An AI-driven cross-device semantic understanding home interaction method

Abstract

The application relates to the technical field of smart homes, in particular to a home interaction method for AI-driven cross-device semantic understanding, which comprises the following steps: firstly, acquiring a user initial instruction and extracting environment control type reference data; secondly, identifying reference words in subsequent cross-space instructions, tracing continuous intentions and updating and adjusting targets; thirdly, combining device sensor data and a historical sharing mechanism to determine an execution instruction; fourthly, synchronizing multiple device states to generate a global interaction log, complementing missing information to form a complete intention chain; fifthly, correcting system semantic understanding deviation; sixthly, broadcasting interaction confirmation based on a demand satisfaction index; and finally, iteratively optimizing a device state library and an interaction framework through user experience data. With the aid of AI semantic analysis, anaphora resolution and intention continuity algorithms, cross-device context sharing and accurate intention transmission are realized, the coherence, accuracy and intelligent level of home interaction are significantly improved, and a seamless cross-space interaction experience is provided for users.

Description

Technical Field

[0001] This invention belongs to the field of smart home technology, specifically an AI-driven cross-device semantic understanding home interaction method. Background Technology

[0002] In the smart home field, with the rapid development of IoT and AI technologies, the number of smart devices deployed in home environments continues to grow, and users' demand for seamless cross-device interaction is becoming increasingly urgent. Building a system capable of coherent and natural interaction across different spaces and types of devices has become a key objective in improving home comfort, convenience, and intelligence. While current mainstream smart home systems have achieved basic voice control, device linkage, and scene automation functions, they still have significant limitations in handling continuous interactions across devices and spaces. Especially when users move between different rooms and continuously express their intentions, existing systems often lack the global perception of user dialogue history and contextual intent, leading to misunderstandings or even failures in understanding subsequent commands. For example, if a user activates environmental control in the living room using the voice command "turn up the temperature," and then enters the bedroom and issues the command "turn it up a little higher," if the bedroom device cannot recognize that this expression is a continuation of the previous command, it may misjudge it as an independent operation or completely ignore it, thus failing to accurately meet the user's true needs. The root cause of this problem lies in the lack of a unified mechanism for recording and transmitting intent in existing technologies. Devices struggle to share interaction states and semantic context in real time, leading to information gaps, intent breaks, or execution misalignments during multi-device collaboration. Furthermore, traditional systems typically employ static rules or isolated local semantic parsing models, making it difficult to dynamically adapt to the continuity of user behavior and changes in environmental states, further limiting their practicality and reliability in complex and dynamic home scenarios. Therefore, there is an urgent need for a novel home interaction method driven by AI, possessing cross-device semantic understanding and intent continuity maintenance capabilities. This would enable accurate capture of user intent, context awareness, cross-domain transmission, and adaptive correction, thereby constructing a truly coherent, intelligent, and reliable whole-house interactive experience. Summary of the Invention

[0003] The purpose of this invention is to address the aforementioned shortcomings in the prior art by providing an AI-driven, cross-device semantic understanding method for home interaction.

[0004] The objective of this invention is achieved through the following technical solution: an AI-driven cross-device semantic understanding home interaction method, comprising the following steps: S1. Obtain the user's initial command content on the living room device, extract the temperature adjustment intention from the initial command content, determine the intention type as environmental control, and obtain the initial adjustment value as the reference data by matching the intention type with the preset device status library. S2. Based on the baseline data and the user's subsequent instructions when moving to the bedroom, determine whether there are quoted words in the subsequent instructions. If there are quoted words, obtain the aforementioned initial adjustment value from the dialogue history, determine that the continuous intention is to further modify the temperature, and obtain the updated adjustment target. S3. Based on the updated adjustment target, the intent continuity algorithm is used to process the current environmental data from the bedroom device sensor. It is determined whether the current environmental data is lower than the preset threshold. If it is lower than the preset threshold, the baseline data and the updated adjustment target are transmitted through the history sharing mechanism to determine that the final execution instruction is the temperature increment operation. S4. Based on the final execution instruction, activate the control module of the bedroom device, obtain feedback signals from the control module, determine whether the feedback signals meet the continuity requirements, and if they meet the continuity requirements, use a real-time interaction algorithm to synchronize the status of all multiple devices and obtain a unified record as a global interaction log. S5. Based on the global interaction log, extract relevant entries for cross-space movement, determine whether there is missing information in the entries, and if there is missing information, supplement the missing part through the transmission mechanism, determine that the supplemented log is a complete intent chain, and obtain the optimized system response sequence. S6. Based on the optimized system response sequence, the history sharing algorithm is used to analyze and obtain potential system deviations from the sequence. It is determined whether the deviations are due to errors in instruction understanding. If they are due to errors in instruction understanding, the sequence parameters are adjusted through the deviation correction process to obtain the corrected response output. S7. Based on the corrected response output, extract the user demand satisfaction index, determine whether the index reaches the preset threshold, and if it reaches the preset threshold, broadcast the output through the multi-device interaction channel to determine that the broadcast content is a continuous interactive confirmation, and obtain the final user experience data. S8. Based on the final user experience data, update the preset device state library, obtain the benchmark data for the next cycle from the updated library, determine whether the benchmark data needs further optimization, and if it needs further optimization, iterate through the intent continuity algorithm to obtain an enhanced cross-device interaction framework.

[0005] The present invention is further configured such that: the baseline data generation includes intent type identification, device state library matching, and initial adjustment value extraction; the updated adjustment target includes reference word detection, historical intent backtracking, and continuous intent determination; the final execution instruction includes environmental data comparison, historical sharing mechanism triggering, and operation type confirmation; the global interaction log includes multi-device state synchronization, interaction event marking, and timestamp binding; the optimized system response sequence includes intent chain integrity verification, missing information location, and context completion; the corrected response output includes deviation source identification, parameter dynamic correction, and output consistency verification; the final user experience data includes requirement satisfaction measurement, interaction fluency evaluation, and feedback loop construction; and the enhanced cross-device interaction framework includes dynamic state library updating, intent continuity enhancement, and adaptive optimization of interaction paths.

[0006] The present invention is further configured to obtain the user's initial command content on the living room device, extract the temperature adjustment intention from the initial command content, determine the intention type as environmental control, and obtain the initial adjustment value as reference data by matching the intention type with a preset device state library. The specific steps are as follows: The system acquires the user's initial voice or text commands on the living room device, uses semantic parsing technology to identify the temperature adjustment keywords contained therein, and generates preliminary intent fragments. Based on the preliminary intent fragment, combined with contextual analysis, it is determined whether the intent category it belongs to is environmental control, and an intent type label is generated; Based on the intent type tag, query the parameter configurations related to environmental control in the preset device status library, match the current operating status of the living room temperature control device, and generate the initial adjustment value; Based on the initial adjustment values, and combined with device capability boundary verification, ensure that it is within the executable range and generate baseline data for subsequent interactions.

[0007] The present invention is further configured such that, based on the baseline data and the user's subsequent instructions when moving to the bedroom, it determines whether there are quoted words in the subsequent instructions; if there are quoted words, it obtains the aforementioned initial adjustment value from the dialogue history, determines the continuous intent as further modifying the temperature, and obtains the updated adjustment target. The specific steps are as follows: The system receives subsequent instructions input by the user in the bedroom device, uses a substitution resolution technique to identify whether they contain quotations, and generates a quotation word determination result; based on the quotation word determination result; If a reference word exists, retrieve the most recent instruction record related to environmental control from the cross-device dialogue history cache and extract the corresponding initial adjustment value; Based on the initial adjustment value and the semantic direction of the current subsequent instruction, determine whether the user's intent is a continuous modification of the original adjustment value, and generate a continuous intent identifier; Based on the continuous intent identifier and combined with the incremental adjustment rules, a new target temperature value is calculated, and an updated adjustment target is generated.

[0008] The present invention is further configured such that, based on the updated adjustment target, an intent continuity algorithm is used to process and obtain current environmental data from the bedroom device sensor, determining whether the current environmental data is lower than a preset threshold. If it is lower than the preset threshold, the baseline data and the updated adjustment target are transmitted through a history sharing mechanism. The specific steps for determining that the final execution instruction is a temperature increment operation are as follows: The current environmental data is obtained from the temperature and humidity sensor built into the bedroom device. The current environmental data is compared with a preset comfort threshold. If it is lower than the threshold, the intent continuity verification process is triggered. Based on the intent continuity verification process, the baseline data of the living room stage is associated and mapped with the updated adjustment target of the bedroom stage through the historical sharing mechanism to generate a joint intent context. Based on the aforementioned joint intent context and combined with the device control logic, a clear temperature increment operation instruction is generated as the final execution instruction.

[0009] The present invention is further configured such that, based on the final execution instruction, the control module of the bedroom device is activated, a feedback signal is obtained from the control module, and it is determined whether the feedback signal meets the continuity requirement. If it meets the continuity requirement, a real-time interaction algorithm is used to synchronize the status of all multiple devices to obtain a unified record as a global interaction log. The specific steps are as follows: Send the final execution command to the control module of the bedroom thermostat to initiate the temperature adjustment action; The control module collects execution feedback signals in real time, including the actual rate of temperature change and the equipment response status. The feedback signal is compared with the expected continuous demand. If the deviation is within the allowable range, it is determined that the continuous demand is met. Based on the judgment result, the time, location, device, intent, and execution status of this interaction event are synchronized to all associated devices through a real-time interaction algorithm, generating a global interaction log with a unified identifier.

[0010] The present invention is further configured to extract relevant entries for cross-spatial movement based on the global interaction log, determine whether there is missing information in the entries, and if there is missing information, supplement the missing part through a transmission mechanism, determine that the supplemented log is a complete intent chain, and obtain the optimized system response sequence. The specific steps are as follows: Filter out interaction records involving user movement across space from the global interaction log to form a cross-space entry set; Perform a structural integrity check on the cross-space entry set to identify any missing information such as intentional breakpoints or unsynchronized states; If information is missing, the corresponding context data is extracted from the source device logs through the historical transmission mechanism between devices to fill in the missing fields; Based on the filled complete log, the user intent evolution path is reconstructed to generate a logically coherent optimized system response sequence.

[0011] The present invention is further configured such that, based on the optimized system response sequence, a history sharing algorithm is used to analyze and obtain potential system deviations from the sequence, and it is determined whether the deviations originate from instruction comprehension errors. If they originate from instruction comprehension errors, the sequence parameters are adjusted through a deviation correction process to obtain the corrected response output. The specific steps are as follows: The optimized system response sequence is backtracked to identify inconsistencies between the execution results and the user's original intent, and potential deviation markers are generated. Based on the potential deviation markers, and combined with the multi-turn interaction context, we analyze whether the deviation is caused by semantic misjudgment or referential confusion. If the deviation is determined to be due to an error in understanding the instruction, a deviation correction process is initiated to adjust key parameters in the response sequence, including the target value, the execution device, or the type of operation. The response output is regenerated based on the corrected parameters to ensure that it is consistent with the user's true intent, thus obtaining the corrected response output.

[0012] The present invention is further configured such that, based on the corrected response output, user demand satisfaction indicators are extracted, it is determined whether the indicators have reached a preset threshold, and if the preset threshold is reached, the results are broadcast through a multi-device interaction channel, and the broadcast content is determined to be a continuous interactive confirmation, thus obtaining the final user experience data. The specific steps are as follows: Extract user demand satisfaction metrics from the corrected response output, including goal achievement, response timeliness, and operational accuracy; The user demand satisfaction index is compared with the preset satisfaction threshold. If all indexes are met, a confirmation broadcast mechanism is triggered. A unified interaction confirmation message is pushed to all relevant devices through a multi-device interaction channel. The message includes information such as operation success, current status, and follow-up suggestions. The final user experience data is compiled based on users' implicit or explicit feedback on confirmation information.

[0013] The present invention is further configured to, based on the final user experience data, update a preset device state database, obtain baseline data for the next period from the updated database, determine whether the baseline data needs further optimization, and if further optimization is needed, obtain an enhanced cross-device interaction framework through iterative processing using an intent continuity algorithm. The specific steps are as follows: Inject the valid parameters from the final user experience data into the preset device state library to update the environmental control strategy and user preference model; Based on the updated device status database, new baseline data is automatically extracted at the start of the next interaction cycle; An adaptive evaluation is performed on the benchmark data to determine whether it can effectively support the intent continuity requirements in complex scenarios. If the evaluation results indicate that further optimization is needed, the intent continuity algorithm is invoked to iteratively refine the baseline data, generating an enhanced cross-device interaction framework with stronger generalization capabilities.

[0014] The beneficial effects of this invention are: I. This invention establishes a unified transmission channel for cross-device interaction context through a history sharing mechanism and referential resolution technology, enabling the system to accurately identify the continuous intent of users when moving between different rooms; it solves the problem of instruction gaps caused by isolated devices and lack of information sharing in the prior art, ensuring that the continuous intent of users is not lost or misunderstood in cross-space and cross-device scenarios.

[0015] Second, relying on the intent continuity algorithm, global interaction log integrity verification, and deviation correction mechanism, this invention can identify and correct semantic misjudgment, referential confusion, and other problems in real time. By dynamically adjusting the parameters of the response sequence, it ensures that the system execution result is highly consistent with the user's true intent. Compared with the traditional static rule parsing model, the accuracy of instruction understanding and execution correctness are significantly improved, reducing the user experience loss caused by erroneous operations.

[0016] Third, when users move across spaces, they do not need to repeat their initial intent. The system can automatically trace the context of historical commands and adapt to the current environment. This achieves a natural interaction mode where one intent is initiated and multiple devices respond seamlessly, greatly reducing the complexity of user operations. It is especially suitable for frequent cross-room scenarios in the home, improving the convenience and comfort of home control.

[0017] Fourth, this invention injects user experience data into the device state database in a closed loop, and iteratively optimizes the baseline data and interaction framework through an intent continuity algorithm. The system can dynamically adapt to different users' operating preferences and environmental changes, gradually forming a personalized interaction mode. Compared with traditional fixed rule systems, it has stronger generalization ability and scenario adaptability.

[0018] Fifth, this invention effectively avoids risks such as information gaps and execution misalignments during multi-device collaboration by completing the global interaction log, synchronizing device status in real time, and implementing deviation correction processes. Even in complex scenarios such as temporary device offline or ambiguous command descriptions, it can still ensure interaction continuity through context completion and historical data backtracking, thereby improving the operational stability and reliability of the smart home system. Attached Figure Description

[0019] The invention will be further illustrated with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the invention. For those skilled in the art, other drawings can be obtained based on the following drawings without any creative effort.

[0020] Figure 1 This is a schematic diagram of the present invention. Detailed Implementation

[0021] The present invention will be further described in conjunction with the following embodiments.

[0022] Depend on Figure 1 As can be seen, this embodiment provides an AI-driven cross-device semantic understanding home interaction method, including the following steps: S1. Obtain the user's initial command content on the living room device, extract the temperature adjustment intention from the initial command content, determine the intention type as environmental control, and obtain the initial adjustment value as the reference data by matching the intention type with the preset device status library. S2. Based on the baseline data and the user's subsequent instructions when moving to the bedroom, determine whether there are quoted words in the subsequent instructions. If there are quoted words, obtain the aforementioned initial adjustment value from the dialogue history, determine that the continuous intention is to further modify the temperature, and obtain the updated adjustment target. S3. Based on the updated adjustment target, the intent continuity algorithm is used to process the current environmental data from the bedroom device sensor. It is determined whether the current environmental data is lower than the preset threshold. If it is lower than the preset threshold, the baseline data and the updated adjustment target are transmitted through the history sharing mechanism to determine that the final execution instruction is the temperature increment operation. S4. Based on the final execution instruction, activate the control module of the bedroom device, obtain feedback signals from the control module, determine whether the feedback signals meet the continuity requirements, and if they meet the continuity requirements, use a real-time interaction algorithm to synchronize the status of all multiple devices and obtain a unified record as a global interaction log. S5. Based on the global interaction log, extract relevant entries for cross-space movement, determine whether there is missing information in the entries, and if there is missing information, supplement the missing part through the transmission mechanism, determine that the supplemented log is a complete intent chain, and obtain the optimized system response sequence. S6. Based on the optimized system response sequence, the history sharing algorithm is used to analyze and obtain potential system deviations from the sequence. It is determined whether the deviations are due to errors in instruction understanding. If they are due to errors in instruction understanding, the sequence parameters are adjusted through the deviation correction process to obtain the corrected response output. S7. Based on the corrected response output, extract the user demand satisfaction index, determine whether the index reaches the preset threshold, and if it reaches the preset threshold, broadcast the output through the multi-device interaction channel to determine that the broadcast content is a continuous interactive confirmation, and obtain the final user experience data. S8. Based on the final user experience data, update the preset device state library, obtain the benchmark data for the next cycle from the updated library, determine whether the benchmark data needs further optimization, and if it needs further optimization, iterate through the intent continuity algorithm to obtain an enhanced cross-device interaction framework.

[0023] This embodiment provides an AI-driven cross-device semantic understanding home interaction method. The baseline data generation includes intent type identification, device state library matching, and initial adjustment value extraction. The updated adjustment targets include reference word detection, historical intent backtracking, and continuous intent determination. The final execution instructions include environmental data comparison, historical sharing mechanism triggering, and operation type confirmation. The global interaction log includes multi-device state synchronization, interaction event marking, and timestamp binding. The optimized system response sequence includes intent chain integrity verification, missing information location, and context completion. The corrected response output includes deviation source identification, parameter dynamic correction, and output consistency verification. The final user experience data includes demand satisfaction measurement, interaction fluency evaluation, and feedback loop construction. The enhanced cross-device interaction framework includes dynamic state library updates, intent continuity enhancement, and adaptive optimization of interaction paths.

[0024] This embodiment provides an AI-driven cross-device semantic understanding home interaction method. The specific steps are as follows: First, the method obtains the user's initial command content on a living room device. Then, it extracts the temperature adjustment intent from the initial command content, determines the intent type as environmental control, and matches the intent type with a preset device state library to obtain an initial adjustment value as baseline data. The system acquires the user's initial voice or text commands on the living room device, uses semantic parsing technology to identify the temperature adjustment keywords contained therein, and generates preliminary intent fragments. Based on the preliminary intent fragment, combined with contextual analysis, it is determined whether the intent category it belongs to is environmental control, and an intent type label is generated; Based on the intent type tag, query the parameter configurations related to environmental control in the preset device status library, match the current operating status of the living room temperature control device, and generate the initial adjustment value; Based on the initial adjustment values, and combined with device capability boundary verification, ensure that it is within the executable range and generate baseline data for subsequent interactions.

[0025] This embodiment provides an AI-driven cross-device semantic understanding home interaction method. Based on the baseline data and subsequent instructions when the user moves to the bedroom, it determines whether there are quoted words in the subsequent instructions. If quoted words are present, the method retrieves the aforementioned initial adjustment value from the dialogue history, determines the continuous intent as further temperature modification, and obtains the updated adjustment target. The specific steps are as follows: The system receives subsequent commands input by the user in the bedroom device, uses a substitution resolution technique to identify whether they contain references such as "higher" or "keep raising," and generates a reference word determination result; based on the reference word determination result; If a reference word exists, retrieve the most recent instruction record related to environmental control from the cross-device dialogue history cache and extract the corresponding initial adjustment value; Based on the initial adjustment value and the semantic direction of the current subsequent instruction, determine whether the user's intent is a continuous modification of the original adjustment value, and generate a continuous intent identifier; Based on the continuous intent identifier and combined with the incremental adjustment rules, a new target temperature value is calculated, and an updated adjustment target is generated.

[0026] This embodiment provides an AI-driven cross-device semantic understanding home interaction method. Based on the updated adjustment target, it uses an intent continuity algorithm to process the current environmental data obtained from the bedroom device sensors, determines whether the current environmental data is lower than a preset threshold, and if it is lower than the preset threshold, transmits the baseline data and the updated adjustment target through a history sharing mechanism. The specific steps for determining the final execution command as a temperature increment operation are as follows: The current environmental data is obtained from the temperature and humidity sensor built into the bedroom device. The current environmental data is compared with a preset comfort threshold. If it is lower than the threshold, the intent continuity verification process is triggered. Based on the intent continuity verification process, the baseline data of the living room stage is associated and mapped with the updated adjustment target of the bedroom stage through the historical sharing mechanism to generate a joint intent context. Based on the aforementioned joint intent context and combined with the device control logic, a clear temperature increment operation instruction is generated as the final execution instruction.

[0027] This embodiment provides an AI-driven cross-device semantic understanding home interaction method. Based on the final execution command, the control module of the bedroom device is activated, feedback signals are obtained from the control module, and it is determined whether the feedback signals meet the continuity requirement. If they meet the continuity requirement, a real-time interaction algorithm is used to synchronize the status of all multiple devices to obtain a unified record as a global interaction log. The specific steps are as follows: Send the final execution command to the control module of the bedroom thermostat to initiate the temperature adjustment action; The control module collects execution feedback signals in real time, including the actual rate of temperature change and the equipment response status. The feedback signal is compared with the expected continuous demand. If the deviation is within the allowable range, it is determined that the continuous demand is met. Based on the judgment result, the time, location, device, intent, and execution status of this interaction event are synchronized to all associated devices through a real-time interaction algorithm, generating a global interaction log with a unified identifier.

[0028] This embodiment provides an AI-driven cross-device semantic understanding home interaction method. Based on the global interaction log, relevant entries for cross-spatial movement are extracted, and it is determined whether there is missing information in the entries. If there is missing information, the missing part is supplemented through a transmission mechanism. The supplemented log is determined to be a complete intent chain, and the optimized system response sequence is obtained. The specific steps are as follows: Filter out interaction records involving user movement across space from the global interaction log to form a cross-space entry set; Perform a structural integrity check on the cross-space entry set to identify any missing information such as intentional breakpoints or unsynchronized states; If information is missing, the corresponding context data is extracted from the source device logs through the historical transmission mechanism between devices to fill in the missing fields; Based on the filled complete log, the user intent evolution path is reconstructed to generate a logically coherent optimized system response sequence.

[0029] This embodiment provides an AI-driven cross-device semantic understanding home interaction method. Based on the optimized system response sequence, a history sharing algorithm is used for analysis to obtain potential system deviations from the sequence. The method determines whether the deviation stems from an instruction comprehension error. If it does, the sequence parameters are adjusted through a deviation correction process to obtain the corrected response output. The specific steps are as follows: The optimized system response sequence is backtracked to identify inconsistencies between the execution results and the user's original intent, and potential deviation markers are generated. Based on the potential deviation markers, and combined with the multi-turn interaction context, we analyze whether the deviation is caused by semantic misjudgment or referential confusion. If the deviation is determined to be due to an error in understanding the instruction, a deviation correction process is initiated to adjust key parameters in the response sequence, including the target value, the execution device, or the type of operation. The response output is regenerated based on the corrected parameters to ensure that it is consistent with the user's true intent, thus obtaining the corrected response output.

[0030] This embodiment provides an AI-driven cross-device semantic understanding home interaction method. Based on the corrected response output, user demand satisfaction indicators are extracted, and it is determined whether the indicators have reached a preset threshold. If the preset threshold is reached, the method broadcasts the output through a multi-device interaction channel, confirming that the broadcast content is a coherent interaction confirmation, and obtaining the final user experience data. The specific steps are as follows: Extract user demand satisfaction metrics from the corrected response output, including goal achievement, response timeliness, and operational accuracy; The user demand satisfaction index is compared with the preset satisfaction threshold. If all indexes are met, a confirmation broadcast mechanism is triggered. A unified interaction confirmation message is pushed to all relevant devices through a multi-device interaction channel. The message includes information such as operation success, current status, and follow-up suggestions. The final user experience data is compiled based on users' implicit or explicit feedback on confirmation information.

[0031] This embodiment provides an AI-driven cross-device semantic understanding home interaction method. Based on the final user experience data, a preset device state database is updated. The baseline data for the next cycle is obtained from the updated database. It is determined whether the baseline data needs further optimization. If further optimization is needed, the enhanced cross-device interaction framework is obtained through iterative processing using an intent continuity algorithm. The specific steps are as follows: Inject the valid parameters from the final user experience data into the preset device state library to update the environmental control strategy and user preference model; Based on the updated device status database, new baseline data is automatically extracted at the start of the next interaction cycle; An adaptive evaluation is performed on the benchmark data to determine whether it can effectively support the intent continuity requirements in complex scenarios. If the evaluation results indicate that further optimization is needed, the intent continuity algorithm is invoked to iteratively refine the baseline data, generating an enhanced cross-device interaction framework with stronger generalization capabilities.

[0032] The following is in conjunction with the appendix Figure 1This invention provides a detailed description of a specific implementation of an AI-driven, cross-device semantic understanding-based home interaction method. In this embodiment, the user resides in a smart home system equipped with smart living room devices (such as smart speakers and smart thermostats) and smart bedroom devices (such as air conditioners with voice assistants and environmental sensors). All devices are connected to a central AI interaction engine via a home LAN. This engine, deployed on a local edge server or cloud platform, possesses semantic parsing, intent recognition, context modeling, and multi-device collaborative control capabilities. Figure 1 As shown, the entire interaction process includes eight steps from S1 to S8, forming a closed-loop optimized cross-space continuous interaction system.

[0033] In step S1, the user in the living room issues a voice command to the living room device: "Set the temperature to 24 degrees." The voice recognition module built into the living room device converts the audio signal into text, and the semantic parsing module extracts the keywords "temperature" and "24 degrees." Combining this with the domain knowledge base, the system determines that the intent belongs to the "environmental control" category and generates an intent type label T_env=1. Subsequently, the system queries a preset device status database, which stores the historical settings, current operating modes, capability boundaries (such as minimum / maximum settable temperatures), and user preference data for each room's temperature control device. Assuming the living room air conditioner is currently in cooling mode, the room temperature is 26℃, and the device supports a setting range of 16–30℃, the initial adjustment value ΔT_init=24℃ is confirmed as valid. After verification by the device capability verification module, it is stored as baseline data B_0 in the context cache. This baseline data includes not only the target temperature value but also metadata such as timestamp t_0, spatial location loc="living room", and device ID=device_A.

[0034] When the user moves to the bedroom and says "higher" on the bedroom device, step S2 is initiated. The input module of the bedroom device receives this subsequent instruction, and the referencing resolution module, based on a pre-trained language model, identifies "higher" as a typical reference word, generating a reference judgment result R_ref=1. The system then retrieves the most recent record related to environmental control from the cross-device dialogue history cache, i.e., the baseline data B_0 generated in S1, and extracts the initial adjustment value of 24℃. Next, the semantic direction analysis module determines that "higher" represents a positive incremental adjustment. Combining this with preset incremental rules (such as a default increment of 1℃ each time), the updated adjustment target T_target=24+1=25℃ is calculated, and a continuous intent identifier C_intent="continuous temperature increase" is generated, forming the updated adjustment target data packet U_1.

[0035] Enter step S3, and the system calls the intention continuity algorithm to process U_1. The algorithm first obtains the current environmental data T_current = 22.5°C from the temperature and humidity sensor built in the bedroom device. The preset comfort threshold T_th = 23°C (which can be dynamically adjusted according to seasons or user habits). Since T_current < T_th, the system triggers the intention continuity verification process. During this process, the historical sharing mechanism spatio-temporally correlates the reference data B_0 in the living room stage with U_1 in the bedroom stage to construct the joint intention context C_joint = {loc_prev = "living room", T_prev = 24, loc_curr = "bedroom", ΔT_dir = +1}. Based on this context, the control logic engine determines that the user's real demand is to continue the original adjustment intention in the new space, so it generates the final execution instruction I_exec: "Adjust the set temperature of the bedroom air conditioner to 25°C and perform a temperature increase operation."

[0036] In step S4, the system sends I_exec to the control module of the bedroom temperature control device to start the action of heating or reducing the cooling power. After the control module executes, the real-time feedback signal F_fb is collected, including the actual temperature change rate v_T = 0.3°C / min and the device response status status = "active". The system compares F_fb with the expected continuous demand (such as the target temperature should reach within the error range of ±0.5°C within 5 minutes) for consistency, and uses the following deviation calculation formula: ; where N is the number of sampling points, and T_actual(t_i) is the actual temperature at the i-th time point. If ε ≤ ε_max (the preset tolerance, such as 0.4°C), it is determined that the continuous demand is met. At this time, the real-time interaction algorithm synchronizes the information such as the time t_1 of this event, the location "bedroom", the device ID = device_B, the intention type "environmental control", and the execution status "success" to all associated devices, and generates a global interaction log L_global with the global unique identifier GID_001, which is stored in the distributed log database.

[0037] Step S5 focuses on maintaining log integrity. The system extracts all entries involving cross-space movement (such as GID_001 and its predecessor GID_000) from L_global, forming a cross-space entry set S_cross. The structural integrity check module scans S_cross to detect whether there are intent breakpoints (such as missing initial settings) or unsynchronized state fields (such as the bedroom device not recording the original living room command). If missing information M_missing is found (e.g., ΔT_init is not included in the bedroom log), the corresponding context data is extracted from the living room device log through the inter-device transfer mechanism, and the fields are filled. The filled complete log constitutes a complete intent chain Chain_full, which is used to reconstruct the user's intent evolution path from the living room to the bedroom, generating an optimized system response sequence Seq_opt to ensure logical coherence without jumps.

[0038] In step S6, the system performs deviation analysis on Seq_opt. The history-sharing algorithm traces the user's behavior trajectory and calculates the consistency score between the execution result and the original intent. If a potential deviation is detected (e.g., the actual setting is 26℃ instead of 25℃), a deviation flag D_flag=1 is generated. The deviation source identification module, combined with multi-turn context, determines whether it is caused by semantic misjudgment (e.g., misinterpreting "a little higher" as +2℃) or referential confusion (e.g., incorrectly associating with an earlier lighting command). If it is confirmed to originate from a misunderstanding of the command, the deviation correction process is initiated: key parameters in the response sequence are adjusted, such as correcting the target value T_target ← 25℃, or specifying the execution device as device_B instead of device_C. The corrected parameters are re-injected into the execution engine, generating a corrected response output O_corr to ensure consistency with the user's true intent. Figure 1 To.

[0039] In step S7, the system extracts user requirement satisfaction metrics from O_corr, including goal achievement rate. The system evaluates user performance based on several metrics, including response time (T_resp, time from command issuance to completion) and operational accuracy (P_op, probability of correct execution). These metrics are compared to preset satisfaction thresholds (e.g., A_acc≤0.02, T_resp≤30s, P_op≥0.95). If all criteria are met, a multi-device interaction channel broadcasts a confirmation message, such as: "The bedroom temperature has been adjusted to 25℃. The current room temperature is 24.8℃. It is expected to reach the target in 2 minutes." This confirmation message is simultaneously pushed to living room devices, the mobile app, and the smartwatch. Users may reply with "okay" or no action (implicit feedback). The system then aggregates this information to form the final user experience data U_exp, which includes explicit ratings, interaction duration, and number of error corrections.

[0040] Finally, in step S8, U_exp is used for system self-optimization. Valid parameters (such as user preferred temperature 25℃, incremental habit +1℃) are injected into the device state library to update the environmental control strategy and personalized user model. At the start of the next interaction cycle, the system automatically extracts new baseline data B_1 from the updated library. The adaptive evaluation module tests B_1 to determine its ability to support intent continuity in complex scenarios (such as multi-user conflicting commands, device offline). If the evaluation score is below the threshold, the intent continuity algorithm is invoked for iterative refinement, such as introducing a reinforcement learning mechanism to optimize the incremental rule weights or expanding the context window length. This ultimately generates an enhanced cross-device interaction framework F_enhanced, which has stronger generalization capabilities and can support more device types (such as underfloor heating, fresh air systems) and more complex cross-spatial intent chains (such as "dim the lights like in the living room").

[0041] In summary, this embodiment, through the closed-loop process of steps S1-S8, achieves continuous cross-device interaction based on AI semantic understanding, effectively solving the problems of intent breakage, context loss, and command misunderstanding caused by spatial switching in traditional smart homes. The entire system relies on core technologies such as semantic parsing, referential resolution, intent continuity algorithms, history sharing mechanisms, and real-time feedback correction, combined with the attached... Figure 1 The various functional modules shown work together to ensure that the home system can seamlessly understand and execute the user's continuous commands when the user moves between different rooms, significantly improving the naturalness of the interaction and the user experience.

[0042] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the essence and scope of the technical solutions of the present invention.

Claims

1. An AI-driven, cross-device semantic understanding method for home interaction, characterized in that: Includes the following steps: S1. Obtain the user's initial command content on the living room device, extract the temperature adjustment intention from the initial command content, determine the intention type as environmental control, and obtain the initial adjustment value as the reference data by matching the intention type with the preset device status library. S2. Based on the baseline data and the user's subsequent instructions when moving to the bedroom, determine whether there are quoted words in the subsequent instructions. If there are quoted words, obtain the aforementioned initial adjustment value from the dialogue history, determine that the continuous intention is to further modify the temperature, and obtain the updated adjustment target. S3. Based on the updated adjustment target, the intent continuity algorithm is used to process the current environmental data from the bedroom device sensor. It is determined whether the current environmental data is lower than the preset threshold. If it is lower than the preset threshold, the baseline data and the updated adjustment target are transmitted through the history sharing mechanism to determine that the final execution instruction is the temperature increment operation. S4. Based on the final execution instruction, activate the control module of the bedroom device, obtain feedback signals from the control module, determine whether the feedback signals meet the continuity requirements, and if they meet the continuity requirements, use a real-time interaction algorithm to synchronize the status of all multiple devices and obtain a unified record as a global interaction log. S5. Based on the global interaction log, extract relevant entries for cross-space movement, determine whether there is missing information in the entries, and if there is missing information, supplement the missing part through the transmission mechanism, determine that the supplemented log is a complete intent chain, and obtain the optimized system response sequence. S6. Based on the optimized system response sequence, the history sharing algorithm is used to analyze and obtain potential system deviations from the sequence. It is determined whether the deviations are due to errors in instruction understanding. If they are due to errors in instruction understanding, the sequence parameters are adjusted through the deviation correction process to obtain the corrected response output. S7. Based on the corrected response output, extract the user demand satisfaction index, determine whether the index reaches the preset threshold, and if it reaches the preset threshold, broadcast the output through the multi-device interaction channel to determine that the broadcast content is a continuous interactive confirmation, and obtain the final user experience data. S8. Based on the final user experience data, update the preset device state library, obtain the benchmark data for the next cycle from the updated library, determine whether the benchmark data needs further optimization, and if it needs further optimization, iterate through the intent continuity algorithm to obtain an enhanced cross-device interaction framework.

2. The AI-driven cross-device semantic understanding home interaction method according to claim 1, characterized in that: The baseline data generation includes intent type identification, device state library matching, and initial adjustment value extraction. The updated adjustment targets include reference word detection, historical intent backtracking, and continuous intent determination. The final execution instructions include environmental data comparison, historical sharing mechanism triggering, and operation type confirmation. The global interaction log includes multi-device state synchronization, interaction event marking, and timestamp binding. The optimized system response sequence includes intent chain integrity verification, missing information location, and context completion. The corrected response output includes deviation source identification, parameter dynamic correction, and output consistency verification. The final user experience data includes requirement satisfaction measurement, interaction fluency evaluation, and feedback loop construction. The enhanced cross-device interaction framework includes dynamic state library updates, intent continuity enhancement, and adaptive optimization of interaction paths.

3. The AI-driven cross-device semantic understanding home interaction method according to claim 1, characterized in that: The specific steps for obtaining the user's initial command content on the living room device, extracting the temperature adjustment intent from the initial command content, determining the intent type as environmental control, and matching the intent type with a preset device state library to obtain the initial adjustment value as baseline data are as follows: The system acquires the user's initial voice or text commands on the living room device, uses semantic parsing technology to identify the temperature adjustment keywords contained therein, and generates preliminary intent fragments. Based on the preliminary intent fragment, combined with contextual analysis, it is determined whether the intent category it belongs to is environmental control, and an intent type label is generated; Based on the intent type tag, query the parameter configurations related to environmental control in the preset device status library, match the current operating status of the living room temperature control device, and generate the initial adjustment value; Based on the initial adjustment values, and combined with device capability boundary verification, ensure that it is within the executable range and generate baseline data for subsequent interactions.

4. The AI-driven cross-device semantic understanding home interaction method according to claim 1, characterized in that: Based on the baseline data and the user's subsequent instructions when moving to the bedroom, the steps to determine whether there are quoted words in the subsequent instructions are as follows: If there are quoted words, the initial adjustment value is obtained from the dialogue history, and the continuous intent is to further modify the temperature. The updated adjustment target is obtained by retrieving the initial adjustment value from the dialogue history. The system receives subsequent instructions input by the user in the bedroom device, uses a substitution resolution technique to identify whether they contain quotations, and generates a quotation word determination result; based on the quotation word determination result; If a reference word exists, retrieve the most recent instruction record related to environmental control from the cross-device dialogue history cache and extract the corresponding initial adjustment value; Based on the initial adjustment value and the semantic direction of the current subsequent instruction, determine whether the user's intent is a continuous modification of the original adjustment value, and generate a continuous intent identifier; Based on the continuous intent identifier and combined with the incremental adjustment rules, a new target temperature value is calculated, and an updated adjustment target is generated.

5. The AI-driven cross-device semantic understanding home interaction method according to claim 1, characterized in that: Based on the updated adjustment target, the intent continuity algorithm is used to process the current environmental data obtained from the bedroom device sensors. It is then determined whether the current environmental data is below a preset threshold. If it is below the preset threshold, the baseline data and the updated adjustment target are transmitted through a history sharing mechanism. The specific steps for determining the final execution instruction as a temperature increment operation are as follows: The current environmental data is obtained from the temperature and humidity sensor built into the bedroom device. The current environmental data is compared with a preset comfort threshold. If it is lower than the threshold, the intent continuity verification process is triggered. Based on the intent continuity verification process, the baseline data of the living room stage is associated and mapped with the updated adjustment target of the bedroom stage through the historical sharing mechanism to generate a joint intent context. Based on the aforementioned joint intent context and combined with the device control logic, a clear temperature increment operation instruction is generated as the final execution instruction.

6. The AI-driven cross-device semantic understanding home interaction method according to claim 1, characterized in that: Based on the final execution instruction, the control module of the bedroom device is activated, feedback signals are obtained from the control module, and it is determined whether the feedback signals meet the continuity requirements. If they do meet the continuity requirements, a real-time interaction algorithm is used to synchronize the status of all multiple devices and obtain a unified record as a global interaction log. The specific steps are as follows: Send the final execution command to the control module of the bedroom thermostat to initiate the temperature adjustment action; The control module collects execution feedback signals in real time, including the actual rate of temperature change and the equipment response status. The feedback signal is compared with the expected continuous demand. If the deviation is within the allowable range, it is determined that the continuous demand is met. Based on the judgment result, the time, location, device, intent, and execution status of this interaction event are synchronized to all associated devices through a real-time interaction algorithm, generating a global interaction log with a unified identifier.

7. The AI-driven cross-device semantic understanding home interaction method according to claim 1, characterized in that: Based on the global interaction log, relevant entries for cross-spatial movement are extracted. It is determined whether any information is missing from these entries. If missing information is found, the missing portion is supplemented through a transmission mechanism. The supplemented log is then confirmed to be a complete intent chain, leading to the optimized system response sequence. The specific steps are as follows: Filter out interaction records involving user movement across space from the global interaction log to form a cross-space entry set; Perform a structural integrity check on the cross-space entry set to identify any missing information such as intentional breakpoints or unsynchronized states; If information is missing, the corresponding context data is extracted from the source device logs through the historical transmission mechanism between devices to fill in the missing fields; Based on the filled complete log, the user intent evolution path is reconstructed to generate a logically coherent optimized system response sequence.

8. The AI-driven cross-device semantic understanding home interaction method according to claim 1, characterized in that: Based on the optimized system response sequence, a history sharing algorithm is used to analyze and obtain potential system biases from the sequence. It is then determined whether the bias stems from an instruction comprehension error. If it does, the sequence parameters are adjusted through a bias correction process to obtain the corrected response output. The specific steps are as follows: The optimized system response sequence is backtracked to identify inconsistencies between the execution results and the user's original intent, and potential deviation markers are generated. Based on the potential deviation markers, and combined with the multi-turn interaction context, we analyze whether the deviation is caused by semantic misjudgment or referential confusion. If the deviation is determined to be due to an error in understanding the instruction, a deviation correction process is initiated to adjust key parameters in the response sequence, including the target value, the execution device, or the type of operation. The response output is regenerated based on the corrected parameters to ensure that it is consistent with the user's true intent, thus obtaining the corrected response output.

9. The AI-driven cross-device semantic understanding home interaction method according to claim 1, characterized in that: Based on the corrected response output, the user demand satisfaction index is extracted, and it is determined whether the index reaches a preset threshold. If the preset threshold is reached, the response is broadcast through a multi-device interaction channel, and the broadcast content is determined to be a continuous interaction confirmation. The specific steps to obtain the final user experience data are as follows: Extract user demand satisfaction metrics from the corrected response output, including goal achievement, response timeliness, and operational accuracy; The user demand satisfaction index is compared with the preset satisfaction threshold. If all indexes are met, a confirmation broadcast mechanism is triggered. A unified interaction confirmation message is pushed to all relevant devices through a multi-device interaction channel. The message includes information such as operation success, current status, and follow-up suggestions. The final user experience data is compiled based on users' implicit or explicit feedback on confirmation information.

10. The AI-driven cross-device semantic understanding home interaction method according to claim 1, characterized in that: Based on the final user experience data, the preset device state database is updated. The baseline data for the next cycle is obtained from the updated database. It is determined whether the baseline data needs further optimization. If further optimization is needed, the enhanced cross-device interaction framework is obtained through iterative processing using an intent continuity algorithm. The specific steps are as follows: Inject the valid parameters from the final user experience data into the preset device state library to update the environmental control strategy and user preference model; Based on the updated device status database, new baseline data is automatically extracted at the start of the next interaction cycle; An adaptive evaluation is performed on the benchmark data to determine whether it can effectively support the intent continuity requirements in complex scenarios. If the evaluation results indicate that further optimization is needed, the intent continuity algorithm is invoked to iteratively refine the baseline data, generating an enhanced cross-device interaction framework with stronger generalization capabilities.

Images