameters, including exhaust frequency and exhaust power.

[0132] The control module 503 is used to control the exhaust fan to exhaust air from the bathroom space based on the exhaust frequency and exhaust power.

[0133] The deep learning model includes a first feature extraction branch, a second feature extraction branch, a feature fusion layer, and a detection head;

[0134] Furthermore, the input module 502 includes:

[0135] The first input submodule is used to input point cloud data and thermal radiation distribution data into the first feature extraction branch for feature extraction to obtain spatial features;

[0136] The second input submodule is used to input pressure distribution data and environmental time series data into the second feature extraction branch for feature extraction to obtain time series-pressure joint features.

[0137] The third input submodule is used to input spatial features and temporal-stress joint features into the feature fusion layer for feature fusion to obtain fused features;

[0138] The fourth input submodule is used to input the fused features into the detection head for detection to obtain the exhaust control parameters.

[0139] Furthermore, the first feature extraction branch includes an input layer, a farthest point sampling layer, a multilayer perceptron layer, a max pooling layer, and a first fully connected layer; the first input submodule includes:

[0140] The first input unit is used to input point cloud data and thermal radiation distribution data into the input layer for splicing to obtain the first intermediate data;

[0141] The second input unit is used to input the first intermediate data into the farthest point sampling layer for downsampling to obtain the second intermediate data;

[0142] The third input unit is used to input the second intermediate data into the multilayer perceptron layer to extract local features point by point and increase the dimensionality to obtain the third intermediate data.

[0143] The fourth input unit is used to input the third intermediate data into the max pooling layer for local feature aggregation to obtain the fourth intermediate data;

[0144] The fifth input unit is used to input the fourth intermediate data into the first fully connected layer for feature mapping to obtain spatial features.

[0145] Furthermore, the second feature extraction branch includes a hidden layer, a second fully connected layer, and a concatenation layer; the second input submodule includes:

[0146] The sixth input unit is used to input environmental time series data into the hidden layer for embedding representation to obtain time series features;

[0147] The seventh input unit is used to input pressure distribution data into the second fully connected layer for feature mapping to obtain pressure features;

[0148] The eighth input unit is used to input the temporal features and stress features into the splicing layer for splicing to obtain the temporal-stress joint features.

[0149] Furthermore, the bathroom exhaust control device 500 also includes:

[0150] The desensitization module is used to desensitize the fused features to obtain desensitized features;

[0151] The sending module is used to send desensitized features, environmental time-series data, and ventilation control parameters to the cloud platform, so that the cloud platform can store the desensitized features, environmental time-series data, and ventilation control parameters in a preset database, and use the desensitized features, environmental time-series data, and ventilation control parameters to perform federated learning to update its own deep learning model.

[0152] Furthermore, the bathroom exhaust control device 500 also includes:

[0153] The second acquisition module is used to acquire the desensitized features of the current user and the current environmental time series data in the bathroom space when the exhaust control of the bathroom space needs to be re-controlled. The module then sends the desensitized features of the current user and the current environmental time series data to the cloud platform so that the cloud platform can match the desensitized features of the current user with the desensitized features stored in the preset database. If the desensitized features of the current user match the desensitized features stored in the preset database, the cloud platform determines and returns the exhaust control parameters that best match the current environmental time series data.

[0154] The receiving module is used to receive the exhaust control parameters returned by the cloud platform and to perform exhaust control according to the exhaust control parameters returned by the cloud platform.

[0155] Furthermore, the bathroom exhaust control device 500 also includes:

[0156] The third acquisition module is used to acquire the humidity value in the bathroom space in real time, and calculate the difference between the humidity value in the bathroom space and the preset humidity value to obtain the humidity deviation;

[0157] The correction module is used to dynamically correct the exhaust frequency and exhaust power based on humidity deviation.

[0158] It should be noted that the bathroom exhaust control device 500 can implement the bathroom exhaust control method provided in any of the aforementioned method embodiments and achieve the same technical effect, which will not be elaborated here.

[0159] See Figure 6 , Figure 6 This is a schematic diagram of the structure of a home host provided in an embodiment of this application, such as... Figure 6 As shown, the home console includes a processor 611, a communication interface 612, a memory 613, and a communication bus 614. The processor 611, communication interface 612, and memory 613 communicate with each other via the communication bus 614.

[0160] Memory 613 is used to store computer programs;

[0161] In one embodiment of this application, the processor 611, when executing the program stored in the memory 613, implements the bathroom ventilation control method provided in any of the foregoing method embodiments.

[0162] See Figure 7 , Figure 7 This is a schematic diagram of a bathroom ventilation control system provided in an embodiment of this application. Figure 7 As shown, the bathroom ventilation control system 700 includes a cloud platform 710 and a home host 720 in the aforementioned embodiment; wherein, the cloud platform 710 and the home host 720 are communicatively connected.

[0163] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the bathroom ventilation control method as provided in any of the foregoing method embodiments.

[0164] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0165] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0166] It should be understood that the terminology used herein is for the purpose of describing particular exemplary embodiments only and is not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “described” as used herein may also include the plural forms. The terms “comprising,” “including,” “containing,” and “having” are inclusive and therefore indicate the presence of the stated features, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, elements, components, and / or combinations thereof. The method steps, processes, and operations described herein are not construed as requiring them to be performed in a particular order described or illustrated unless the order of performance is explicitly indicated. It should also be understood that additional or alternative steps may be used.

[0167] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A bathroom exhaust ventilation control method, characterized in that, The method includes: Acquire multimodal data, wherein the multimodal data includes point cloud data, pressure distribution data, thermal radiation distribution data, and environmental time series data within the bathroom space; The point cloud data, the pressure distribution data, the thermal radiation distribution data, and the environmental time series data are input into a pre-trained deep learning model to obtain exhaust control parameters, wherein the exhaust control parameters include exhaust frequency and exhaust power. Based on the exhaust frequency and the exhaust power, the exhaust fan is controlled to exhaust air from the bathroom space.

2. The method according to claim 1, characterized in that, The deep learning model includes a first feature extraction branch, a second feature extraction branch, a feature fusion layer, and a detection head; The step of inputting the point cloud data, the pressure distribution data, the thermal radiation distribution data, and the environmental time series data into a pre-trained deep learning model to obtain exhaust control parameters includes: The point cloud data and the thermal radiation distribution data are input into the first feature extraction branch for feature extraction to obtain spatial features; The pressure distribution data and the environmental time-series data are input into the second feature extraction branch for feature extraction to obtain time-series-pressure joint features. The spatial features and the temporal-stress joint features are input into the feature fusion layer for feature fusion to obtain fused features; The fused features are input into the detection head for detection to obtain the exhaust control parameters.

3. The method according to claim 2, characterized in that, The first feature extraction branch includes an input layer, a farthest point sampling layer, a multilayer perceptron layer, a max pooling layer, and a first fully connected layer; The step of inputting the point cloud data and the thermal radiation distribution data into the first feature extraction branch for feature extraction to obtain spatial features includes: The point cloud data and the thermal radiation distribution data are input into the input layer and spliced ​​together to obtain the first intermediate data; The first intermediate data is input into the farthest point sampling layer for downsampling to obtain the second intermediate data; The second intermediate data is input into the multilayer perceptron layer to extract local features point by point and increase the dimensionality to obtain the third intermediate data; The third intermediate data is input into the max pooling layer for local feature aggregation to obtain the fourth intermediate data; The fourth intermediate data is input into the first fully connected layer for feature mapping to obtain the spatial features.

4. The method according to claim 2, characterized in that, The second feature extraction branch includes a hidden layer, a second fully connected layer, and a splicing layer; The step of inputting the pressure distribution data and the environmental time-series data into the second feature extraction branch for feature extraction to obtain time-series-pressure joint features includes: The environmental time-series data is input into the hidden layer for embedding representation to obtain time-series features; The pressure distribution data is input into the second fully connected layer for feature mapping to obtain pressure features; The time-series features and the pressure features are input into the splicing layer for splicing to obtain the time-series-pressure joint features.

5. The method according to claim 2, characterized in that, After inputting the point cloud data, the pressure distribution data, the thermal radiation distribution data, and the environmental time-series data into a pre-trained deep learning model to obtain exhaust control parameters, the method further includes: The fusion features are desensitized to obtain desensitized features; The desensitization features, the environmental time-series data, and the exhaust control parameters are sent to the cloud platform, so that the cloud platform stores the desensitization features, the environmental time-series data, and the exhaust control parameters in a preset database, and uses the desensitization features, the environmental time-series data, and the exhaust control parameters to perform federated learning to update its own deep learning model.

6. The method according to claim 5, characterized in that, After sending the desensitization features, the environmental time-series data, and the exhaust control parameters to the cloud platform, the method further includes: When the bathroom space needs to be re-controlled for ventilation, the de-identified features of the current user and the current environmental time series data in the bathroom space are obtained, and the de-identified features of the current user and the current environmental time series data are sent to the cloud platform. The cloud platform then matches the de-identified features of the current user with the de-identified features stored in the preset database. If the de-identified features of the current user and the de-identified features stored in the preset database are successfully matched, the ventilation control parameters that best match the current environmental time series data are determined and returned. Receive the exhaust control parameters returned by the cloud platform, and perform exhaust control according to the exhaust control parameters returned by the cloud platform.

7. The method according to claim 1, characterized in that, After controlling the exhaust fan to ventilate the bathroom space based on the exhaust frequency and the exhaust power, the method further includes: The humidity value in the bathroom space is acquired in real time, and the difference between the humidity value in the bathroom space and the preset humidity value is calculated to obtain the humidity deviation; Based on the humidity deviation, the exhaust frequency and the exhaust power are dynamically corrected.

8. A bathroom exhaust ventilation control device, characterized in that, The device includes: The first acquisition module is used to acquire multimodal data, wherein the multimodal data includes point cloud data, pressure distribution data, thermal radiation distribution data and environmental time series data within the bathroom space; The input module is used to input the point cloud data, the pressure distribution data, the thermal radiation distribution data and the environmental time series data into a pre-trained deep learning model to obtain exhaust control parameters, wherein the exhaust control parameters include exhaust frequency and exhaust power. The control module is used to control the exhaust fan to exhaust air from the bathroom space based on the exhaust frequency and the exhaust power.

9. A home console, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, when executing a program stored in a memory, implements the bathroom ventilation control method according to any one of claims 1-7.

10. A bathroom exhaust ventilation control system, characterized in that, The bathroom ventilation control system includes a cloud platform and the home host as described in claim 9; The cloud platform is communicatively connected to the home host.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the bathroom ventilation control method according to any one of claims 1-7.

Images