Interaction control method and system based on smart home equipment

Abstract

The invention belongs to the field of computer control systems, and provides an interaction control method and system based on smart home equipment, and the method comprises the steps: extracting the context information of a current user when there is missing control parameters, the context information comprising time information, a spatial position, an environment state and a user identity; constructing a control context map of the current user; searching a neighborhood user set similar to the behavior portrait characteristics of the current user, and extracting control parameter records of the neighborhood users in the context similar to the current user; performing weighted fusion on the historical control parameter of the current user and the control parameter of the neighborhood user to generate a target control parameter; forming a complete control command, and issuing the complete control command to the target smart home device to execute a control operation; and updating the control context map based on user feedback or a system monitoring result.

Description

Technical Field

[0001] The present invention belongs to the field of computer control systems, and more particularly relates to an interactive control method and system based on smart home devices. Background Art

[0002] With the rapid development of technologies such as the Internet of Things, artificial intelligence, and edge computing, smart home systems have gradually become popular and are widely used in domestic life scenarios. Smart homes usually achieve unified management and intelligent control of home devices such as lighting, air conditioners, security systems, audio systems, and curtains through the coordinated operation of sensing terminals, control centers, and execution devices. Users can send control requests to the system through various interaction methods such as voice commands, mobile terminals, touch panels, and gesture recognition, so as to achieve remote control and automatic response of the devices.

[0003] In existing smart home systems, the control methods are mostly based on clear instructions issued by users, such as "set the air conditioner to 26 degrees" and "adjust the light brightness to 50%". For such control requirements with clear parameters, the system can directly generate commands according to the device control protocol and execute them. However, in actual use, users often express control intentions in a vague or incomplete manner, such as "turn on the air conditioner", "make the light brighter", and "play music", without clearly giving the required parameters. At this time, traditional systems usually handle it by setting default values, asking users to supplement information, or calling preset scenario modes, but the above methods have problems such as inflexible response, weak adaptability, and insufficient personalization.

[0004] The existing technologies have the following deficiencies in dealing with fuzzy instructions: First, relying on default values for response is likely to deviate from user expectations; second, lacking an intelligent reasoning mechanism based on user context and historical behavior, it cannot fully reflect personalized needs; third, ignoring the time evolution of user behavior and the migration law of group behavior, the system cannot effectively cope with complex situations such as new users, sparse data, or changing habits, and it is difficult to achieve the intelligent interaction goal of truly "understanding users and adapting to users". Summary of the Invention

[0005] To solve the problems in the existing technologies, the present invention provides an interactive control method based on smart home devices, including the following steps:

[0006] Step S10: Receive a control instruction input by a user, and identify whether there is a lack of control parameters in the control instruction;

[0007] When there is a lack of control parameters, extract the context information of the current user, where the context information includes time information, spatial location, environmental state, and user identity identifier;

[0008] Step S20: Based on the context information, construct a control context map for the current user. The control context map is a weighted graph structure, where nodes represent the historical control intentions of the user in the corresponding context, and edges represent the correlations between control intentions at different time sequences;

[0009] Step S30: In the user behavior portrait space, retrieve a set of neighboring users whose features are similar to those of the current user behavior portrait, and extract the control parameter records of the neighboring users in the context similar to that of the current user;

[0010] Step S40: Perform weighted fusion of the historical control parameters of the current user and the control parameters of the neighboring users to generate target control parameters;

[0011] Step S50: Complete the target control parameters into the control instruction to form a complete control command, and send the complete control command to the target smart home device to perform a control operation;

[0012] Step S60: After the control operation is completed, update the control context map based on user feedback or system monitoring results.

[0013] Further, step S10 includes the following sub-steps:

[0014] Obtain the voice input, graphical interface input or sensor trigger event of the current user, and parse it into a control intention instruction;

[0015] Perform integrity recognition of the parameter field based on whether the control intention contains explicit control parameters;

[0016] When it is identified that the control parameters are missing, extract the context information of the current user, where:

[0017] The time information obtains the current time through the local clock or network clock and matches it to the preset time period label;

[0018] The spatial location performs room positioning through the physical location information of the user operation terminal, Bluetooth beacon or wearable device bound to the user;

[0019] The environmental state is obtained through temperature and humidity sensors, light sensors or air quality sensors deployed in the room;

[0020] The user identity identifier is determined by voiceprint recognition, face recognition or user device ID recognition methods.

[0021] Further, step S20 includes the following sub-steps:

[0022] Based on the historical control behavior records of the current user, extract the context labels and control parameters of each control behavior;

[0023] Aggregate the behaviors with the same control intention under the same context label into nodes, and record the control device type, control parameters, context label, and behavior frequency in the nodes;

[0024] Connect the continuously occurring control behaviors in chronological order to form an edge relationship, where the edge represents the behavior path of the front and back control actions;

[0025] Assign edge weights to the edges between nodes according to the co-occurrence frequency, time interval, and context similarity of the user's control behaviors;

[0026] The constructed control context graph is a weighted directed graph structure, which is used to express the evolution trend of the user's control intention in different behavior paths.

[0027] Furthermore, step S30 includes the following sub-steps:

[0028] Encode the behavior pattern of the current user in the control context graph into a behavior portrait feature vector;

[0029] In the user behavior portrait space, use the behavior feature similarity calculation method to identify users with similar behavior patterns;

[0030] When the similarity between the behavior portraits is higher than the set threshold, determine that the user is a neighborhood user;

[0031] Screen the control behaviors of the neighborhood users under similar time periods, spatial positions, and environmental states, and extract the corresponding control parameter values;

[0032] Record the screened control parameter values together with the context label as the neighborhood sample set.

[0033] Furthermore, step S40 includes the following sub-steps:

[0034] Merge the historical control parameter samples of the current user under similar contexts with the control parameter samples of the neighborhood users to form a candidate parameter set;

[0035] For each control parameter sample, set its fusion weight, and the weight is determined by the following factors:

[0036] The behavior occurrence frequency of the current user's parameter sample;

[0037] The similarity between the neighborhood user portrait and the current user portrait;

[0038] The matching degree of the sample with the current context label;

[0039] According to the weighting rule, sort, screen, or perform weighted averaging on the candidate parameter samples to generate the target control parameter;

[0040] The target control parameter is used to complete the missing parameter part in the user control instruction to form a complete control command.

[0041] On the other hand, the present invention also provides an interactive control system based on a smart home device, characterized in that the system includes the following modules:

[0042] An input recognition module, configured to receive a control instruction input by a user and identify whether there is a missing control parameter in the control instruction;

[0043] A context extraction module, configured to extract the context information of the current user when it is recognized that there is a missing control parameter, where the context information includes time information, spatial location, environmental state, and user identity identifier;

[0044] A graph construction module, configured to construct a control context graph of the current user based on the context information, where the control context graph is a weighted graph structure, and nodes represent the historical control intentions of the user in the corresponding context, and edges represent the correlation between different time-sequence control intentions;

[0045] A similar retrieval module, configured to retrieve a set of neighboring users whose behavioral portrait features are similar to those of the current user in the user behavior portrait space, and extract the control parameter records of the neighboring users in the context similar to that of the current user;

[0046] A fusion calculation module, configured to perform weighted fusion on the historical control parameters of the current user and the control parameters of the neighboring users to generate a target control parameter;

[0047] An instruction generation module, configured to supplement the target control parameter to the control instruction to form a complete control command, and send the complete control command to the target smart home device to perform a control operation;

[0048] A graph update module, configured to update the control context graph based on user feedback or system monitoring results after the control operation is completed.

[0049] Further, the context extraction module includes:

[0050] A time extraction unit, configured to obtain the current time through a local clock or a network clock and match it to a preset time period label;

[0051] A location recognition unit, configured to determine the spatial location where the user is located through the physical location information of the user terminal, Bluetooth beacon, or communication information of a wearable device;

[0052] An environment perception unit, configured to collect environmental state information through deployed temperature and humidity sensors, light sensors, or air quality sensors;

[0053] An identity recognition unit for determining the issuer of the current control instruction through voiceprint recognition, face recognition, or user device identification information.

[0054] Further, the map construction module includes:

[0055] A behavior aggregation unit for aggregating historical control behaviors with the same context label and control intention into the same node;

[0056] A path construction unit for establishing edge relationships between adjacent nodes in the order of behavior occurrence, where the edges record the time sequence and co-occurrence frequency of control behaviors;

[0057] A weight assignment unit for assigning weights to the edges according to the node occurrence frequency, context similarity, and behavior transition probability, to represent the correlation between different time-sequence control intentions;

[0058] A map output unit for outputting a control context map structure with nodes and weighted edges.

[0059] Further, the similarity retrieval module includes:

[0060] A portrait encoding unit for encoding the behavior pattern of the current user in the control context map into a behavior portrait feature vector;

[0061] A similarity calculation unit for calculating the similarity between the behavior portrait of the current user and other user portraits, and determining users with a similarity higher than the set threshold as neighborhood users;

[0062] A sample extraction unit for extracting control parameter values from the control behaviors of neighborhood users under similar context conditions, and recording them together with their context labels as a neighborhood parameter sample set.

[0063] Further, the fusion calculation module includes:

[0064] A sample sorting unit for merging the historical control parameters of the current user and the control parameters of neighborhood users into a candidate parameter sample set;

[0065] A weight calculation unit for determining the fusion weights of each parameter sample according to the behavior frequency of the current user sample, the similarity between the neighborhood user and the current user portrait, and the context matching degree;

[0066] A parameter generation unit for performing weighted average, sorting, or screening on the candidate parameter samples according to the weighting rule to generate target control parameters to complete the original control instruction.

[0067] The interactive control method and system based on smart home devices proposed by the present invention can, in the case of users inputting fuzzy control instructions, combine the current context information, user historical behavior, and neighborhood user control data, and through constructing a control context map and a behavior portrait space, realize the intelligent inference and completion of missing control parameters, effectively improving the system's understanding ability of natural instructions and response accuracy. This method supports the dynamic evolution of behavior data and the map update mechanism, and has good adaptability and continuous learning ability.

[0068] The present invention not only improves the personalization and scenario adaptability of control decisions, but also enables the system to have the control ability from "habit recognition" to "behavior prediction" through a multi-source information fusion and map-driven inference mechanism, showing a higher intelligent level and stability when dealing with complex application scenarios such as user preference changes, family member differences, and fuzzy language input, and significantly improving the user's intelligent interaction experience. Brief Description of the Drawings

[0069] 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 to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0070] Figure 1 is the flowchart of the method of the present invention;

[0071] Figure 2 is the system block diagram of the present invention. Detailed Embodiments

[0072] The following will make a preferred description of the invention in combination with the drawings and specific embodiments.

[0073] This embodiment solves the above problems through the following steps:

[0074] Smart home refers to a residential environment system that uses modern sensing technology, computing technology, communication technology, and artificial intelligence technology to interconnect various electrical appliances, lighting systems, security systems, environmental monitoring systems, entertainment terminals, etc. in the home through a network, and combines a human-computer interaction interface to achieve unified, coordinated, and automated control. The core goal of smart home is to improve residential comfort, security, energy efficiency, and user experience, enabling the home space to have the intelligent characteristics of "perception - understanding - response".

[0075] Common smart home devices include but are not limited to: smart lighting systems, smart air conditioners, smart curtains, smart door locks, voice assistants, smart cameras, smart speakers, etc. These devices generally have programmability, remote access capabilities, and status feedback functions.

[0076] The interactive control of a smart home refers to the process in which a user issues a control intention to home devices through various interaction methods (such as voice control, touch panel, mobile terminal APP, gesture recognition, etc.), and the system makes interpretations, decisions and issues corresponding control instructions according to the control intention, driving the devices to perform the expected behaviors.

[0077] In the actual use process, users often express control intentions in a "fuzzy" and "incomplete" way. For example:

[0078] The user says "turn on the air conditioner", but does not specify specific parameters such as temperature, mode (cooling / heating), wind speed, etc.;

[0079] The user says "make the light brighter", but does not specify the brightness value or specific lamps;

[0080] The user says "play music", but does not specify the music genre, volume or device terminal.

[0081] To solve the above problems, in one embodiment, an interactive control method based on smart home devices is proposed, which belongs to the field of computer control. According to the control instructions issued by the user and combined with the environmental context, user behavior characteristics and historical control preferences, it automatically completes the control parameters and realizes an interactive control method with dynamic response of smart devices, aiming to solve the problems of insufficient response ability to fuzzy instructions, poor user experience and lack of personalization and context adaptability in control decisions in the prior art. As Figure 1 shown, the method specifically includes the following steps:

[0082] Step S10: Receive the control instructions input by the user, and identify whether there are missing control parameters in the control instructions;

[0083] When there are missing control parameters, extract the context information of the current user, and the context information includes time information, spatial location, environmental state and user identity identifier;

[0084] Step S20: Based on the context information, construct a control context graph of the current user. The control context graph is a weighted graph structure, where the nodes represent the historical control intentions of the user in the corresponding context, and the edges represent the correlation between different time-sequence control intentions;

[0085] Step S30: In the user behavior portrait space, retrieve the set of neighboring users with similar characteristics to the current user behavior portrait, and extract the control parameter records of the neighboring users in the context similar to the current user;

[0086] Step S40: Weightedly fuse the historical control parameters of the current user and the control parameters of the neighboring users to generate target control parameters;

[0087] Step S50: Complete the target control parameters in the control instruction to form a complete control command, and send the complete control command to the target smart home device to perform a control operation;

[0088] Step S60: After the control operation is completed, update the control context map based on user feedback or system monitoring results.

[0089] To address the possible problem of missing control parameters when the user interacts with a smart home device, the system needs to first determine whether the instruction content contains necessary parameter information after receiving a control instruction. When the system identifies that the instruction input by the user has missing parameters, it needs to further obtain the context information of the current user to provide semantic and environmental basis support for subsequent parameter completion and control intention reasoning. Context information can reflect the user's usage scenario and preference characteristics and is a key factor in achieving personalized and context-aware control responses.

[0090] Among them, context information refers to the environmental and identity-related elements that reflect the current control situation of the user, usually including four categories: time information, spatial location, environmental status, and user identity identifier. Among them, time information includes the time point and time period when the control occurs, the spatial location refers to the area where the user is in the home, the environmental status includes the data collected by sensors such as the current indoor temperature, humidity, and light, and the user identity identifier is used to distinguish the control habit differences between different family members and is the basis for realizing personalized control.

[0091] To implement the above functions, further, in a specific implementation, step S10 specifically includes the following sub-steps:

[0092] Step S101: Receive the control instruction input by the user. The system can receive control instruction information through methods such as a voice recognition module, a mobile terminal input module, or a smart panel input module, and parse it into structured instruction data. The instruction data includes at least an operation object, an operation type, and a parameter field. Among them, if the parameter field is empty or is a fuzzy expression (such as "turn on the air conditioner", "make the light brighter"), it is determined that the control parameter is missing.

[0093] Step S102: Identify whether there is a missing control parameter in the control instruction. The system performs field matching on the current instruction structure through a preset control template library. If a required parameter field for a device control template is missing or a valid value cannot be recognized, a parameter missing identification signal is generated. The template library can pre-store the control parameter structure requirements of different smart devices. For example, an air conditioner needs to include fields such as a temperature value, a mode type, and a wind speed level.

[0094] Step S103: Extract the context information of the current user. After the system determines that a parameter is missing, it immediately retrieves the context information related to the current user for subsequent reasoning and parameter completion. The extraction of context information specifically includes the following sub-items:

[0095] Step S1031: Extract time information. The system obtains the current timestamp through the local time service module or the unified time management module, and classifies the timestamp into the corresponding time period label according to the preset time period division rules (such as morning, daytime, evening, night).

[0096] Step S1032: Extract spatial location information. The system determines the room location where the user is currently located through the positioning module. Optional implementation schemes include: automatically determining the room area based on the bound location of the device controlled by the user, or using the Bluetooth beacon signal of the wearable device bound to the user for positioning, or identifying the room through an indoor camera combined with image recognition.

[0097] Step S1033: Obtain environmental status information. The system obtains the current environmental parameter values in real time through modules such as temperature and humidity sensors, light sensors, and carbon dioxide sensors deployed in the home space, and performs normalization and labeling processing. For example, when the room temperature is greater than 28 degrees Celsius, it can be classified as the "relatively hot" label, and when the light is lower than a certain threshold, it can be classified as the "relatively dim light" label.

[0098] Step S1034: Obtain the user identity identifier. The system identifies the identity of the user who issues the current command through the voiceprint characteristics of voice recognition, the face information of image recognition, the identity ID of the wearable device, etc. Optional implementation schemes include: using the voiceprint recognition module to compare user identities, or automatically identifying the bound user through the family member account system, or using the user login information of the terminal device for determination.

[0099] Through the above steps, the system can obtain a complete context information structure including time period, room location, environmental status label, and user identity label, which is used as the input condition for the subsequent semantic reasoning and parameter completion module.

[0100] The advantage of this step design is that it can provide key support for the completion of fuzzy commands and accurate responses by extracting the current control scenario context of the user in real time. Compared with the traditional method of using default parameters for response, this method realizes a comprehensive understanding of "who, when, where, and in what environment" the control command is issued, which is beneficial to improving the personalization, adaptability, and intelligent response ability of the system, and avoiding control failures or experience degradation caused by inconsistent default settings and user expectations.

[0101] In a specific example, when the user says the command "Turn on the air conditioner" by voice in the bedroom at 21:00 in the evening, the system recognizes that the command lacks parameter information such as the target temperature, and then executes this step to extract the context information, obtaining the comprehensive context information that the current time is at night, the location is the bedroom, the ambient temperature is 28 degrees Celsius, and the user is User_7. This context information will be used in subsequent steps to reason about and complete the temperature parameter of the "Turn on the air conditioner" command.

[0102] To accurately complete the missing parameter information in the user's control command, the system not only needs to perceive the current context state, but also needs to perform intention reasoning and behavior prediction based on the user's past control behaviors in similar or related contexts. For this purpose, it is necessary to construct a graph structure based on the user's historical control behaviors and with the ability to express context semantics. Through the association relationships between the nodes and edges in the graph, the temporal and logical connections between different control intentions in a specific context are revealed, so as to provide a basis for completing fuzzy commands. This graph structure can capture the laws of the evolution of the user's behavior over time and scenarios, and is an important basis for supporting the intelligent home system to implement personalized control strategies.

[0103] The control context graph refers to a graph structure constructed based on the user's historical control behaviors, which is used to express the distribution and evolution path of the user's control intentions under various context conditions. Each node in the graph represents a specific control intention, which occurs under a combination of a specific time period, spatial location, and environmental state. The edges in the graph represent the association relationships between the control behaviors that occur continuously in adjacent times or similar contexts. The edges can carry weights, which are used to represent the frequency, similarity, or probabilistic intensity between behaviors. Through this weighted graph structure, a semantic association link between control behaviors can be established, thus providing a reference path for the parameter reasoning of the current command.

[0104] To achieve the above purpose, in a specific implementation, step S20 includes the following sub-steps:

[0105] Step S201, collect and normalize the user's historical control behavior records.

[0106] The system extracts the historical control behaviors of a preset time window (such as the recent 30 days) from the control logs of the user on the intelligent home platform. Each control behavior includes, but is not limited to, the following fields: timestamp, device name, device operation type, parameter value, execution location, user identity, environmental state, etc.

[0107] For example, a certain user controls the "bedroom air conditioner to be set to 23 degrees" by voice at "21:04 on April 10, 2025", and the system will record the following structured data:

[0108] Time: 21:04;

[0109] Controlled device: Air conditioner;

[0110] Operation type: Turn on;

[0111] Parameter value: Temperature 23 degrees;

[0112] Room: Bedroom;

[0113] User ID: User_7;

[0114] Ambient temperature: 28 degrees Celsius.

[0115] After the data is standardized in format, it is used as the input source data for the graph.

[0116] Step S202, construct graph nodes and extract context tags.

[0117] The system abstracts each control intention with complete parameters issued by the user in a specific context as a node in the graph. Each node has the following attributes:

[0118] Controlled target device (such as air conditioner);

[0119] Control parameter (such as 23 degrees);

[0120] Context tag set, including:

[0121] Time tag (such as night);

[0122] Space tag (such as bedroom);

[0123] Environment tag (such as relatively hot);

[0124] User identity tag (such as User_7);

[0125] For example, node N1 represents "User_7 sets the air conditioner to 23 degrees in the bedroom at night", and the node attributes include:

[0126] N1. Device: Air conditioner;

[0127] N1. Operation: Turn on;

[0128] N1. Parameter: 23 degrees;

[0129] N1. Context: {night, bedroom, relatively hot, User_7}.

[0130] If similar control behaviors occur repeatedly, the system can add a counting weight to this node to represent the frequency of occurrence of this control intention in this context.

[0131] Step S203, establish edge relationships between nodes to represent temporal correlations.

[0132] The system analyzes the consecutive control behaviors of the user in chronological order. If the time interval between two control intentions is less than a preset threshold (e.g., 10 minutes), and there is context overlap or linkage logic between the behaviors, a directed edge is established between the corresponding two nodes. The direction of the edge represents the order of occurrence of the behaviors, and the weight of the edge can represent their co-occurrence frequency or context similarity.

[0133] The correlation between different sequential control intentions is mainly reflected in the following categories:

[0134] Continuous operation mode, such as "turn on the air conditioner" followed by "draw the curtains";

[0135] Time coincidence mode, such as the device switching behavior that is fixed at 21:00 every night;

[0136] Linkage control mode, such as automatically starting the background music after dimming the lights.

[0137] For example:

[0138] If node N1 is "bedroom at night, turn on the air conditioner to 23 degrees";

[0139] Node N2 is "bedroom at night, draw the curtains";

[0140] And these two types of behaviors have occurred continuously more than 15 times in the past 30 days;

[0141] Then the system establishes edge E12, representing N1 → N2, and the weight of the edge is 15.

[0142] This weight can be used as the basis for the priority of behavior recommendation.

[0143] Step S204, the graph structure is completed and persistently stored.

[0144] After completing the construction of nodes and edges, the system organizes the graph structure into a graph database or a graph index structure, and persistently stores the attribute set of nodes, the weight parameters of edges, and the context label index respectively.

[0145] Optional implementation solutions include: using the graph database Neo4j to construct the graph structure, or using a JSON-based index structure to store each node and its outgoing edge information for quick retrieval and reasoning.

[0146] In this step, the control context graph is used to represent the user's control behavior in the form of a graph structure, which is different from the traditional behavior data stored linearly by time or defined in the form of a rule list. By using nodes to represent control intentions and edges to represent the association paths between behaviors, a structured, traversable, and computable behavior graph is formed. This enables the system to not only determine "what the user has done" in a certain scenario but also judge "what the user is likely to do after that", providing a basis for behavior chain prediction and reasoning.

[0147] Each node in the graph carries complete context tags, including information such as time period, spatial location, environmental state, and user identity, tightly binding the control behavior to its occurrence context, thus enabling context-based behavior screening, similarity matching, and habit deduction. After the user issues a fuzzy instruction, the system can retrieve the "most likely matching node" in the graph based on the current context to achieve natural and fitting parameter completion.

[0148] By introducing an edge structure in the graph, the system can record the behavior sequence of the user's consecutive operations at different times. Combining the weights of the edges can reflect the frequency, stability, and evolution trend of the behaviors. This path structure enables the system to recommend the next possible action after judging the current operation, thereby achieving chain control recommendation, linkage response, and pre-judgment decision-making.

[0149] Through the above steps, the system can construct a control context graph with the ability to structurally express behavior context, which can not only reveal the distribution of the user's control intentions in different situations but also express the temporal logic relationship between the user's behaviors. This graph provides an inferable path for completing the missing control parameters in practical applications, improving the accuracy of control response and the personalized adaptation ability.

[0150] In a complete specific example:

[0151] Taking the user User_7 as an example, in the recent 30 days, the user has performed the following control behaviors in the "bedroom" "after 21:00 at night" multiple times:

[0152] "Turn on the air conditioner to 23 degrees";

[0153] "Draw the curtains";

[0154] "Turn off the main light";

[0155] "Turn on the bedside lamp".

[0156] The control context graph constructed by the system includes the following nodes:

[0157] N1: "Bedroom at night, turn on the air conditioner to 23 degrees";

[0158] N2: "Bedroom at night, draw the curtains";

[0159] N3: "Turn off the main light in the bedroom at night";

[0160] N4: "Turn on the bedside lamp in the bedroom at night".

[0161] And the following edges:

[0162] E12: N1 → N2, indicating that the curtain is drawn immediately after the air conditioner is turned on, with a weight of 18;

[0163] E23: N2 → N3, indicating that the main light is usually turned off after the curtain is drawn, with a weight of 17;

[0164] E34: N3 → N4, indicating that the bedside lamp is turned on after the main light is turned off, with a weight of 16.

[0165] When the user only says the vague instruction "Turn on the air conditioner" in the current night context, the system identifies the closest node as N1 in the context graph, and determines the complemented temperature parameter as 23 degrees based on the context of N1 and the high-frequency behavior path, thus completing the automatic complementation of the control instruction. Subsequently, the system can also automatically suggest a linked control sequence according to the path N1 → N2 → N3 → N4 to improve the user experience.

[0166] In the process of smart home interactive control, in the case of missing parameters in the control instructions issued by the user, simply relying on the historical behavior of the current user may lead to inaccurate complementation due to insufficient samples or sparse context. To improve the reliability and personalized adaptability of the complementation, the system introduces the concept of "neighboring users" on this basis, that is, to find other user groups with similar behavior characteristics to the current user in the behavior feature space, and retrieve the control parameter records of these neighboring users in similar contexts. By referring to the behavior preferences of similar users in similar scenarios, the coverage range of behavior samples can be effectively expanded, and the accuracy and robustness of parameter complementation can be improved.

[0167] In this step, the user behavior portrait space refers to a multi-dimensional feature space constructed based on the control behavior sequence of the user in a specific time period, spatial location, and environmental state, which is used to express features such as the user's control style, preference type, and behavior periodicity. The neighboring user set refers to other user sets in this behavior portrait space whose portrait feature distances from the current user are within a certain threshold range, and these users are considered to have similar control modes or usage habits. The control parameter record refers to the parameter values used by neighboring users to control devices in a context environment similar to that of the current user, such as temperature setting values, brightness levels, or volume values.

[0168] In a specific implementation, step S30 includes the following sub-steps:

[0169] Step S301, construct a behavior portrait feature vector.

[0170] Based on the node features and behavior statistical information involved by the current user in the control context map, the system constructs a user behavior portrait feature vector, and the feature dimensions include but are not limited to the following:

[0171] The device usage frequency of the user in different time periods;

[0172] The control preference in different spatial positions;

[0173] The centralized distribution value of control parameters;

[0174] The pattern number of common behavior sequences.

[0175] Optional implementation schemes include: using a behavior frequency vector based on statistics, or using a graph neural network to embed the nodes of the map to obtain a dense representation vector. The system can represent the current user behavior portrait vector as:

[0176] Vector U = {f1, f2, ..., f n};

[0177] Where f1 to f n represent n-dimensional behavior features, including numerical values such as time preference, spatial preference, and parameter distribution center.

[0178] Step S302, calculate the user portrait similarity and select neighboring users.

[0179] In the user behavior portrait space, the system calculates the similarity between the current user portrait vector U and all other user portrait vectors. Optional implementation schemes include:

[0180] Using the Euclidean distance to measure the spatial distance between portraits;

[0181] Using the cosine similarity to measure the consistency of control styles;

[0182] Using the Mahalanobis distance to consider the distribution correlation.

[0183] Let U be the current user portrait vector and V be a neighboring candidate user portrait vector. Define the similarity function as Sim(U, V). If Sim(U, V) is greater than the set threshold θ, then the user corresponding to V is considered a neighboring user.

[0184] Where:

[0185] U and V represent user portrait vectors;

[0186] Sim(U, V) represents the similarity;

[0187] θ is the neighborhood determination threshold;

[0188] N(U) represents the set of all users who satisfy Sim(U, V) ≥ θ.

[0189] The system selects the top K users that meet the conditions to form the neighborhood user set N(U), where K is the neighborhood capacity parameter set by the system.

[0190] Step S303: Match similar contexts and extract control parameter records.

[0191] For each user in the neighborhood user set N(U), the system filters out the nodes with relatively high similarity to the current user's context tags (including time period, spatial location, environmental status) from their control context graphs, and extracts the control parameters used from these nodes.

[0192] The similar context matching rule can determine similarity based on the label matching score of context tags, label embedding distance, or context classifier. The nodes that meet the conditions will be regarded as candidate nodes, and the control parameter values they carry will be used as the candidate set of recommended parameters. The system summarizes these parameter records to form a neighborhood control parameter sample set for subsequent fusion calculations.

[0193] The implementation of this step enables the system to break through the limitations of single-user behavior data, introduce a group behavior learning mechanism, and provide a "analogical decision-making" complementation method for the current user by finding similar users in the behavior portrait space and referring to their behavior decisions in similar scenarios. This method has the characteristics of rich samples, clear structure, and strong fusion ability, effectively improving the accuracy and diversity of parameter complementation, especially performing more excellently in cases where user behavior is novel, data is sparse, or control intentions are ambiguous.

[0194] In a specific example, continuing the previous example, the current user User_7 says the instruction "turn on the air conditioner" in the bedroom at 21:00 in the evening but does not specify the temperature. After the system obtains its current context information through steps S10 and S20 and constructs the control context graph, and determines that the historical behavior data is insufficient to form a definite recommendation, this step is executed:

[0195] In step S301, the system encodes User_7's air conditioner usage preferences in each time period and each room in the past 30 days into a vector U;

[0196] In step S302, the system searches for other users similar to User_7's portrait, such as User_15 and User_22, with a similarity greater than 0.85 in the portrait space;

[0197] In step S303, the system retrieves that the air conditioner temperatures set by User_15 and User_22 in the scenario of "bedroom at night with environmental temperature higher than 27°C" are 22 degrees and 23 degrees respectively.

[0198] Finally, the system records these neighborhood parameter records as a candidate set, which is used in the subsequent fusion module to infer that the current air conditioner temperature should be set to 22.5 degrees, thus completing the parameter complementation of the fuzzy instruction.

[0199] To improve the accuracy and personalized matching degree of fuzzy control instruction parameter complementation, the system not only needs to refer to the historical control parameters of the current user himself, but also needs to make full use of the control parameters extracted from neighboring users in similar contexts. Since the behavior habits of different users have certain differences, when using the data of neighboring users, it is necessary to reasonably integrate the individual preferences of the current user and the group behavior trends of neighboring users to avoid the problem of individual misjudgment caused by directly adopting neighboring behaviors. By setting a weighting mechanism to fuse the control parameters of the current user and neighboring users, the system can enhance the inference robustness under the condition of sparse samples while retaining personalized features, thus generating target control parameters that better conform to the current situation.

[0200] To achieve the above purpose, a specific implementation of step S40 includes:

[0201] Step S401, organize the historical control parameter samples of the current user.

[0202] The system extracts the control parameters set by the current user in the nodes that match or are similar to the current context conditions from the control context map, and conducts statistical analysis. This analysis includes the occurrence frequency of parameter values, the concentration degree of parameter intervals, and the similarity with the current context labels.

[0203] For example, if the user sets the air conditioner to 22 degrees three times and 23 degrees twice under the condition of "bedroom at night, slightly hot", the system will record these parameter values and their corresponding context similarity information as the self-behavior sample set.

[0204] Step S402, organize the control parameter samples of neighboring users.

[0205] The system summarizes all the control parameter records set by neighboring users in the neighboring user set under similar contexts, and attaches weight information to each parameter sample. The weights can be set based on the following factors:

[0206] The similarity between the neighboring user portrait and the current user portrait;

[0207] The matching degree between this control behavior and the current context;

[0208] The frequency of this control parameter appearing in the user map;

[0209] Optional implementation schemes include: setting a comprehensive weight factor for each parameter record, which is composed of the above indicators weighted.

[0210] Step S403: Perform weighted fusion to generate the target control parameter.

[0211] The system combines the two sample sets of the current user and neighboring users, and performs operations such as weighted averaging, weighted sorting, or voting to generate the final target control parameter according to the source type (own or neighboring) of each parameter sample and its corresponding weight coefficient.

[0212] Optional implementation solutions include:

[0213] Use the weighted average method to calculate the fusion result of continuous value parameters;

[0214] Use the weighted voting method to determine the recommended result of discrete value parameters;

[0215] Use the hierarchical priority strategy to preferentially use the current user's parameters when the own samples are sufficient, and enhance the influence of neighboring parameters when the samples are sparse.

[0216] The target control parameter finally output by the system will be merged into the original fuzzy control instruction as the complement part to form a complete control command.

[0217] By executing this step, the system can fully integrate individual experience and group wisdom to achieve the adaptive complement of control parameters. In the case of sufficient samples, the original preferences of the current user can be maintained, and in the case of sparse samples, reasonable reasoning can also be carried out with the reference values of neighboring users. The weighted fusion mechanism not only improves the personalization and credibility of parameter generation, but also enhances the intelligent level of the system to handle uncertain instructions.

[0218] In a specific example, User_7 issued a fuzzy instruction of "turn on the air conditioner" in the bedroom at 21:00 in the evening. The system has recognized that the target temperature parameter is missing in this instruction, and obtained the following parameter samples through the control context map and the behavior portrait space:

[0219] The historical setting values of the current user are 23 degrees for 2 times and 22 degrees for 1 time;

[0220] Neighboring users User_15 and User_22 set the air conditioner temperature to 22 degrees, 23 degrees, and 24 degrees respectively in a similar context. Among them, 22 degrees appears 2 times, 23 degrees appears 2 times, and 24 degrees appears 1 time;

[0221] When the system calculates the sample weights, the current user's samples are given higher weights due to high behavior consistency, and the neighboring samples are weighted according to similarity.

[0222] Finally, the system synthesizes the current user preference (biased towards 23 degrees) and the neighborhood sample distribution (concentrated between 22 and 24 degrees), determines the target control parameter to be 23 degrees, and complements this parameter into the "turn on the air conditioner" instruction to form a complete command "turn on the air conditioner, set the temperature to 23 degrees", which is then sent to the device for execution. The system also records the execution result of this time for subsequent graph update and behavior model iteration.

[0223] To achieve the closed-loop response of the user's fuzzy control instruction, after generating the target control parameter in the previous steps, the system needs to accurately complement this parameter into the user's original instruction to form a control command with a complete structure, and send this complete command to the corresponding smart home device to drive the actual execution.

[0224] In a specific implementation, step S50 includes the following sub-steps:

[0225] Step S501, construct the complete control command structure.

[0226] The system merges the target control parameter with the original control instruction in terms of structure to construct a complete control command that conforms to the device control interface protocol. This command structure includes but is not limited to the following fields:

[0227] Target device identifier, such as device ID, type, installation location, etc.;

[0228] Control action type, such as turn on, turn off, adjust, switch, etc.;

[0229] Complemented control parameter, such as temperature value, brightness percentage, volume size, etc.;

[0230] Command effective context information, such as current timestamp, issuing user identifier, etc.

[0231] Optional implementation solutions include: using a unified JSON control command format, or adopting the XML structure defined by the manufacturer's protocol, or generating a structured data object through the local control API format.

[0232] Step S502, establish a control channel and send the control command.

[0233] The system sends the complete control command to the target smart home device through the corresponding control channel according to the communication protocol and access method of the target device. The control channel can include the following methods:

[0234] Local local area network device control channel, through the home gateway or edge controller for protocol encapsulation and command issuance;

[0235] Cloud device control channel, through the cloud control platform to call the open interface of the device manufacturer to forward the command to the device;

[0236] Point-to-point communication channel, directly controlling devices through short-distance protocols such as Bluetooth and ZigBee.

[0237] After receiving the command, the device will parse the command field and execute the corresponding operation, and at the same time feedback the execution status to the system. The system can record information such as the time of this issuance, the execution response code, and the change of device status, which serves as the basic data for subsequent control graph update and execution quality evaluation.

[0238] In a specific example, User_7 input the fuzzy instruction "turn on the air conditioner" by voice in the bedroom at 21:00 in the evening. The system has recognized that the instruction lacks temperature parameters, and determines the target control parameter as 23 degrees through the context graph, neighborhood user portrait and weighted fusion mechanism. The system executes this step as follows:

[0239] In step S501, the system constructs a complete control command with the following format:

[0240] Device type: air conditioner;

[0241] Operation type: turn on;

[0242] Parameter value: set the temperature to 23 degrees;

[0243] Control location: bedroom;

[0244] Issuing user: User_7;

[0245] Time tag: night.

[0246] In step S502, the system sends this command to the bedroom air conditioner host through the gateway device in the home local area network using the local area network control protocol. After receiving the command, the device immediately turns on and sets the target temperature to 23 degrees. The device feedback status shows that the command has been successfully executed, and the system records this behavior for subsequent graph node update and control habit learning. Thus, the fuzzy instruction processing process is completed, forming a closed-loop control process, which improves the response intelligence of the system and user satisfaction.

[0247] In the smart home interaction control system, the control context graph is used to describe the control intention and behavior path of the user in a specific context, providing a decision-making basis for parameter completion and linkage control of fuzzy control instructions. To maintain the timeliness and accuracy of this graph, after completing the control operation, the system should dynamically update the graph according to user feedback or device status monitoring results. Graph update can not only be used to correct behavior deviations, but also adapt to changes in user behavior habits, enhance the system's adaptability to long-term preferences and short-term changes, and improve the accuracy of personalized control and the stability of behavior prediction.

[0248] To achieve the above purpose, in a specific implementation, step S60 specifically includes:

[0249] Step S601: Receive user feedback or collect system monitoring data.

[0250] After the control command is issued, the system actively waits for the user's feedback behavior or system state change within a preset time window. The sources of feedback information include:

[0251] The user confirms whether the operation is satisfactory through the App;

[0252] The user issues oral feedback commands, such as "Turn it down a bit", "No", etc.;

[0253] The user manually adjusts the device state within a short time, such as immediately changing the temperature or turning off the device;

[0254] Optional implementation solutions include: monitoring the time interval between the device's feedback status and the user's touch actions to determine whether there is a control rollback behavior.

[0255] All feedback and monitoring data will be uniformly encoded into control response evaluation results, which are divided into labels such as "Confirmed valid", "Unsatisfied", "Repeated adjustment", "Revoke execution", etc.

[0256] Step S602: Identify the relationship between the current control path and the feedback result.

[0257] Based on the context tags in the control command, the system locates the nodes and edges to which the current behavior path belongs in the control context graph, such as the node "Turn on the air conditioner to 23 degrees at night in the bedroom" and its previous nodes "Draw the curtains", "Turn off the main light", etc.

[0258] The system judges according to the feedback result:

[0259] If it is "Confirmed valid", then increase the weights of the current node and the front and back edges;

[0260] If it is "Revoke" or "Parameter manual adjustment", then mark the current parameter as a potential deviation parameter;

[0261] If it is "Repeated control" or "Timeout adjustment", then reduce the confidence of the node in this context.

[0262] The system can establish a mapping relationship table between the feedback result and the behavior correction for subsequent automatic maintenance of the graph.

[0263] Step S603: Update the weights of the nodes and edges in the control context graph.

[0264] Based on the feedback relationship judged in step S602, the system performs the following dynamic update of the graph structure:

[0265] If the control parameter represented by the node is used or confirmed again by the user, increase the node weight;

[0266] If a new combination is formed in the behavior path, increase the occurrence frequency of the edges between nodes and update the edge weights;

[0267] If an unsatisfactory or rollback behavior occurs in the original parameters, reduce the node weights or add alternative parameter candidates;

[0268] Optional implementation solutions include: using a sliding window mechanism to only retain the latest behavior distribution for updating, or introducing a time decay coefficient to handle historical weights.

[0269] The updated graph will take effect immediately and be used as the reasoning basis for subsequent fuzzy control instructions.

[0270] Through this step, the system establishes a complete closed-loop update mechanism for the control context graph, which can not only continuously learn the user's control preferences, but also adaptively adjust the reasoning strategy and behavior judgment logic. This mechanism helps the system maintain a high response accuracy and personality matching ability in the long term, improves the overall intelligent control quality, and has significant advantages especially in scenarios where user preferences change frequently or multiple family members use the system.

[0271] In a specific example, User_7 uttered the fuzzy instruction "Turn on the air conditioner" in the bedroom at 21:00 in the evening. The system completed the temperature as 23 degrees and successfully issued a control command. After the control was executed, the system entered this step to execute the following process:

[0272] In step S601, the system detected that User_7 manually adjusted the air conditioner temperature to 22 degrees through the App within about 20 seconds after the control was executed, and determined it as a "parameter manual adjustment" feedback.

[0273] In step S602, the system traced back to the node "Bedroom, Night, Turn on the air conditioner, 23 degrees" that was just executed, marked this node as "parameters actively corrected", and recorded it as a deviation execution.

[0274] In step S603, the system reduced the weight of this node, and at the same time increased the weight of the node "Bedroom, Night, Turn on the air conditioner, 22 degrees" in the same context. If the system receives a similar fuzzy instruction again later, it is more likely to infer that the set temperature is 22 degrees rather than 23 degrees.

[0275] This example demonstrates the adaptive feature of the graph update mechanism. The system continuously optimizes the control parameter selection path through real-time learning and fine-tuning, and continuously conforms to the user's true intention.

[0276] As Figure 2 shown, in another embodiment, the present invention also provides an interactive control system based on smart home devices, including:

[0277] An input recognition module, configured to receive a control instruction input by a user and identify whether there is a missing control parameter in the control instruction;

[0278] A context extraction module, configured to extract context information of the current user when it is identified that there is a missing control parameter, where the context information includes time information, spatial location, environmental status, and user identity identifier;

[0279] A graph construction module, configured to construct a control context graph of the current user based on the context information, where the control context graph is a weighted graph structure, in which nodes represent the historical control intentions of the user in the corresponding context, and edges represent the correlation between control intentions at different time sequences;

[0280] A similarity retrieval module, configured to retrieve a set of neighboring users with similar characteristics to the current user behavior portrait in the user behavior portrait space, and extract the control parameter records of the neighboring users in the context similar to that of the current user;

[0281] A fusion calculation module, configured to perform weighted fusion on the historical control parameters of the current user and the control parameters of the neighboring users to generate target control parameters;

[0282] An instruction generation module, configured to supplement the target control parameters into the control instruction to form a complete control command, and send the complete control command to a target smart home device to perform a control operation;

[0283] A graph update module, configured to update the control context graph based on user feedback or system monitoring results after the control operation is completed.

[0284] In a further implementation, the context extraction module includes:

[0285] A time extraction unit, configured to obtain the current time through a local clock or a network clock and match it to a preset time period label;

[0286] A location recognition unit, configured to determine the spatial location where the user is located through the physical location information of the user terminal, Bluetooth beacon, or communication information of a wearable device;

[0287] An environment perception unit, configured to collect environmental status information through deployed temperature and humidity sensors, light sensors, or air quality sensors;

[0288] An identity recognition unit, configured to determine the issuer of the current control instruction through voiceprint recognition, face recognition, or user device identification information.

[0289] In a further implementation, the graph construction module includes:

[0290] A behavior aggregation unit, configured to aggregate historical control behaviors with the same context label and control intention into the same node;

[0291] A path construction unit, configured to establish edge relationships between adjacent nodes according to the order of behavior occurrence, where the edge records the time sequence and co-occurrence frequency of control behaviors;

[0292] A weight assignment unit, configured to assign weights to edges based on node occurrence frequency, context similarity, and behavior transition probability, for representing the correlation between different time-sequence control intentions;

[0293] A graph spectrum output unit, configured to output a control context graph spectrum structure with nodes and weighted edges.

[0294] In a further implementation, the similarity retrieval module includes:

[0295] A portrait encoding unit, configured to encode the behavior pattern of the current user in the control context graph spectrum into a behavior portrait feature vector;

[0296] A similarity calculation unit, configured to calculate the similarity between the behavior portrait of the current user and other user portraits, and determine users with a similarity higher than a set threshold as neighborhood users;

[0297] A sample extraction unit, configured to extract control parameter values from the control behaviors of neighborhood users under similar context conditions, and record them together with their context labels as a neighborhood parameter sample set.

[0298] In a further implementation, the fusion calculation module includes:

[0299] A sample arrangement unit, configured to merge the historical control parameters of the current user and the control parameters of neighborhood users into a candidate parameter sample set;

[0300] A weight calculation unit, configured to determine the fusion weights of each parameter sample according to the behavior frequency of the current user sample, the similarity between the neighborhood user and the current user portrait, and the context matching degree;

[0301] A parameter generation unit, configured to perform weighted average, sorting, or screening on the candidate parameter samples according to the weighting rule, and generate target control parameters to complete the original control instruction.

[0302] It should be noted that the foregoing explanatory description of the embodiment of the interactive control method based on smart home devices also applies to the device of the embodiment of the present application, and will not be elaborated here.

[0303] Those of ordinary skill in the art will realize that the various units and algorithm steps described in the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0304] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0305] In several embodiments provided in this application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The foregoing storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memory (hereinafter referred to as ROM), random access memory (hereinafter referred to as RAM), magnetic disks, or optical discs.

[0306] The above is only the specific implementation manner of this application. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application and should be covered by the protection scope of this application. For the part of the module structure that is not specifically defined in this invention, it shall be subject to the content recorded in the prior art. The prior art mentioned in the foregoing background art part and specific embodiment part of this invention can be used as a part of this invention to understand the meaning of some technical features or parameters.

Claims

1. An interactive control method based on smart home devices, characterized in that, It includes the following steps: Step S10: Receive a control instruction input by the user, and identify whether there is a missing control parameter in the control instruction; When there is a missing control parameter, extract the context information of the current user, where the context information includes time information, spatial location, environmental status, and user identity identifier; Step S20: Based on the context information, construct a control context map for the current user. The control context map is a weighted graph structure, where nodes represent the historical control intentions of the user in the corresponding context, and edges represent the correlation between control intentions at different time sequences; Step S30: In the user behavior portrait space, retrieve a set of neighboring users whose behavior portrait features are similar to those of the current user, and extract the control parameter records of the neighboring users in the context similar to that of the current user; Step S40: Weightedly fuse the historical control parameters of the current user with the control parameters of the neighboring users to generate target control parameters; Step S50: Complete the target control parameters to the control instruction to form a complete control command, and send the complete control command to the target smart home device to execute a control operation; Step S60: After the control operation is completed, update the control context map based on user feedback or system monitoring results.

2. The interactive control method based on the smart home device according to claim 1, characterized in that, The step S10 includes the following sub-steps: Obtain the voice input, graphical interface input, or sensor trigger event of the current user, and parse it into a control intention instruction; Based on whether the control intention contains explicit control parameters, perform parameter field integrity identification; When it is identified that there is a missing control parameter, extract the context information of the current user, where: The time information obtains the current time through the local clock or network clock and matches it to the preset time period label; The spatial location performs room positioning through the physical location information of the user operation terminal, Bluetooth beacon, or wearable device bound to the user; The environmental status is obtained through temperature and humidity sensors, light sensors, or air quality sensors deployed in the room; The user identity identifier is determined through voiceprint recognition, face recognition, or user device ID recognition methods.

3. The interactive control method based on smart home devices according to claim 1, characterized in that The step S20 includes the following sub-steps: Based on the historical control behavior records of the current user, extract the context label and control parameter of each control behavior; Aggregate the behaviors with the same control intention under the same context label into nodes, and record the control device type, control parameter, context label, and behavior frequency in the nodes; Connect the continuously occurring control behaviors in chronological order to form an edge relationship, where the edge represents the behavior path of the front and back control actions; The edges between nodes are assigned edge weights according to the co-occurrence frequency, time interval, and context similarity of the user control behaviors; The constructed control context map is a weighted directed graph structure, which is used to express the evolution trend of the user control intention in different behavior paths.

4. The interactive control method based on smart home devices according to claim 1, characterized in that The step S30 includes the following sub-steps: Encode the behavior pattern of the current user in the control context map into a behavior portrait feature vector; In the user behavior portrait space, use the behavior feature similarity calculation method to identify users with similar behavior patterns; When the similarity between behavior portraits is higher than the set threshold, determine that the user is a neighborhood user; Filter the control behaviors of neighborhood users in similar time periods, spatial positions, and environmental states, and extract the corresponding control parameter values; Record the filtered control parameter values together with the context tags as a neighborhood sample set.

5. The interactive control method based on the smart home device according to claim 1, wherein, The step S40 includes the following sub-steps: Merge the historical control parameter samples of the current user in similar contexts with the control parameter samples of neighborhood users to form a candidate parameter set; For each control parameter sample, set its fusion weight, and the weight is determined by the following factors: The occurrence frequency of the behavior of the current user parameter sample; The similarity between the neighborhood user portrait and the current user portrait; The matching degree of the sample with the current context tag; According to the weighting rule, sort, filter, or perform weighted averaging on the candidate parameter samples to generate the target control parameter; The target control parameter is used to complete the missing parameter part in the user control instruction to form a complete control command.

6. An interactive control system based on smart home devices, characterized in that, The system includes the following modules: An input recognition module, which is used to receive the control instruction input by the user and identify whether there are missing control parameters in the control instruction; A context extraction module, which is used to extract the context information of the current user when it is recognized that there are missing control parameters. The context information includes time information, spatial position, environmental state, and user identity identifier; A graph construction module, which is used to construct a control context graph of the current user based on the context information. The control context graph is a weighted graph structure, where nodes represent the historical control intentions of the user in the corresponding context, and edges represent the correlation between control intentions at different time sequences; A similarity retrieval module, which is used to retrieve a set of neighborhood users whose behavior portrait features are similar to those of the current user in the user behavior portrait space, and extract the control parameter records of the neighborhood users in the context similar to that of the current user; A fusion calculation module, which is used to perform weighted fusion on the historical control parameters of the current user and the control parameters of the neighborhood users to generate the target control parameter; An instruction generation module, which is used to complete the target control parameter into the control instruction to form a complete control command, and send the complete control command to the target smart home device to perform a control operation; A graph update module, which is used to update the control context graph based on user feedback or system monitoring results after the control operation is completed.

7. The interactive control system based on smart home devices according to claim 6, wherein The context extraction module includes: A time extraction unit, which is used to obtain the current time through the local clock or network clock and match it to the preset time period label; A position recognition unit, which is used to determine the spatial position where the user is located through the physical position information of the user terminal, Bluetooth beacon, or communication information of the wearable device; An environment perception unit, which is used to collect environmental state information through deployed temperature and humidity sensors, light sensors, or air quality sensors; An identity recognition unit, which is used to judge the issuer of the current control instruction through voiceprint recognition, face recognition, or user device identification information.

8. The interactive control system based on smart home devices according to claim 6, wherein The graph construction module includes: A behavior aggregation unit, which is used to aggregate historical control behaviors with the same context tag and control intention into the same node; A path construction unit, which is used to establish edge relationships between adjacent nodes in the order of behavior occurrence, and the edges record the time sequence and co-occurrence frequency of control behaviors; A weight assignment unit, which is used to assign weights to the edges according to the node occurrence frequency, context similarity and behavior transition probability, and is used to represent the correlation between different time-sequence control intentions; A graph spectrum output unit, which is used to output a control context graph spectrum structure with nodes and weighted edges.

9. The interactive control system based on smart home devices according to claim 6, wherein The similarity retrieval module includes: A portrait encoding unit, which is used to encode the behavior pattern of the current user in the control context graph spectrum into a behavior portrait feature vector; A similarity calculation unit, which is used to calculate the similarity between the behavior portrait of the current user and other user portraits, and determine the users with similarity higher than the set threshold as neighborhood users; A sample extraction unit, which is used to extract control parameter values from the control behaviors of neighborhood users under similar context conditions, and record them together with their context labels as a neighborhood parameter sample set.

10. The interactive control system based on smart home devices according to claim 6, characterized in that, The fusion calculation module includes: A sample sorting unit, which is used to merge the historical control parameters of the current user and the control parameters of neighborhood users into a candidate parameter sample set; A weight calculation unit, which is used to determine the fusion weights of each parameter sample according to the behavior frequency of the current user sample, the similarity between the neighborhood user and the current user portrait, and the context matching degree; A parameter generation unit, which is used to perform weighted average, sorting or screening on the candidate parameter samples according to the weighting rules to generate target control parameters to complement the original control instructions.

Images