Automatic control method and control system of intelligent sofa

Abstract

The invention relates to the technical field of smart home, in particular to an automatic control method and control system of a smart sofa. Comprising the following steps: firstly, identifying the identity of a user of the intelligent sofa to obtain corresponding personal data; secondly, a posture recognition model is adopted to recognize the posture of the user, the model is combined with a pressure recognition technology and an image recognition technology for comprehensive recognition, and the posture recognition accuracy is improved; and finally, automatically controlling the intelligent sofa, including temperature control and posture control. According to the temperature control, a temperature influence curve is established according to the historical sofa regulation and control temperature, the environment temperature and the body temperature of the user for analysis, so that the intelligent sofa can realize accurate temperature adjustment according to the body temperature of the user and the environment temperature; the posture control is analyzed according to the current state of the user, precise automatic control is achieved in combination with personal data of the user in different states, and the experience feeling of the user is improved.

Description

Technical Field

[0001] This invention relates to the field of smart home technology, specifically to an automatic control method and control system for a smart sofa. Background Technology

[0002] Sofas, as a common piece of furniture, provide a comfortable space for sitting or lying down, offering both relaxation and social interaction. They are typically placed in the living room or family room, which are their most common settings. However, they can also be placed in offices, lounges, and studies.

[0003] For sofas, the most important thing is to provide users with a comfortable and relaxing experience. This includes good support, softness, a suitable sitting angle, and overall comfort, allowing users to easily relax both physically and mentally. Traditional design processes did not explore automatic adjustment of sofas; instead, users achieved a good experience by changing their posture or making manual adjustments. With the continuous development of technology, researchers have combined automation and recognition technologies to produce a wide variety of fully functional smart sofas.

[0004] Both smart sofas and traditional sofas provide users with basic rest and relaxation, but smart sofas, through technological means, achieve automated control, greatly enhancing user comfort and convenience. Currently, smart sofas on the market include many functions, such as intelligent adjustment, massage, temperature control, and health monitoring. While smart sofas are quite comprehensive, some shortcomings remain; for example, the automatic control process follows a standardized pattern, making it impossible to adjust control parameters according to individual user needs, resulting in a less than ideal user experience.

[0005] Therefore, this invention proposes an automatic control method and control system for an intelligent sofa. Summary of the Invention

[0006] The purpose of this invention is to propose an automatic control method and control system for a smart sofa, achieving automated control of the sofa's temperature and posture. This invention focuses on the following aspects: First, it uses facial recognition technology to acquire the user's personal data from a database to accurately determine the user's height, weight, and health status. Second, it proposes a posture recognition model to identify the user's posture. This model includes a pressure recognition module, an image recognition module, and a joint output module. The pressure recognition module acquires corresponding pressure values ​​through sensors placed at different locations on the smart sofa, and analyzes these pressure values ​​to determine the distribution of the user's body. The image recognition module divides the human body into multiple points based on the acquired user image and performs posture recognition based on the features of these points and their connections to adjacent points. The joint output module integrates the recognition results from the pressure recognition module and the image recognition module. Finally, this invention automatically controls the temperature and posture of the smart sofa. Regarding temperature control, this invention utilizes historical temperature settings, historical ambient temperature, and the user's historical temperature data. A temperature influence curve is established based on body temperature. This curve allows for the adjustment of the smart sofa's temperature according to real-time monitoring of human body temperature and ambient temperature, enabling both heating and cooling functions. Regarding posture control, the invention divides the control scenario into a first state and a second state. The first state refers to activities such as using a mobile phone, reading, or watching television. During this process, the relative positions of the smart sofa, the user, and the target object are obtained through spatial coordinates. The smart sofa is initially adjusted based on the calculated distances between them. A second adjustment is then made based on the user's personal data and current posture information. The second state refers to the user being asleep. During the adjustment process, data mapping is performed by collecting past user sleep postures and the smart sofa's control parameters. The mapped data is then clustered and input into a machine learning model to obtain the correspondence between sleep posture and control parameters. The similarity between the current body posture and the historical control set is calculated to obtain the control parameters in the corresponding set. Finally, the user's personal data is combined to control the smart sofa's parameters.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] Obtain the user image; perform facial recognition based on the user image to obtain the user ID;

[0009] Furthermore, the user's personal data is obtained based on the user ID; the personal data includes: basic personal information and physical condition information;

[0010] Furthermore, the user's current body posture is identified; wherein, the current body posture is determined using a posture recognition model; the specific identification process includes:

[0011] The pressure values ​​from pressure sensors located at different positions on the smart sofa are obtained and input to the pressure recognition module.

[0012] Furthermore, the pressure values ​​are denoised to obtain standard pressure data;

[0013] Furthermore, the standard pressure data are fused to construct a pressure distribution map;

[0014] Furthermore, feature extraction is performed on the pressure distribution map to obtain a pressure feature set; wherein, the pressure feature set includes: location features and pressure features; and the location features and the pressure features form a corresponding relationship;

[0015] Furthermore, the set of stress features is input into a pre-trained classification model to obtain stress posture recognition results;

[0016] The user image is obtained and sent to the preprocessing layer of the image recognition module to obtain a standard user image;

[0017] Furthermore, the standard user image is input into the feature extraction layer to obtain human skeleton features and pose features, respectively;

[0018] Further, the human skeletal features and the pose features are input into a feature fusion layer to obtain pose recognition features; wherein, the fusion process of the feature fusion layer includes: obtaining the connection relationship of each skeletal point in the skeletal features; when the connection relationship is a direct connection, calculating the connection slope of the skeletal point from the pose features; wherein, the formula for calculating the connection slope is: Among them, M(P i ,P j ) represents the skeleton point P. i With skeleton point P j The connection slope; (x) i ,y i ) represents the skeleton point P. i The coordinates; (x j ,y j ) represents the skeleton point P. j The coordinates;

[0019] Furthermore, the human skeletal features are spliced ​​with the corresponding posture features to obtain spliced ​​features;

[0020] Furthermore, the splicing features are weighted using the connection slope to obtain the pose recognition features;

[0021] Furthermore, the key features are input into the classification layer to obtain the image pose recognition result;

[0022] Further, the stress posture recognition result and the image posture recognition result are input into the joint output module to obtain the user's current body posture. The specific implementation process of the joint output module includes: obtaining the corresponding stress body posture and image body posture based on the stress posture recognition result and the image posture recognition result; acquiring the features of the stress body posture and the image body posture respectively to obtain a first posture feature and a second posture feature; calculating the error between the first posture feature and the second posture feature to construct a posture error feature vector; correcting the posture error feature vector based on posture feature values ​​in the body posture library to obtain a unified posture feature; and passing the unified posture feature through the output layer to obtain the current body posture.

[0023] Furthermore, the current state of the user is determined; the current state includes: a first state and a second state;

[0024] Furthermore, the smart sofa is automatically controlled according to the current state; wherein, the automatic control includes: automatic temperature adjustment and automatic posture adjustment;

[0025] The automatic temperature adjustment process includes:

[0026] Obtain the user's historical temperature adjustment parameters, historical ambient temperature, and historical body temperature;

[0027] Furthermore, the historical temperature adjustment parameters, the historical ambient temperature, and the user's historical body temperature are constructed into a historical temperature matrix; wherein, the historical temperature matrix is ​​represented as: Y = θX + ε; where Y represents the vector of the user's historical body temperature; X represents the matrix of the historical temperature adjustment parameters and the historical ambient temperature; θ represents the adjustment parameters calculated using the least squares method; and ε represents the error term.

[0028] Furthermore, based on the adjustment parameters and the error term, a curve showing the effect of body temperature is obtained, expressed as: T b,i =θ(αT) s,i +βT e,i )+ε; where, T b,i T represents the user's body temperature recorded in the i-th entry; s,i This represents the sofa temperature setting recorded in the i-th record; T e Let represent the ambient temperature of the i-th record; α and β represent the parameters of the body temperature influence curve.

[0029] Furthermore, the smart sofa's temperature is automatically adjusted based on the monitored user's body temperature and ambient temperature using the body temperature influence curve; wherein, the smart sofa's automatic posture adjustment process includes:

[0030] Obtain the current state of the user;

[0031] When the user is in the first state; the control process of the smart sofa includes: acquiring the first spatial relative position information of the smart sofa, the second spatial relative position information of the user, the third spatial relative position information of the first object, and the size information of the first object;

[0032] Furthermore, based on the user's current posture recognition result, the user's posture spatial information is obtained;

[0033] Furthermore, an adjustment set is obtained based on the user's physical condition information and posture spatial information;

[0034] Further, based on the first spatial relative position information, the second spatial information, and the third spatial relative position information, the first relative distance between the smart sofa and the user, and the second relative distance between the smart sofa and the first object are calculated respectively;

[0035] Furthermore, the smart sofa is initially adjusted based on the first relative distance, the second relative distance, and the size information;

[0036] Furthermore, the initial adjustment is automatically adjusted a second time based on the adjustment set and the personal basic information; wherein the second automatic adjustment includes any one of the following: the height and angle of the backrest, the position of the armrests, and the hardness and curvature of the seat cushion.

[0037] When the user is in the second state; the control process of the smart sofa includes: acquiring the user's historical sleep state; wherein, the historical sleep state includes: historical sleep posture and historical control parameters of the smart sofa;

[0038] Furthermore, a sleep data mapping is formed by mapping each historical sleep posture to the corresponding historical control parameters;

[0039] Furthermore, the sleep data mappings are clustered to obtain N similar mapping sets;

[0040] Furthermore, the N mapping sets are input into a machine learning model to obtain the correspondence between N sleep posture control parameters.

[0041] Furthermore, the current body posture is compared with the similar mapping set to obtain similar sleep posture features;

[0042] Furthermore, the control parameters of the smart sofa are obtained by analyzing the corresponding relationship between the sleep posture control parameters.

[0043] Furthermore, the parameters are adjusted based on the control parameters and the individual data to obtain the optimal control parameters;

[0044] Furthermore, the smart sofa automatically adjusts any one of the backrest, armrests, footrests, and seat cushions according to the optimal control parameters.

[0045] An automatic control system for a smart sofa includes: an image acquisition unit, a sensor unit, an information storage unit, an identity recognition unit, a posture recognition unit, a status judgment unit, an automatic control unit, a voice control unit, and a parameter display unit;

[0046] The image acquisition unit is used to acquire the user's image information;

[0047] The sensor unit is used to collect pressure values ​​at various locations of the smart sofa;

[0048] The information storage unit is used to store user information and the historical control information of the smart sofa;

[0049] The identity recognition unit is used to identify the user;

[0050] The posture recognition unit is used to recognize the user's body posture; wherein, the posture recognition unit uses a posture recognition model to perform posture recognition; wherein, the posture recognition model includes: a pressure recognition module, an image recognition module, and a joint output module;

[0051] The pressure recognition module's recognition process includes: acquiring the pressure value of the smart sofa through the sensor unit; denoising the pressure value to obtain standard pressure data; fusing the standard pressure data to construct a pressure distribution map; extracting features from the pressure distribution map to obtain a pressure feature set; wherein the pressure feature set includes: location features and pressure features; and the location features and pressure features form a corresponding relationship; inputting the pressure feature set into a pre-trained classification model to obtain the pressure posture recognition result.

[0052] The recognition process of the image recognition module includes: acquiring a user image through the image acquisition unit; inputting the user image into an image preprocessing layer to obtain a standard user image; inputting the standard user image into a feature extraction layer to obtain human skeleton features and pose features respectively; inputting the human skeleton features and pose features into a feature fusion layer to obtain pose recognition features; wherein, the fusion process of the feature fusion layer includes: obtaining the connection relationship of each skeleton point in the skeleton features; when the connection relationship is a direct connection, calculating the connection slope of the skeleton point from the pose features; wherein, the formula for calculating the connection slope is: Among them, M(P i ,P j ) represents the skeleton point P. i With skeleton point P j The connection slope; (x) i ,y i ) represents the skeleton point P. i The coordinates; (x j ,y j ) represents the skeleton point P. j The coordinates are determined; the human skeleton features are concatenated with the corresponding pose features to obtain concatenated features; the concatenated features are weighted using the connection slope to obtain the pose recognition features; an attention mechanism is used to focus on the pose recognition features of key skeleton points and obtain the key features of key skeleton points; the key features are input into the classification layer to obtain the image pose recognition result.

[0053] The pressure posture recognition result and the image posture recognition result are input into the joint output module to obtain the user's current body posture.

[0054] The status determination unit is used to determine the user's current status;

[0055] The automatic control unit is used to achieve automated control of the smart sofa, and includes: an automatic temperature control module and an automatic posture control module;

[0056] The specific implementation of the aforementioned automatic temperature control module includes: acquiring the user's historical temperature adjustment parameters, historical ambient temperature, and historical body temperature; constructing a historical temperature matrix from the historical temperature adjustment parameters, historical ambient temperature, and historical body temperature; wherein the historical temperature matrix is ​​represented as: Y = θX + ε; where Y represents the vector of the user's historical body temperature; X represents the matrix of the historical temperature adjustment parameters and the historical ambient temperature; θ represents the adjustment parameters calculated using the least squares method; ε represents the error term; and a body temperature influence curve is obtained based on the adjustment parameters and the error term, represented as: T b,i=θ(αT) s,i +βT e,i )+ε; where, T b,i T represents the user's body temperature recorded in the i-th entry; s,i This represents the sofa temperature setting recorded in the i-th record; T e The ambient temperature is represented by the i-th record; α and β are the parameters of the body temperature influence curve; the temperature of the smart sofa is automatically controlled based on the monitored body temperature and ambient temperature using the body temperature influence curve.

[0057] The posture control module includes a first state control function and a second state control function. The first state control function includes: acquiring first spatial relative position information of the smart sofa, second spatial relative position information of the user, third spatial relative position information of a first object, and size information of the first object; acquiring the user's posture space information based on the user's current posture recognition result; obtaining an adjustment set based on the user's body condition information and the posture space information; calculating a first relative distance between the smart sofa and the user, and a second relative distance between the smart sofa and the first object, based on the first, second, and third spatial relative position information; performing initial adjustments to the smart sofa based on the first, second, and size information; and performing secondary automatic adjustments to the initial adjustments based on the adjustment set and the user's basic information. The secondary automatic adjustments include any one of the following: backrest height and angle, armrest position, and seat cushion hardness and curvature.

[0058] The second state control function includes: acquiring the user's historical sleep state; wherein the historical sleep state includes: historical sleep posture and historical control parameters of the smart sofa; forming a sleep data mapping between each historical sleep posture and the corresponding historical control parameter; clustering the sleep data mapping to obtain N similar mapping sets; inputting the N mapping sets into a machine learning model to obtain N sleep posture-control parameter analysis correspondences; calculating the similarity between the current body posture and the similar mapping sets to obtain similar sleep posture features; acquiring the control parameters of the smart sofa based on the corresponding sleep posture-control parameter analysis correspondences; adjusting the parameters based on the control parameters and personal data to obtain optimal control parameters; and automatically adjusting any one of the backrest, armrests, footrests, and seat cushions according to the optimal control parameters.

[0059] The voice control unit is used by the user to control the smart sofa via voice.

[0060] The parameter display unit is used to display the current control price parameters of the smart sofa.

[0061] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0062] 1. This invention proposes a posture recognition model for recognizing the posture of users on a smart sofa. The model includes a pressure recognition module, an image recognition module, and a joint output module. The pressure recognition module obtains pressure posture recognition results by analyzing the pressure distribution map of the smart sofa. The image recognition module obtains image posture recognition results by weighted fusion of skeletal point features and positional information. The joint output module corrects the posture recognition results using both the pressure and image posture recognition results. This model can accurately determine the user's body posture, providing effective data support for subsequent automatic control of the smart sofa and improving the user experience.

[0063] 2. This invention proposes an automatic temperature control method for a smart sofa. This method can establish a temperature influence curve by collecting historical smart sofa temperature adjustment parameters, historical ambient temperature, and historical body temperature of the user. This curve reflects the user's temperature adjustment of the smart sofa under the dual effects of body temperature and ambient temperature. The smart sofa can perform real-time and precise automatic temperature control for specific users through temperature monitoring equipment, thereby improving the user's experience.

[0064] 3. This invention proposes a posture adjustment method for a smart sofa. This method can automatically control the user based on their recognition status. In the first state, an initial adjustment is performed by calculating the relative spatial positions of the smart sofa, the user, and the target object. A second adjustment is then performed by considering the user's basic personal information, physical condition, and target dimensions. In the second state, the method analyzes the user's past sleeping postures and the smart sofa's historical control parameters. It calculates the similarity between the current body posture and the historical data set to obtain initial adjustment parameters. These parameters are then adjusted based on the user's personal data to obtain the optimal adjustment parameters. This method achieves two dynamic adaptive controls in different scenarios, which helps improve the user experience. Attached Figure Description

[0065] Figure 1 A flowchart illustrating an automatic control method for a smart sofa provided in an embodiment of the present invention;

[0066] Figure 2 This is a structural diagram of an automatic control system for a smart sofa provided in an embodiment of the present invention;

[0067] Figure 3This is a structural diagram of the posture recognition model provided in an embodiment of the present invention;

[0068] Figure 4 A schematic diagram of pressure distribution provided for an embodiment of the present invention;

[0069] Figure 5 The provided illustration is a scenario diagram of Embodiment 1 for the present invention.

[0070] In the picture: 1. Smart sofa; 101. Armrest; 102. Backrest; 103. Seat cushion; 104. Pedal; 2. User A; 3. Television. Detailed Implementation

[0071] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0072] Sofas, as a common piece of furniture in daily life, provide people with comfortable rest and social spaces. However, as people's needs continue to rise, traditional sofas can no longer meet them; to satisfy these needs, many sofas today are combined with high technology, resulting in smart sofas.

[0073] Various types of smart sofas have appeared on the market, and sales are steadily increasing each year. Smart sofas achieve automated control through technological means, greatly enhancing user comfort and convenience. Today's smart sofas include many functions such as intelligent adjustment, massage, temperature control, and health monitoring. However, despite their comprehensive functionality, some shortcomings remain. Their automatic control process follows a standardized model, failing to adapt to different users. Therefore, this invention proposes an automatic control method and control system for smart sofas; the following two embodiments will provide a detailed description of the invention.

[0074] Example 1:

[0075] This application embodiment describes the automatic control process of a smart sofa for user A; to specifically illustrate the automatic control process, this application embodiment will combine... Figure 1 and Figure 2 The content will be explained in detail; Figure 1The present invention proposes an automatic control method for a smart sofa, comprising the following steps: S10. Acquiring a user image; S20. Performing facial recognition on the user to obtain a user ID; S30. Obtaining personal data based on the user ID; S40. Identifying the user's current body posture; S50. Determining the user's state; S60. Performing automatic control of the smart sofa. Figure 2 The present invention proposes a structural diagram of an automatic control system for a smart sofa, comprising: an image acquisition unit, a sensor unit, an information storage unit, an identity recognition unit, a posture recognition unit, a status judgment unit, an automatic control unit, a voice control unit, and a parameter display unit. According to... Figure 1 and Figure 2 The content will be described in detail;

[0076] The image of the user is acquired by the image acquisition unit of the system, corresponding to step S10 above; wherein, the image acquisition unit takes pictures of the user of the smart sofa through a camera installed inside the house, and the camera can achieve 360-degree, no blind spot image acquisition;

[0077] Further, the identity recognition unit of the system performs identity recognition on the facial features of the image acquired by the image acquisition unit, corresponding to step S20 above; wherein, the identity recognition unit adopts the FaceNet model, and performs feature comparison by comparing the facial features of the acquired image with the stored facial images; and obtains the user number based on the recognition result; wherein, the user number is a unique identifier for the user, and the two have a one-to-one correspondence.

[0078] Further, based on the user ID, the corresponding personal data is retrieved from the information storage unit of the system, corresponding to step S30 above; the smart sofa includes a memory storage unit for storing the facial information, ID information, and personal data of the relevant user; wherein, the personal data mainly includes: the user's basic personal information, such as gender, age, height, and weight; and physical condition information, such as whether there are problems with the lumbar spine, spine, or neck; see Table 1, which shows the personal data stored in the information storage unit;

[0079] Table 1. Stored data of the information storage unit

[0080] user User ID gender age Height / m Weight / kg Description of physical condition User A PC001 male 38 1.81 80 There is damage to the lumbar spine and vertebrae. User B PC002 female 36 1.72 65 Occasional shoulder discomfort User C PC003 male 13 1.70 55 Less severe scoliosis

[0081] According to the face matching of the identity recognition unit, the current user of the smart sofa is user A; the saved information of user A can be found in Table 1. User A's user number is PC001, gender is male, age is 38 years old, height is 1.81m, weight is 80kg, and the description of physical condition is: damage to the lumbar spine and spine.

[0082] Further, posture recognition is performed on user A, corresponding to step S40 above. Posture recognition refers to the posture of user A on the smart sofa. Generally, the body postures presented on the sofa mainly include the following: sitting posture (upright sitting, slanted sitting), semi-reclining posture (lying down, legs apart), fully reclining posture (supine, side-lying), slanted reclining posture (semi-side-lying, vertically slanted reclining), crossed posture (legs crossed, arms crossed), and resting posture, etc. For posture recognition, the smart sofa's memory storage stores common body postures in life. If a posture occurs during the use of the smart sofa that is not saved in the memory storage, the smart sofa can record the unsaved posture.

[0083] Regarding the use of a pose recognition model for user A's pose recognition, please refer to [link / reference]. Figure 3 The posture recognition model includes: a pressure recognition module, an image recognition module, and a joint output module.

[0084] The pressure values ​​obtained by the sensor unit from pressure sensors arranged at different positions on the smart sofa are input into the pressure recognition module; wherein, sensors are placed in the seat cushion area, backrest area, armrest area and leg support area of ​​the smart sofa.

[0085] Furthermore, the pressure values ​​are denoised to obtain standard pressure data; wherein, the denoising process involves treating pressure values ​​obtained from non-body parts contacting the smart sofa as noise and removing them.

[0086] Furthermore, the standard pressure data are fused to obtain a pressure distribution map; Figure 4 The paper presents the pressure distribution at various points in a planar coordinate system established on the smart sofa. The larger the value, the greater the pressure at that point.

[0087] Furthermore, feature extraction is performed on the pressure distribution map to obtain a pressure feature set; wherein, the pressure feature set includes: location features and pressure features; and the location features and the pressure features form a corresponding relationship;

[0088] Furthermore, the set of stress features is input into a pre-trained classification model to obtain stress posture recognition results;

[0089] Furthermore, the full-body image of user A acquired by the image acquisition unit is input into the image recognition module;

[0090] Furthermore, the image of user A is input into the preprocessing layer to obtain the standard image of user A;

[0091] Furthermore, the standard user A image is input into the feature extraction layer to obtain human skeleton features and pose features. During feature extraction, OpenPose is used to detect the entire body of user A using 25 key points. OpenPose is a real-time multi-person 2D pose estimation system based on convolutional neural networks and supervised learning, written using Caffe as the framework. It can track the human torso and limbs, and can obtain skeletal key point information of human movements in a timely and accurate manner.

[0092] Further, the human skeletal features and the pose features are input into a feature fusion layer to obtain pose recognition features; wherein, the fusion process of the feature fusion layer includes: obtaining the connection relationship of each skeletal point in the skeletal features; when the connection relationship is a direct connection, calculating the connection slope of the skeletal point from the pose features; wherein, the formula for calculating the connection slope is: Among them, M(P i ,P j ) represents the skeleton point P. i With skeleton point P j The connection slope; (x) i ,y i ) represents the skeleton point P. i The coordinates; (x j ,y j ) represents the skeleton point P. j coordinates

[0093] In the calculation process based on slope, a slope of 0 indicates that the skeleton points are aligned on a horizontal line; a slope of positive or negative infinity indicates that the skeleton points are aligned on a vertical line. The magnitude and sign of the slope are also important for pose recognition. Slope magnitude: The larger the absolute value of the slope, the more significant the tilting relationship between the skeleton points. Slope sign: A positive slope indicates that the skeleton points are tilted upwards, and a negative slope indicates that the skeleton points are tilted downwards.

[0094] For example, P i P j These represent the shoulder and elbow of the body, respectively; the shoulder and elbow are connected; from the identification process, we can know that P i The coordinates are (2,3), P j The coordinates are (5,7); the slope can be calculated using the slope formula: The calculated slope results indicate that the arm is tilted upward relative to the horizontal, which can be identified as an upward movement or an arm bending movement.

[0095] Furthermore, the human skeletal features are spliced ​​with the corresponding posture features to obtain spliced ​​features;

[0096] Furthermore, the splicing features are weighted using the connection slope to obtain the pose recognition features;

[0097] Furthermore, the key features are input into the classification layer to obtain the image pose recognition result;

[0098] Further, the stress posture recognition result and the image posture recognition result are input into the joint output module to obtain the user's current body posture. The specific implementation process of the joint output module includes: obtaining the corresponding stress body posture and image body posture based on the stress posture recognition result and the image posture recognition result; acquiring the features of the stress body posture and the image body posture respectively to obtain a first posture feature and a second posture feature; calculating the error between the first posture feature and the second posture feature to construct a posture error feature vector; correcting the posture error feature vector based on posture feature values ​​in the body posture library to obtain a unified posture feature; and passing the unified posture feature through the output layer to obtain the user A's current body posture.

[0099] By adding a joint output module to the posture recognition model, the posture recognition results are corrected by combining the pressure posture recognition results and the image posture recognition results, thereby improving the accuracy of posture recognition.

[0100] In this embodiment, a posture recognition model is used to recognize the posture of user A. The model includes a pressure recognition module, an image recognition module, and a joint output module. The pressure recognition module acquires the user's pressure values ​​on the smart sofa using pressure sensors installed inside, and establishes a pressure distribution map based on these values ​​to identify user A's body posture, making posture recognition more timely and accurate, and allowing for prompt detection and feedback of changes in the user's posture. The image recognition module divides the user image into multiple skeleton points and acquires the position information of each skeleton point. It then performs a weighted fusion operation on the skeleton point features and position information, and uses an attention mechanism to focus on key points to obtain the image posture recognition result. This module, through skeleton point analysis and the weighted fusion of the attention mechanism, not only provides high-precision posture recognition.

[0101] Further, the state determination unit performs a state determination on user A, corresponding to step S50 above; wherein, in performing a state determination on user A, the main identification is between a non-sleep state and a sleep state; if the user is in a non-sleep state, the present invention considers that user A is looking at some kind of object, such as a book, mobile phone, television, or computer, etc., not limited to the objects mentioned above, and regards it as the first state; sleep is regarded as the second state.

[0102] When user A is in the first state;

[0103] Furthermore, the automatic control unit of the system realizes the automatic control of the smart sofa, corresponding to step S60 above; wherein, the automatic control unit includes: an automatic temperature control module and an automatic posture control module; wherein, the automatic temperature control module includes:

[0104] The historical temperature adjustment parameters, historical ambient temperature, and historical body temperature of user A are obtained. The historical temperature can be a one-month historical record or a one-week historical record for analysis. In this embodiment, the data of user A within one week is used as the analysis standard. Refer to Tables 2, 3, and 4 to see the temperature records of user A in the morning, noon, and evening.

[0105] Table 2 Morning Temperature Records Over One Week

[0106]

[0107] Table 3. Midday Temperature Records Over One Week

[0108]

[0109] Table 4. Evening Temperature Records Over One Week

[0110]

[0111] Furthermore, the historical temperature adjustment parameters, the historical ambient temperature, and the user's historical body temperature are constructed into a historical temperature matrix;

[0112] The historical temperature matrix can be established by creating corresponding matrices for morning, noon, and evening; represented as: Where Y represents the vector of the user's historical body temperature; X represents the matrix of the historical temperature adjustment parameters and the historical ambient temperature; θ · ε is represented as the adjustment parameter calculated using the least squares method. · Represented as an error term;

[0113] Based on the historical temperature matrices corresponding to Tables 2, 3, and 4, the values ​​of the adjustment parameters and error terms can be calculated as follows:

[0114] morning:

[0115] noon:

[0116] Night:

[0117] Furthermore, a body temperature influence curve is obtained based on the calculated adjustment parameters and error terms; wherein, the body temperature influence curve is represented as: T b,i =θ(αT) s,i +βT e,i)+ε; where, T b,i T represents the user's body temperature recorded in the i-th entry; s,i This represents the sofa temperature setting recorded in the i-th record; T e Let represent the ambient temperature recorded in the i-th record; α and β represent the parameters of the body temperature influence curve; in this embodiment, the body temperature influence curve at different time points can be obtained based on the adjustment parameters and error terms at different times.

[0118] Furthermore, by monitoring the body temperature of user A and the ambient temperature, the body temperature influence curve at the corresponding time is selected, and the adjustment temperature of the smart sofa is obtained by performing inverse calculation.

[0119] Furthermore, the smart sofa automatically controls the temperature based on the calculated adjustment temperature.

[0120] The attitude automatic control module includes:

[0121] According to the state determination unit, user A's current state is determined to be the first state, which is watching television 3. Figure 5 As shown;

[0122] Furthermore, a spatial coordinate system is established based on the room where user A is located. Based on this spatial coordinate system, the first spatial relative position information of smart sofa 1, the second spatial relative position information of user A, and the third spatial relative position information of television 3 can be obtained. Among them, the relative position information includes: angle information and coordinate information.

[0123] In this embodiment of the application, the coordinate position of the smart sofa 1 is denoted as (x1, y1, z1) and the facing angle is set as θ1; the coordinate position of user A is denoted as (x2, y2, z2) and the facing angle is set as θ2; the coordinate position of the television 3 is denoted as (x3, y3, z3) and the facing angle is set as θ3.

[0124] Furthermore, based on the posture recognition results of user A, the posture spatial information of user A is obtained, including the height of the head and the direction it faces; the orientation of the body and the absolute angle between the reference point of the spatial coordinate axis, etc.

[0125] Furthermore, based on the data information of user A in Table 1 and the body space information, an adjustment set is obtained;

[0126] Furthermore, based on the first spatial relative position information, the second spatial relative position information, and the third spatial relative position information, the first relative distance and the second relative distance are calculated; wherein, the first relative distance is the straight-line distance between the smart sofa 1 and the user A; and the second relative distance is the straight-line distance between the smart sofa 1 and the television 3.

[0127] The first relative distance is: The second relative distance is:

[0128]

[0129] Furthermore, the smart sofa 1 is initially adjusted based on the first relative distance, the second relative distance, and the size of the television 3; the initial adjustment mainly involves automatic control of the height and tilt angle of the backrest 102; the initial height setting of the backrest 102 is such that the user A's eyes are approximately at the center of the television 3, therefore the initial height setting of the backrest 102 is: Where H represents the height of the TV 3; the initial setting of the deflection angle of the backrest 102 is such that the user A's line of sight is aligned with the center of the TV 3 screen, so the initial setting of the deflection angle of the backrest 102 is: ρ-(φ-θ2+θ1); where ρ represents the angle between the user A and the TV 3; φ represents the directional angle between the TV 3 and the smart sofa 1.

[0130] Furthermore, the initial adjustment is automatically adjusted a second time based on the adjustment set and the user's basic information. During the second adjustment, the user A's age, height, and weight, as well as the description of the user A's physical condition in Table 1, are taken into consideration. The height and tilt of the backrest 102 in the smart sofa 1 are adjusted again. The softness and curvature of the part of the seat cushion 103 that contacts the user A's condition are automatically adjusted appropriately. The position and height of the armrest 101 are adjusted according to the user A's hand posture. The height and length of the footrest 104 are automatically controlled according to whether the user A needs to use it and the degree of leg bending.

[0131] In this embodiment, the posture of the smart sofa is adjusted when user A is in a first state. This adjustment process is performed by calculating the relative spatial positions of the smart sofa, the user, and the target object. Based on this, the control parameters of the smart sofa are adjusted a second time by analyzing user A's height, weight, and physical condition. This allows for adaptive adjustment based on user A's personal data.

[0132] The system of this application embodiment also proposes a voice control unit, which can automatically control the smart sofa 1 according to the voice commands of user A, such as temperature control and posture control, and is not limited to all the functions of the above-mentioned method and system of the present invention.

[0133] Furthermore, the parameter display unit of the system is used to display all parameters of the smart sofa 1 during its operation.

[0134] In this embodiment, automatic temperature and posture control are implemented for user A during the use of the smart sofa. This mainly includes the following aspects: First, facial recognition technology is used to obtain the user's personal data from a database to accurately determine the user's height, weight, and health status. Second, this invention proposes a posture recognition model to identify the user's posture. This model determines the user's posture by analyzing the force on the smart sofa and the user's image, and calculates the error of the pressure recognition and image recognition results during the output process to more accurately grasp the user's posture. Finally, this invention automatically controls the temperature and posture of the smart sofa. The invention provides automatic control of the smart sofa's posture. In terms of temperature control, it collects historical temperature data about the user, performs correlation analysis, and establishes a temperature influence curve. Based on this curve, the smart sofa's temperature can be adjusted in real time. Regarding posture control, the invention divides the control scenario into a first state and a second state. When user A is identified as being in the first state, the relative positions of the smart sofa, the user, and the target object are obtained through spatial coordinates. The smart sofa is then initially adjusted based on the calculated distances between them. A second adjustment is made based on the user's personal data and current posture information. This achieves precise automatic control of the smart sofa, enhancing the user experience.

[0135] Example 2:

[0136] In Embodiment 1, the method and system of the present invention are combined to achieve automatic control of user A when using the smart sofa. To illustrate the effectiveness of the present invention for different users, this embodiment demonstrates automatic control of the smart sofa for user B in Table 1 of Embodiment 1 when using the smart sofa. The specific process includes:

[0137] Obtain the body image of the current user of the smart sofa;

[0138] Furthermore, facial recognition is performed based on the acquired image to determine that the current user is user B;

[0139] Further, the user ID corresponding to user B is obtained as PC002;

[0140] Further, based on the ID information, user B's personal data was retrieved from the database; user B is female, 36 years old, 1.72m tall, and weighs 65kg; her physical condition is described as: occasional shoulder discomfort.

[0141] Furthermore, the current body posture of user B is identified; based on the posture recognition model, user B's posture is identified as lying flat on their back.

[0142] Furthermore, the current state of user B is determined, and user B is in the second state, which is sleep.

[0143] Furthermore, automatic control is implemented for user B when using the smart sofa, including automatic temperature control and automatic posture control;

[0144] The method of automatic temperature control is the same as that of the automatic temperature control module in Embodiment 1; the difference is that in this embodiment, the automatic temperature control of the smart sofa is achieved by collecting and analyzing historical temperature data of B.

[0145] The automatic attitude control method differs from that in Example 1; the user B's state is determined to be the second state, therefore the automatic attitude control process for the second state is as follows:

[0146] Obtain the historical sleep state of user B; wherein, the historical sleep state includes: historical sleep posture and historical control parameters; see Table 5, which provides an example of the historical sleep state of user B;

[0147] Table 5 User B's historical sleep status

[0148]

[0149] Furthermore, a sleep data mapping is formed by mapping each historical sleep posture to the corresponding historical control parameters;

[0150] Furthermore, the sleep data mappings are clustered to obtain N similar mapping sets; in this embodiment, the K-means algorithm is used to implement clustering; specifically as follows: the number of similar mapping sets is set to N=3; the K-means algorithm is executed; the clustering results in Table 5 are shown in Table 6:

[0151] Table 6 Examples of historical sleep state clustering

[0152]

[0153] In this example, set 1 contains similar sleep settings, while sets 2 and 3 represent different sleep preferences. Such clustering can help identify common sleep patterns and optimize the sleep environment.

[0154] Furthermore, the N mapping sets are input into a machine learning model to obtain N correspondences between sleep posture and control parameters. This machine learning model can employ a random forest regression model to analyze the correspondence between sleep posture and control parameters. The following are the analysis results of the random forest model on Table 6:

[0155] Set 1 (Side lying, head angle 0°, back angle 30°): Mattress firmness: medium; backrest angle: 15°; foot support height: 5cm;

[0156] Set 2 (Supine position, head angle 5°, back angle 0°): Mattress firmness: soft; backrest angle: 0°; foot support height: 3cm;

[0157] Set 3 (prone position, head angle 10°, back angle 10°): Mattress firmness: firm; backrest angle: 10°; foot support height: 7cm;

[0158] These analyses provide optimal control parameter settings for each sleeping position, which can help optimize the user's sleep experience.

[0159] Furthermore, the similarity between user B's current body posture and the similar mapping set is calculated to obtain similar sleep posture features;

[0160] Furthermore, the control parameters of the smart sofa are obtained by analyzing the corresponding relationship between the sleep posture and control parameters.

[0161] Furthermore, the control parameters are adjusted based on user B's age, gender, height, weight, and physical condition to obtain the optimal control parameters;

[0162] Furthermore, the smart sofa automatically controls itself based on optimal control parameters, including automatic adjustment of the backrest, armrests, footrests, and seat cushion.

[0163] In this embodiment, the user B is in a second state, namely a sleeping state, and the smart sofa is automatically controlled. The control process analyzes the user B's past sleeping postures and historical control parameters to obtain the correspondence between sleeping posture and control parameters. The initial adjustment parameters are obtained by calculating the similarity between the current body posture and the historical data set. The optimal adjustment parameters are then obtained by combining the user's personal data. This enables the user B to effectively adjust the smart sofa while sleeping.

[0164] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An automatic control method for a smart sofa, characterized in that, include: Obtain user image; Facial recognition is performed based on the user image to obtain the user ID; Retrieve the user's personal data based on the user ID; The personal data includes: basic personal information and health status information; Identify the user's current body posture; determine the user's current state; the current state includes: a first state and a second state; The smart sofa is automatically controlled based on its current state; the automatic control includes: automatic temperature adjustment and automatic posture adjustment. The automatic temperature adjustment process includes: acquiring the user's historical temperature adjustment parameters, historical ambient temperature, and historical body temperature; constructing a historical temperature matrix from the historical temperature adjustment parameters, historical ambient temperature, and historical body temperature; the historical temperature matrix is ​​represented as: Y = θX + ε; where Y represents the vector of the user's historical body temperature; X represents the matrix of the historical temperature adjustment parameters and the historical ambient temperature; θ represents the adjustment parameters calculated using the least squares method; ε represents the error term; and a body temperature influence curve is obtained based on the adjustment parameters and the error term, represented as: T b,i =θ(αT) s,i +βT e,i )+ε; where, T b,i T represents the user's body temperature recorded in the i-th entry; s,i This represents the sofa temperature setting recorded in the i-th record; T e The ambient temperature is represented by the i-th record; α and β represent the parameters of the body temperature influence curve; the temperature of the smart sofa is automatically adjusted based on the user's body temperature and the ambient temperature using the body temperature influence curve.

2. The automatic control method for a smart sofa according to claim 1, characterized in that, The current body posture is identified using a posture recognition model; The posture recognition model includes: a pressure recognition module, an image recognition module, and a joint output module; The pressure recognition module's recognition process includes: acquiring pressure values ​​from pressure sensors located at different positions on the smart sofa; denoising the pressure values ​​to obtain standard pressure data; fusing the standard pressure data to construct a pressure distribution map; extracting features from the pressure distribution map to obtain a pressure feature set; wherein the pressure feature set includes location features and pressure features; and the location features and pressure features form a corresponding relationship; inputting the pressure feature set into a pre-trained classification model to obtain a pressure posture recognition result. The image recognition module's recognition process includes: inputting the user image into an image preprocessing layer to obtain a standard user image; inputting the standard user image into a feature extraction layer to obtain human skeleton features and pose features; inputting the human skeleton features and pose features into a feature fusion layer to obtain pose recognition features; wherein the feature fusion layer's fusion process includes: obtaining the connection relationships of each skeleton point in the skeleton features; when the connection relationship is a direct connection, calculating the connection slope of the skeleton point from the pose features; wherein the formula for calculating the connection slope is: Among them, M(P i ,P j ) represents the skeleton point P. i With skeleton point P j The connection slope; (x) i ,y i ) represents the skeleton point P. i The coordinates; (x j ,y j ) represents the skeleton point P. j The coordinates are determined; the human skeleton features are concatenated with the corresponding posture features to obtain concatenated features; the concatenated features are weighted using the connection slope to obtain the posture recognition features; an attention mechanism is used to focus on the posture recognition features of key skeleton points and obtain key features of key skeleton points; the key features are input into the classification layer to obtain the image posture recognition result; the pressure posture recognition result and the image posture recognition result are input into the joint output module to obtain the user's current body posture.

3. The automatic control method for a smart sofa according to claim 2, characterized in that, The specific implementation process of the joint output module includes: obtaining the corresponding stress body posture and image body posture based on the stress posture recognition result and the image body posture recognition result; obtaining the features of the stress body posture and the image body posture respectively to obtain the first posture feature and the second posture feature; calculating the error between the first posture feature and the second posture feature to construct a posture error feature vector; correcting the posture error feature vector according to the posture feature values ​​in the body posture library to obtain a unified posture feature; and passing the unified posture feature through the output layer to obtain the current body posture.

4. The automatic control method for a smart sofa according to claim 1, characterized in that, When the current state is the first state, the process of automatically controlling the smart sofa includes: acquiring the first spatial relative position information of the smart sofa, the second spatial relative position information of the user, the third spatial relative position information of the first object, and the size information of the first object; acquiring the user's posture space information based on the user's current posture recognition result; obtaining an adjustment set based on the user's body condition information and the posture space information; calculating the first relative distance between the smart sofa and the user, and the second relative distance between the smart sofa and the first object based on the first spatial relative position information, the second spatial information, and the third spatial relative position information; performing initial adjustments to the smart sofa based on the first relative distance, the second relative distance, and the size information; and performing secondary automatic adjustments to the initial adjustments based on the adjustment set and the user's basic information; wherein, the secondary automatic adjustments include any one of the following: the height and angle of the backrest, the position of the armrests, and the hardness and curvature of the seat cushion.

5. The automatic control method for a smart sofa according to claim 1, characterized in that, When the current state is the second state, the process of automatically controlling the smart sofa includes: acquiring the user's historical sleep state; wherein, the historical sleep state includes: historical sleep posture and historical control parameters of the smart sofa; forming a sleep data mapping between each historical sleep posture and the corresponding historical control parameter; clustering the sleep data mapping to obtain N similar mapping sets; inputting the N mapping sets into a machine learning model to obtain N sleep posture-control parameter analysis correspondences; calculating the similarity between the current body posture and the similar mapping sets to obtain similar sleep posture features; obtaining the control parameters of the smart sofa based on the corresponding sleep posture-control parameter analysis correspondences; adjusting the parameters based on the control parameters and the personal data to obtain the optimal control parameters; and automatically adjusting any one of the backrest, armrests, footrests, and seat cushions according to the optimal control parameters.

6. An automatic control system for an intelligent sofa, characterized in that, The system includes: an image acquisition unit, a sensor unit, an information storage unit, an identity recognition unit, a posture recognition unit, a status judgment unit, an automatic control unit, a voice control unit, and a parameter display unit. The image acquisition unit is used to acquire image information of the user; the sensor unit is used to acquire pressure values ​​at various locations on the smart sofa; the information storage unit is used to store user information and historical control information of the smart sofa; the identity recognition unit is used to identify the user; the posture recognition unit is used to identify the user's body posture; the status judgment unit is used to determine the user's current status; the automatic control unit is used to automate the control of the smart sofa, including an automatic temperature control module and an automatic posture control module; the voice control unit is used for the user to control the smart sofa via voice; and the parameter display unit is used for the smart sofa to display current control parameters.

7. The automatic control system for a smart sofa according to claim 6, characterized in that, The posture recognition unit uses a posture recognition model for posture recognition; wherein, the posture recognition model includes: a pressure recognition module, an image recognition module, and a joint output module; The pressure recognition module's recognition process includes: acquiring the pressure value of the smart sofa through the sensor unit; denoising the pressure value to obtain standard pressure data; fusing the standard pressure data to construct a pressure distribution map; extracting features from the pressure distribution map to obtain a pressure feature set; wherein the pressure feature set includes: location features and pressure features; and the location features and pressure features form a corresponding relationship; inputting the pressure feature set into a pre-trained classification model to obtain the pressure posture recognition result. The recognition process of the image recognition module includes: acquiring a user image through the image acquisition unit; inputting the user image into an image preprocessing layer to obtain a standard user image; inputting the standard user image into a feature extraction layer to obtain human skeleton features and pose features respectively; inputting the human skeleton features and pose features into a feature fusion layer to obtain pose recognition features; wherein, the fusion process of the feature fusion layer includes: obtaining the connection relationship of each skeleton point in the skeleton features; when the connection relationship is a direct connection, calculating the connection slope of the skeleton point from the pose features; wherein, the formula for calculating the connection slope is: Among them, M(P i ,P j ) represents the skeleton point P. i With skeleton point P j The connection slope; (x) i ,y i ) represents the skeleton point P. i The coordinates; (x j ,y j ) represents the skeleton point P. j The coordinates are determined; the human skeleton features are concatenated with the corresponding pose features to obtain concatenated features; the concatenated features are weighted using the connection slope to obtain the pose recognition features; an attention mechanism is used to focus on the pose recognition features of key skeleton points and obtain the key features of key skeleton points; the key features are input into the classification layer to obtain the image pose recognition result. The pressure posture recognition result and the image posture recognition result are input into the joint output module to obtain the user's current body posture.

8. The automatic control system for a smart sofa according to claim 6, characterized in that, The specific implementation of the automatic temperature control module includes: acquiring the user's historical temperature adjustment parameters, historical ambient temperature, and historical body temperature; constructing a historical temperature matrix from the historical temperature adjustment parameters, historical ambient temperature, and historical body temperature; wherein the historical temperature matrix is ​​represented as: Y = θX + ε; where Y represents the vector of the user's historical body temperature; X represents the matrix of the historical temperature adjustment parameters and the historical ambient temperature; θ represents the adjustment parameters calculated using the least squares method; ε represents the error term; and obtaining the body temperature influence curve based on the adjustment parameters and the error term, represented as: T b,i =θ(αT) s,i +βT e,i )+ε; where, T b,i T represents the user's body temperature recorded in the i-th entry; s,i This represents the sofa temperature setting recorded in the i-th record; T e The ambient temperature is represented by the i-th record; α and β are the parameters of the body temperature influence curve; the temperature of the smart sofa is automatically controlled based on the monitored body temperature and ambient temperature using the body temperature influence curve.

9. The automatic control system for a smart sofa according to claim 6, characterized in that, The posture control module includes: a first state control function and a second state control function; wherein, the first state control function includes: acquiring the first spatial relative position information of the smart sofa, the second spatial relative position information of the user, the third spatial relative position information of the first object, and the size information of the first object; acquiring the user's posture space information based on the user's current posture recognition result; obtaining an adjustment set based on the user's body condition information and the posture space information; calculating the first relative distance between the smart sofa and the user, and the second relative distance between the smart sofa and the first object based on the first spatial relative position information, the second spatial information, and the third spatial relative position information; performing initial adjustments to the smart sofa based on the first relative distance, the second relative distance, and the size information; and performing secondary automatic adjustments to the initial adjustments based on the adjustment set and the user's basic information; wherein, the secondary automatic adjustments include: any one of the following: the height and angle of the backrest, the position of the armrests, and the hardness and curvature of the seat cushion.

10. The automatic control system for a smart sofa according to claim 9, characterized in that, The second state control function includes: acquiring the user's historical sleep state; wherein the historical sleep state includes: historical sleep posture and historical control parameters of the smart sofa; forming a sleep data mapping between each historical sleep posture and the corresponding historical control parameter; clustering the sleep data mapping to obtain N similar mapping sets; inputting the N mapping sets into a machine learning model to obtain N sleep posture-control parameter analysis correspondences; calculating the similarity between the current body posture and the similar mapping sets to obtain similar sleep posture features; acquiring the control parameters of the smart sofa based on the corresponding sleep posture-control parameter analysis correspondences; adjusting the parameters based on the control parameters and personal data to obtain optimal control parameters; and automatically adjusting any one of the backrest, armrests, footrests, and seat cushions according to the optimal control parameters.

Images