Smart home equipment control method and smart home

Abstract

The invention discloses a smart home equipment control method and a smart home, and the method comprises the steps: receiving a current instruction action of a user located in a grid region, and judging whether the current instruction action is a specified instruction action or not; if it is judged that the current instruction action is the specified instruction action, target equipment is determined, a control instruction in the current instruction action is obtained, and whether the current control instruction is the control instruction of the target equipment or not is judged; if yes, controlling the target equipment to complete corresponding operation by using the control instruction; if not, generating a prediction instruction to replace the current control instruction in combination with scene analysis; if it is judged that the current instruction action is not the specified instruction action, a predicted instruction action is generated in combination with scene analysis, the predicted instruction action obtains a control instruction, and the target equipment receives the control instruction to execute operation; according to the invention, through dual verification of real-time matching of the user instruction action and the target equipment control instruction, the target equipment can be accurately controlled, false triggering can be reduced, and the interaction efficiency can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent furniture control, and particularly to a control method for intelligent home appliances and an intelligent home. Background Art

[0002] The intelligent furniture product combination breaks the traditional furniture combination mode, gives full play to the subjective creativity of users, and can be freely combined and matched according to personal preferences and the actual situation of the home space in terms of external dimensions and combination modes. By splitting the furniture functions and processing them unitarily, each unit is a product, and the products can be combined by arranging and stacking. At present, the construction operation of the intelligent furniture system is complex and the degree of product humanization is not high. Summary of the Invention

[0003] The present invention overcomes the deficiencies of the prior art and provides a control method for intelligent home appliances and an intelligent home.

[0004] To achieve the above object, the technical solution adopted by the present invention is: a control method for intelligent home appliances, comprising the following steps:

[0005] Step S1, receiving the current instruction action of a user located in a grid area, and determining whether the current instruction action is a specified instruction action;

[0006] Step S2, if it is determined that the current instruction action is a specified instruction action, determining the target device, obtaining the control instruction in the current instruction action, and determining whether the current control instruction is the control instruction of the target device;

[0007] Step S3, if so, controlling the target device with the control instruction to complete the corresponding operation;

[0008] Step S4, if not, generating a prediction instruction to replace the current control instruction by combining scenario analysis;

[0009] Step S5, if it is determined that the current instruction action is not a specified instruction action, generating a prediction instruction action by combining scenario analysis, obtaining the corresponding control instruction with the prediction instruction action, and the target device receiving the control instruction to execute the corresponding operation.

[0010] In a preferred embodiment of the present invention, in the step S1, the instruction actions include: action trajectory feature, joint angle feature; the action trajectory feature and the joint angle feature are extracted by using computer vision analysis.

[0011] In a preferred embodiment of the present invention, in the step S2, a control instruction is generated based on the specified instruction action, and the control instruction controls the target device to execute the corresponding operation.

[0012] In a preferred embodiment of the present invention, in the step S1, the following steps are included:

[0013] Step S11: Divide the user area into a number of grid areas R in the form of a grid i , and each grid area is defined as R i ={(x,y)|x∈[x i1 ,x i2 , y∈[y i1 ,y i2}, where (x,y) represents the area coordinates, and x i1 ,x i2 ,y i1 ,y i2 represents the area boundary;

[0014] Step S12: Each grid area corresponds to a related device, and a "region-device" mapping table is established;

[0015] Step S13: Real-time locate the grid area where the user is currently located, and store the instruction actions of the related devices in the system database, and determine whether the current instruction action is the instruction action specified in the storage.

[0016] In a preferred embodiment of the present invention, in the step S13, the following sub-steps are included:

[0017] Step S131: Real-time capture the instruction actions in the grid area where the user is located, and extract the key action feature vector A = [a 1 , a 2 , where a 1 , a 2 represent the action trajectory feature and the joint angle feature;

[0018] Step S132: Associate the predefined instruction actions in Step S131 with the control instructions to form a mapping M = {(A j , Y j )|j = 1,2,...,n}, A j represents the instruction action, and Y j represents the control instruction corresponding to the instruction action;

[0019] Step S133: Set the similarity threshold of the instruction action, and pair the current instruction action A a of the user with the instruction actions in its current area: X represents the similarity threshold.

[0020] In a preferred embodiment of the present invention, in the step S2, the method for determining the target device includes the following steps:

[0021] Step S21: Obtain the user's orientation feature, grid area location, time feature, and habit feature;

[0022] Step S22: Determine the set of candidate devices {D 1 , D 2 , …, D n} within the grid area;

[0023] Step S23: Initially determine the candidate devices according to the user's orientation feature;

[0024] Step S24: For each candidate device, calculate its comprehensive weight n represents the number of influencing factors, F i represents the i-th influencing factor affecting the device score, a i represents the weight coefficient of the i-th influencing factor, and select the one with the largest proportion as the current target device.

[0025] In a preferred embodiment of the present invention, in the method of combining scenario analysis with the prediction model, it includes the following steps:

[0026] Step S41: Use computer vision to determine the user's real-time location and the target device;

[0027] Step S42: Extract the control instruction of the current command action and calculate the similarity between it and each control instruction of the target device;

[0028] Step S43: Integrate the scenarios of time features and habit features, calculate the weight of each control instruction, and select the instruction with the highest weight as the prediction instruction.

[0029] In a preferred embodiment of the present invention, in step S4, combining scenario analysis further includes: combining the indoor temperature and humidity environment, the outdoor temperature and humidity environment, and the weather conditions.

[0030] A smart home includes: a grid division module, an action detection module, a target device matching module, a prediction control module, and an execution module;

[0031] The grid division module is used to divide the space where the user is located into several grid areas;

[0032] The action detection module is used to receive the user's current command action and determine whether it is a specified command action;

[0033] The target device matching module is used to determine the target device and judge whether the current command action matches the target device;

[0034] The prediction control module is used to generate a prediction instruction by combining scenario analysis;

[0035] The execution module is used to operate the target device according to the generated control instruction;

[0036] The predictive control module includes: a scenario feature extraction unit, an instruction matching unit, and an instruction generation unit;

[0037] The scenario feature extraction unit is used to extract scenario features such as the user's orientation, location, time, and environmental state;

[0038] The instruction matching unit is used to calculate the similarity between the current instruction action and the control instruction of the target device according to the features;

[0039] The instruction generation unit is used to combine scenario analysis and generate the final control instruction of the target device.

[0040] In a preferred embodiment of the present invention, the grid division module dynamically adjusts the range of the grid area according to the spatial layout information and the device distribution information.

[0041] The present invention solves the defects in the background technology and has the following beneficial effects:

[0042] (1) The present invention provides a method for controlling smart home devices. The user's residence is divided into several grid areas. By real-time monitoring the matching between the instruction actions in the area where the user is located and the stored specified instruction actions, it is judged whether the user's instruction actions are specified instruction actions, ensuring that the user's operation intention can be accurately interpreted, avoiding incorrect triggering of devices due to misrecognition. At the same time, the target device is determined, and the control instruction generated by the obtained specified action is matched with the control instruction stored in the target device to judge whether the current control instruction is the control instruction of the target device. Through the double verification of real-time matching of the user's instruction actions and the target device control instructions, it is possible to accurately control the target device while reducing mis-triggering, improving the interaction efficiency, and enhancing the intelligence and fault tolerance ability, significantly improving the user's operation experience.

[0043] (2) By restricting the user's instruction actions within the grid area where the user is located and binding the control relationship between each grid area and specific devices, the present invention not only improves the accuracy and efficiency of instruction transmission, but also can quickly identify the target device and eliminate the interference of cross-area devices, avoiding mis-triggering or multi-device conflicts, significantly enhancing the reliability and safety of operations; at the same time, the division of grid areas only needs to process the device data in the current area, improving the efficiency of instruction matching and providing an intuitive and natural interaction experience for users.

[0044] (3) By generating a prediction instruction through scenario analysis that combines time characteristics and habit characteristics to replace the current control instruction, the present invention can accurately predict the true intention of the user, improve the prediction accuracy, thereby automatically generating a more scenario-compliant control instruction when the user's instruction is unclear or incorrect. At the same time, the prediction reduces the need for the user's repeated operations, enhances the convenience of operation, and continuously accumulates user usage data, achieving efficient, fault-tolerant, and scenario-based device control, and enhancing the practicality and user experience of the smart home. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0046] Figure 1 is a flowchart of a smart home device control method according to a preferred embodiment of the present invention;

[0047] Figure 2 is a flowchart of the grid area definition according to a preferred embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0049] Many specific details are set forth in the following description in order to provide a thorough understanding of the present invention, but the present invention can also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited by the specific embodiments disclosed below.

[0050] APPLICATION OVERVIEW

[0051] As mentioned above, in the field of smart furniture, the interconnection or transmission of information or other signals is utilized to achieve the control of smart furniture.

[0052] In the control of smart furniture, the visual controller of smart furniture can collect the instruction actions of users and then control the smart furniture according to the results of the instruction actions without affecting the current usage process of the smart furniture functions by users. That is, users no longer need to search for remote controls or physical buttons to operate the functions, which improves the convenience of user operations. For example, when a user enters the living room and wants to turn on the TV or adjust the TV channels and volume, they can trigger instructions through instruction actions without using a remote control or physical buttons and directly control functions such as turning on the TV or adjusting the TV channels and volume.

[0053] Moreover, smart furniture such as TVs is placed in places such as the living room, which is a public space in the home. There is only one user using it at a time. When a user uses this smart furniture and uses instruction actions to control functions, it is easily interfered by other smart furniture in the same space or other spaces. For example, the instruction action of "opening the window" for the window in the adjacent area is similar to the instruction action of "turning on the TV" for the TV. When a user uses the instruction action of "turning on the TV" for the TV, it is easy to cause confusion with the instruction action of "opening the window" for the window, resulting in interference between instruction actions and making it impossible to effectively distinguish and extract the instruction actions of users. Therefore, a smart home device control method and a smart home are designed to solve the problem of confusion caused by similar instruction actions.

[0054] As Figure 1 shown, a smart home device control method includes the following steps:

[0055] Step S1: Receive the current instruction action of the user in the grid area and determine whether the current instruction action is a specified instruction action;

[0056] Step S2: If it is determined that the current instruction action is a specified instruction action, determine the target device, obtain the control instruction in the current instruction action, and determine whether the current control instruction is the control instruction of the target device;

[0057] Step S3: If so, control the target device to complete the corresponding operation with the control instruction;

[0058] Step S4: If not, generate a predicted instruction to replace the current control instruction by combining scenario analysis;

[0059] Step S5: If it is determined that the current instruction action is not a specified instruction action, generate a predicted instruction action by combining scenario analysis, and obtain the corresponding control instruction with the predicted instruction action. The target device accepts the control instruction and executes the corresponding operation.

[0060] In the present invention, in step S1, the instruction actions include: action trajectory features and joint angle features; the action trajectory features and joint angle features are extracted by using computer vision analysis.

[0061] As Figure 2 shown, in step S1, the following steps are included:

[0062] Step S11: Divide the user area into a number of grid areas R in the form of a grid i , and each grid area is defined as R i ={(x,y)|x∈[x i1 ,x i2 , y∈[y i1 ,y i2}, where (x,y) represents the area coordinates, and x i1 ,x i2 ,y i1 ,y i2 represent the area boundaries;

[0063] Step S12: Each grid area corresponds to relevant devices, and a "region-device" mapping table is established. For example, the mapping table for a certain area in the living room includes: R (2，3) ={TV, lights, stereo}. By using computer vision to locate the grid area where the user is currently located, and by binding the devices to the area, the device list in the user's current area can be directly located, reducing the instruction recognition range, thereby determining the range of devices that the user can control; and the operations of the user within a grid area only affect the devices bound to that area, avoiding accidental triggering of devices in other areas;

[0064] Step S13: Real-time locate the grid area where the user is currently located, and store the instruction actions of the relevant devices in the system database. By determining whether the current instruction action is the instruction action specified in the storage, and by monitoring in real time whether the instruction action in the user's area matches the stored specified instruction action, it is judged whether the user's instruction action is the specified instruction action, ensuring that the user's operation intention can be accurately interpreted and avoiding incorrect triggering of devices due to misrecognition;

[0065] In step S13, the following sub-steps are included:

[0066] Step S131: Real-time capture the instruction actions in the grid area where the user is located, and extract the key action feature vector A = [a 1 ,a 2 , where a 1 ,a 2 represent the action trajectory feature and the joint angle feature;

[0067] Step S132: Associate the predefined instruction actions in step S131 with the control instructions, where the predefined instruction actions are the stored instruction actions, and the stored instruction actions and the control instructions form a mapping M = {(A j ,Y j)|j = 1, 2, …, n}, A j represents the instruction action, Y j represents the control instruction corresponding to the instruction action;

[0068] Step S133: Set the similarity threshold of the instruction action, and pair the user's current instruction action A a with the instruction actions within its current area: X represents the similarity threshold. Through the similarity matching of the real-time captured user actions and the stored instruction actions, if the similarity is greater than the similarity threshold, it means that the user's instruction action is standard, that is, it can be matched among the stored instruction actions.

[0069] During the user's use of the smart home, for example, when the user needs to operate the TV in the living room, the user issues a waving finger instruction action within the grid area of the living room. The device list of the living room grid area is R( 2，3 ) = {TV, light, stereo}. Receive the user's waving action, extract the action trajectory and joint angles through computer vision to generate a feature vector, match the features from the action library bound to the living room devices. For example, the feature vector of "turn on the TV" is F, then calculate the similarity. The similarity is 0.92, and its similarity threshold is 0.85. Since it is higher than the similarity threshold, it indicates that the user's instruction action is the specified instruction action. Therefore, it is determined as the "turn on the TV" instruction, and the instruction is transmitted to the living room TV device, and the TV is successfully turned on, while the devices in other areas such as the bedroom light are not affected.

[0070] Therefore, by comparing the user's instruction actions with the stored specified instruction actions one by one within a specific grid area, the search range of the instruction action library is reduced, thereby improving the matching speed and accuracy; for example, when the user issues an instruction action within the grid area of the living room, only need to match the actions bound to the living room devices, without considering the device actions in the kitchen or bedroom; the segmentation of the grid area clarifies the user's operation range, can quickly respond to the user's needs, reduce the complexity of the user's operation, and provide a personalized smart home experience.

[0071] In the present invention, in step S2, based on the specified instruction action, a control instruction is generated, and the control instruction controls the target device to perform corresponding operations.

[0072] In a preferred embodiment of the present invention, in step S2, the method for determining the target device includes the following steps:

[0073] Step S21: Obtain the user's orientation feature, grid area location, time feature, and habit feature. Capture the user's face or gesture direction through computer vision to infer the user's operation target. If there is only one device in the user's face or gesture direction, determine this device as the target device and directly indicate the device the user is concerned about to improve the recognition efficiency; based on grid division, clarify the area where the user is located and determine the device list within the grid area; obtain the time feature by converting the current time into discrete time periods (such as morning, afternoon, evening) to optimize the device selection logic, and obtain the habit feature by analyzing the user's operation preferences in similar scenarios through historical records;

[0074] Step S22: Determine the set of candidate devices {D 1 、D 2 、…、D n} within the grid area;

[0075] Step S23: Initially determine the candidate devices according to the user's orientation feature;

[0076] Step S24: If there are multiple devices in the user's face or gesture direction, for each candidate device, calculate its comprehensive weight n represents the number of influencing factors, F i represents the i-th influencing factor affecting the device score, a i represents the weight coefficient of the i-th influencing factor, and select the one with the largest proportion as the current target device, and its weight coefficient is judged through experience.

[0077] In a preferred embodiment of the present invention, in step S4, since the instruction action made by the user may not be the instruction action stored in the devices within this grid area, it is necessary to combine scenario analysis to generate a predicted instruction to replace the control instruction generated by the current instruction action;

[0078] The method of combining scenario analysis in the prediction model includes the following steps:

[0079] Step S41: Use computer vision to determine the user's real-time position and the target device;

[0080] Step S42: Extract the control instruction of the current instruction action, extract the feature vector of each control instruction from the control instruction library, denoted as the set {N i}, calculate the similarity between the control instruction N of the current instruction action and each control instruction to determine the candidate set of the target device control instruction closest to it. Among them, before calculating the similarity between the control instructions, it is necessary to pre-define and associate its feature vector for each control instruction;

[0081] Use the similarity index to calculate the matching degree between the current instruction action and the target device control instruction: Define a matching degree threshold. If the matching degree S is higher than the threshold, it indicates that the control instruction of the current instruction action is closer to the control instruction of the target device;

[0082] Step S43: Based on the scenarios of comprehensive time features and habit features, calculate the weight of each control instruction, and select the instruction with the highest weight as the prediction instruction;

[0083] Regarding the comprehensive time features, since there is a relationship between the current time and the usage pattern of device control. For example, the probability of operating the lights is higher in the evening, and the probability of opening the curtains is higher in the morning;

[0084] According to the historical operation records, distribute a time probability function for each target device control instruction where ni(T) represents the historical trigger times of control instruction i corresponding to time period T, represents the total trigger times of all control instructions i in time period T. The larger P1 is, the more suitable the current time is for control instruction i of the target device;

[0085] Regarding the comprehensive habit features, based on the user's historical operation habits in the current regional grid, calculate the preference probability of the instruction, and define the habit preference probability function mi(H) represents the cumulative trigger times of control instruction i under habit feature H, represents the total trigger times of all control instructions i under habit feature H. The larger P2 is, the higher the user's preference degree for control instruction i of the target device;

[0086] Based on all the matching probabilities, calculate the comprehensive weight W of the candidate control instructions i ,

[0087] W i = w1·S + w2·P1 + w3·P2, where w1, w2, and w3 are all weight coefficients. Obtain the candidate control instruction set, select the control instruction with the largest weight from the candidate control instruction set as the prediction instruction, replace the current incorrect instruction with the generated prediction instruction, and send it to the target device to complete precise control.

[0088] Match the control instruction generated by the specified action obtained with the control instruction stored in the target device, and determine whether the current control instruction is the control instruction of the target device. Through the double verification of real-time matching of the user instruction action and the target device control instruction, it is possible to reduce mis-triggering while precisely controlling the target device, improve the interaction efficiency, enhance the intelligence and fault tolerance ability, and significantly improve the user's operation experience.

[0089] Scenario analysis that combines time features and habit features generates prediction instructions to replace the current control instructions, which can accurately predict the true intentions of users, improve prediction accuracy, and thus automatically generate more scenario-compliant control instructions when user instructions are unclear or incorrect. At the same time, prediction reduces the need for users to perform repeated operations and enhances the convenience of operations.

[0090] When the user uses smart furniture in the grid area of the living room, the orientation direction of the user is confirmed, and the target device is the TV. The set of control instructions for the TV is {turn on, turn off, adjust volume, switch channels}. The user makes an instruction action of waving upwards, aiming to turn on the TV. The control instruction generated by the instruction action of waving upwards does not match the control instructions in the TV, but rather matches the control instructions for "turn on the sound". The matching degree between the current control instruction and the control instructions for the sound is 1, which is higher than the matching degree threshold (0.85), and the matching degree between the current control instruction and the control instructions for the TV is 0.7, which is lower than the matching degree threshold.

[0091] It is determined that the current time is 7 pm, and the operation probability of the TV is relatively high. The matching probability of the time feature of the TV is 0.85, and the matching probability of the time feature of the sound is 0.3.

[0092] Combined with the analysis of habit features, the control instruction generated by the instruction action of waving upwards is the "turn on" control instruction in the historical data. In this time period, the preference matching probability of the "turn on" control instruction for the TV is 0.9, and the preference matching probability of the "turn on the sound" control instruction for the sound is 0.4.

[0093] In calculating the comprehensive weight of the candidate control instructions, it is defined that the weight coefficient w1 = 0.4, w2 = 0.3, w3 = 0.3. In summary, the matching probabilities are calculated respectively, and the comprehensive weight of the TV is obtained: 0.7×0.4 + 0.85×0.3 + 0.9×0.3 = 0.805, and the comprehensive weight of the sound is obtained: 1×0.4 + 0.3×0.3 + 0.4×0.3 = 0.61. Then, the control instruction with the highest weight is selected, and the prediction instruction generated is the "turn on" control instruction for the TV, replacing the incorrect control instruction of "turn on the sound" for the sound.

[0094] It should be noted that in the allocation of the weight coefficient, in the initial stage, the weight value is manually set according to experience. For example, when the target device is a light or a curtain, the time feature weight is set relatively high because the turning on and off of the light and the opening and closing of the curtain have a greater relationship with time. If the target device is a computer or a sound, the proportion of the habit feature weight is set relatively high because the turning on or off of the computer and the sound have a greater relationship with the habit feature.

[0095] In a preferred embodiment of the present invention, in step S4, the context analysis further includes: combining the indoor temperature and humidity environment, the outdoor temperature and humidity environment, and the weather conditions.

[0096] A smart home includes: a grid division module, an action detection module, a target device matching module, a prediction control module, and an execution module;

[0097] The grid division module is used to divide the space where the user is located into several grid areas;

[0098] The action detection module is used to receive the user's current instruction action and determine whether it is a specified instruction action;

[0099] The target device matching module is used to determine the target device and judge whether the current instruction action matches the target device;

[0100] The prediction control module is used to generate a prediction instruction by combining context analysis;

[0101] The execution module is used to operate the target device according to the generated control instruction;

[0102] The prediction control module includes: a context feature extraction unit, an instruction matching unit, and an instruction generation unit;

[0103] The context feature extraction unit is used to extract context features such as the user's orientation, position, time, and environmental status;

[0104] The instruction matching unit is used to calculate the similarity between the current instruction action and the target device control instruction according to the features;

[0105] The instruction generation unit is used to combine context analysis and generate the final control instruction for the target device.

[0106] In the present invention, the grid division module dynamically adjusts the range of the grid area according to the space layout information and the device distribution information.

[0107] Based on the ideal embodiments of the present invention as inspiration, through the above description, relevant personnel can completely make various changes and modifications without departing from the technical idea of the present invention. The technical scope of the present invention is not limited to the content in the specification, and must be determined according to the scope of the claims.

Claims

1. A smart home device control method, characterized in that: The following steps are involved: Step S1, receiving the current command action of the user in the grid area, and determining whether the current command action is a prescribed command action; Step S2: if the current instruction action is determined to be a prescribed instruction action, the target device is determined, the control instruction in the current instruction action is obtained, and whether the current control instruction is a control instruction of the target device is determined; Step S3: If yes, control the target device with the control instruction to complete the corresponding operation; Step S4: If not, generate a prediction instruction to replace the current control instruction in combination with scenario analysis; Step S5: If it is determined that the current command action is not the prescribed command action, a predicted command action is generated in combination with the scenario analysis, and a corresponding control instruction is obtained based on the predicted command action. The target device receives the control instruction and performs the corresponding operation.

2. A smart home device control method according to claim 1, characterized in that: In the step S1, the command action includes: action trajectory features and joint angle features; the action trajectory features and joint angle features are extracted by computer vision analysis.

3. A smart home device control method according to claim 1, characterized in that: In the step S2, a control instruction is generated based on a specified instruction action, and the control instruction controls the target device to perform a corresponding operation.

4. A smart home device control method according to claim 2, characterized in that: In step S1, the following steps are included: Step S11: Divide the user area into several grid areas R in a grid form i , each grid area is defined as R i ={(x,y)|x∈[x i1 ,x i2 ],y∈[y i1 ,y i2 ]}, where (x,y) represents the region coordinates, x i1 ,x i2 ,y i1 ,y i2 Indicates the region boundary; Step S12: Each grid area corresponds to a related device, and a "area-device" mapping table is established; Step S13: locate the grid area where the user is currently located in real time, and store the command actions of related devices in the system database, and determine whether the current command action is the command action specified in the storage.

5. A smart home device control method according to claim 4, characterized in that: In the step S13, the following sub-steps are included: Step S131, capturing the command action in the grid area in real time, and extracting the key action feature vector A=[a1, a2] of the command action, where a1 and a2 represent the action trajectory feature and the joint angle feature; Step S132: associate the command action predefined in step S131 with the control command to form a mapping M={(A j ,Y j )|j=1,2,…,n},A j Indicates command action, Y j Indicates the control instruction corresponding to the instruction action; Step S133: Set the similarity threshold of the command action and convert the user's current command action A a Paired with the command action in the current area: X represents the similarity threshold.

6. A smart home device control method according to claim 1, characterized in that: In step S2, the method for determining the target device includes the following steps: Step S21, obtaining the user's orientation feature, grid area position, time feature and habit feature; Step S22: Determine a candidate device set {D1, D2, ..., D n }; Step S23: preliminarily determine candidate devices according to the user's orientation characteristics; Step S24: For each candidate device, calculate its comprehensive weight n represents the number of influencing factors, F i represents the i-th influencing factor affecting the device score, a i Represents the weight coefficient of the i-th influencing factor, and selects the one with the largest proportion as the current target device.

7. A smart home device control method according to claim 1, characterized in that: The method of combining the forecast model with scenario analysis comprises the following steps: Step S41, using computer vision to determine the user's real-time location and target device; Step S42, extracting the control instruction of the current instruction action, and calculating the similarity between the control instruction of the target device and each control instruction; Step S43, comprehensively consider the time characteristics and the scenarios of the habit characteristics, calculate the weight of each control instruction, and select the instruction with the highest weight as the predicted instruction.

8. A smart home device control method according to claim 1, characterized in that: In step S4, combining scenario analysis also includes: combining indoor temperature and humidity environment, outdoor temperature and humidity environment, and weather conditions.

9. A smart home, characterized in that: include: Grid division module, motion detection module, target device matching module, prediction control module and execution module; The grid division module is used to divide the space where the user is located into a number of grid areas; The action detection module is used to receive the user's current command action and determine whether it is a specified command action; The target device matching module is used to determine the target device and determine whether the current instruction action matches the target device; The prediction control module is used to generate prediction instructions in combination with scenario analysis; The execution module is used to operate the target device according to the generated control instruction; The prediction control module includes: a scenario feature extraction unit, an instruction matching unit and an instruction generation unit; The scenario feature extraction unit is used to extract scenario features of user orientation, location, time, and environmental status; The instruction matching unit is used to calculate the similarity between the current instruction action and the target device control instruction according to the feature; The instruction generation unit is used to combine the scenario analysis and generate the final control instruction of the target device.

10. A smart home according to claim 9, characterized in that: The grid division module dynamically adjusts the range of the grid area according to the spatial layout information and the equipment distribution information.

Images