A home intelligent recommendation method and system based on cross-modal fusion

Abstract

The application relates to a household intelligent recommendation method and system based on cross-modal fusion, which comprises the following steps: collecting user modal data, environment modal data and equipment behavior modal data in real time; converting the environment modal data into quantum state amplitude characteristics through a quantum encoder, and simultaneously obtaining user behavior phase characteristics through a time sequence feature extractor; performing cross-modal fusion on the quantum state amplitude characteristics and the behavior phase characteristics based on quantum entanglement operation to generate a dynamic fusion vector; inputting the dynamic fusion vector into a mixed decision engine, outputting a recommendation instruction set from the mixed decision engine, and verifying the physical feasibility of the recommendation instruction set based on a device state knowledge graph; and generating a final recommendation instruction according to the verification result and executing the final recommendation instruction. The application realizes intelligent matching of job seekers and recruitment demands, and improves the efficiency and accuracy of recruitment.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of smart home, in particular to a smart home recommendation method and system based on cross-modal fusion. BACKGROUND

[0002] With the rapid development of Internet of Things and artificial intelligence technology, modern smart home systems are facing multiple challenges such as massive heterogeneous data processing, real-time decision response and privacy security protection. Traditional smart home recommendation systems mainly rely on classic machine learning algorithms to process multi-source data such as user behavior, environmental sensing and device state. When the system needs to fuse multi-modal data such as voice, image and sensor to realize personalized services, there are generally problems of low cross-modal data fusion efficiency and semantic gap. The static model architecture is difficult to adapt to the continuous environmental evolution and user demand drift of the home scene, resulting in poor dynamic adaptability of the system and delay in responding to sudden scenarios. SUMMARY

[0003] To solve the above technical problems, the application provides a smart home recommendation method and system based on cross-modal fusion, which realizes the deep cooperation of environment-user-device through quantum entanglement operation, and improves the sudden scenario processing speed and dynamic adaptability of the system.

[0004] The above invention purpose of the application is realized by the following technical scheme:

[0005] A smart home recommendation method based on cross-modal fusion, comprising the steps of:

[0006] real-time collection of user modal data, environmental modal data and device behavior modal data;

[0007] convert the environmental modal data into quantum state amplitude features through a quantum encoder, and obtain user behavior phase features through a time sequence feature extractor;

[0008] cross-modal fusion of the quantum state amplitude features and behavior phase features based on quantum entanglement operation to generate a dynamic fusion vector;

[0009] input the dynamic fusion vector into a hybrid decision engine, and the hybrid decision engine outputs a recommendation instruction set, and verifies the physical feasibility of the recommendation instruction set based on a device state knowledge graph;

[0010] generate a final recommendation instruction according to the verification result and execute it.

[0011] By adopting the technical scheme, quantum entanglement operation reduces environment and behavior data fusion delay, is suitable for smart home millisecond level response scene, and through three-dimensional verification of the device state knowledge graph, error instruction interception rate is improved. For example, when gas leakage (environment mutation) and user leaving home (behavior anomaly) occur in the kitchen at the same time, the quantum fusion module generates a "safety risk" dynamic fusion vector, the decision engine outputs "close the gas valve + start ventilation" instruction, and the knowledge graph verification circuit executes immediately after the load is safe, thereby solving the problems of cross-modal data fragmentation and burst scene response delay.

[0012] In a preferred example of the present application: the step of converting the environment modal data into quantum state amplitude features through a quantum encoder, while obtaining user behavior phase features through a time sequence feature extractor, specifically includes the steps of:

[0013] Receiving continuous physical quantity data streams from environment sensors, including temperature, light intensity and device running state, and discretizing the physical quantity values into environment quantum state probability distribution intervals that can be processed by quantum computing;

[0014] Capturing user operation event sequences, including voice instruction trigger time points and device interaction action intervals, constructing a behavior time axis, and extracting user behavior phase features.

[0015] By adopting the technical scheme, physical quantity quantization realizes accurate discretization of physical quantities such as temperature (0-40℃→16 quantum bits) and light (0-1000 lux→8 quantum bits), reduces sampling error compared with traditional ADC, and reduces misjudgment rate by distinguishing between "active adjustment" and "mistouch" behaviors through constructing an "operation interval time axis". For example, in the plum rain season, the system quantizes the environment data "humidity continuously for 12 hours > 75%" into quantum state probability distribution, and captures the behavior time sequence feature of "frequent on-off of the dehumidifier".

[0016] In a preferred example of the present application: the step of cross-modal fusion of the quantum state amplitude features and the behavior phase features based on quantum entanglement operation to generate a dynamic fusion vector, specifically includes the steps of:

[0017] Loading the quantum state amplitude features into a first quantum feature carrier, loading the behavior phase features into a second quantum feature carrier, performing quantum entanglement initialization operation on the two carriers, and establishing a quantum state correlation channel;

[0018] Monitoring real-time changes of the behavior phase features, triggering phase rotation operation, and dynamically adjusting entanglement strength based on environment amplitude features;

[0019] The quantum state amplitude feature of the first quantum feature carrier is taken as a control bit, a conditional rotation is performed on the behavior phase feature of the second quantum feature carrier, and a fusion quantum state is generated;

[0020] The fusion quantum state is projected to a classical data space, a double-modal carrying capacity of a vector is verified, and a dynamic fusion vector carrying the quantum state amplitude feature and the behavior phase feature is generated.

[0021] By adopting the technical solution, quantum entanglement improves the correlation coefficient of the environmental temperature and the user behavior, thereby improving the relevance of the recommendation instruction. The entanglement initialization operation establishes a cross-modal physical correlation channel, avoids manual weight setting, performs a conditional rotation on the behavior phase feature of the second quantum feature carrier, improves the dynamic adaptability of the behavior feature to the environmental state, realizes situation awareness fusion, and realizes flexible switching of the fusion strategy in the case of environmental mutation through graded adjustment of the entanglement strength.

[0022] In a preferred example of the present application, the step of monitoring the real-time change of the behavior phase feature and triggering the phase rotation operation, and dynamically adjusting the entanglement strength based on the environmental amplitude feature, specifically includes the steps of:

[0023] Continuously receiving the user behavior time sequence feature flow, calculating the phase offset of adjacent time windows;

[0024] When a positive mutation pattern is identified, a phase accelerator is started to perform a positive rotation, thereby improving the priority of the feature vector in decision-making;

[0025] When a negative drift pattern is identified, a phase compensator is triggered to perform an inverse rotation, thereby reducing the confidence weight of the feature vector;

[0026] The environmental amplitude feature is analyzed in real time, and the entanglement energy level is adjusted through a quantum coupler. When the amplitude fluctuation is below a preset threshold, a low entanglement strength is maintained, and the independence of each modal feature is preserved;

[0027] When a continuous amplitude step is detected, the entanglement strength is switched to high, and the double-modal feature is forced to update synchronously.

[0028] By adopting the technical solution, the calculation of the phase offset quantifies the mutation degree of the behavior pattern, avoids subjective threshold setting, the positive / inverse rotation mechanism dynamically normalizes the behavior feature, eliminates the time scale difference, the amplitude step detection can accurately capture the moment of environmental mutation, and the burst response sensitivity is improved.

[0029] In a preferred example of the present application, the step of taking the quantum state amplitude feature of the first quantum feature carrier as a control bit, performing a conditional rotation on the behavior phase feature of the second quantum feature carrier, and generating a fusion quantum state, specifically includes the steps of:

[0030] Extracting the environmental data stability index and the sudden event marker intensity information in the quantum state amplitude feature to generate a quantum control signal bound to the quantum state amplitude feature;

[0031] Inputting the quantum control signal into a second quantum feature carrier interface, activating a rotation mechanism according to the control signal type, triggering a micro-phase calibration for a steady-state control signal, and activating a large-angle phase reset for a mutation suppression signal;

[0032] Performing a rotation operation to enhance the behavior feature weight through forward rotation and suppress abnormal behavior interference through reverse rotation;

[0033] Synchronously maintaining quantum entanglement during the execution of the rotation operation to generate a fusion quantum state containing a main feature layer, a modulation layer, and a correlation layer, the main feature layer inherits the environmental amplitude distribution, the modulation layer carries the phase information after rotation, and the correlation layer records the interaction trajectory of the double carriers.

[0034] By using the above technical solution, the decision value of the environmental feature is quantified by extracting the environmental data stability index and the sudden event marker intensity information, which can avoid invalid signal interference, activate the rotation mechanism according to the control signal type, dynamically match the behavior adjustment intensity, realize precise scene adaptation, and generate new features while retaining original information through the fusion quantum state, thereby breaking through the limitations of simple weighted fusion.

[0035] In a preferred example of the present application, the step of inputting the dynamic fusion vector into a hybrid decision engine, the hybrid decision engine outputting a recommended instruction set, and verifying the physical feasibility of the recommended instruction set based on a device state knowledge graph, specifically includes the following steps:

[0036] Extracting the environmental state code containing the abnormal event marker in the dynamic fusion vector and the user intent probability distribution containing the confidence of each behavior mode to generate a double-channel input data stream;

[0037] Inputting the environmental state code into a context perception module to identify the current home scene type and predict the environmental evolution trend;

[0038] Inputting the user intent probability distribution into an intent reasoning module to analyze the explicit operation demand of the user and mine the implicit behavior preference of the user, and outputting a recommended instruction set with confidence;

[0039] Loading the recommended instruction set into a symbolic rule verifier, performing three-dimensional verification in conjunction with the device state knowledge graph to generate an instruction feasibility marker, and the three-dimensional verification includes electrical compatibility, spatial accessibility, and safety compliance verification.

[0040] By adopting the technical scheme, the physical layer and the semantic layer of the cross-modal feature are decoupled by separating the environment state code and the user intention probability distribution, the context perception module realizes the environment trend prediction for the next 30 minutes by constructing a space-time matrix containing 12 environmental parameters (temperature, humidity, illumination, etc.), the intention reasoning module innovatively introduces the knowledge graph embedding technology, successfully mines the user's implicit demand by constructing an intention reasoning tree containing 3 layers of implicit semantics, and the verification coverage of the physical feasibility of the final instruction is significantly increased through the linkage of the symbolic rule verifier and the knowledge graph.

[0041] In a preferred example of the present application: the instruction feasibility label includes a green instruction of full constraint pass, a yellow instruction of soft constraint violation and a red instruction of hard constraint violation, and the step of generating a final recommended instruction according to the verification result and executing the final recommended instruction specifically includes the steps of:

[0042] Performing hierarchical processing according to the instruction feasibility label, the green instruction directly enters the execution queue, the yellow instruction triggers the constraint optimizer, and the red instruction starts the safety coverage engine;

[0043] Outputting the final recommended instruction with a correction record and executing the final recommended instruction.

[0044] By adopting the technical scheme, a dynamic spectrum from safety to optimization is constructed by using red, yellow and green three-color labels, the risk is visualized and managed in layers, the red instruction corresponds to the hard constraint (such as electrical compatibility ≥ 95% pass rate), the yellow instruction corresponds to the soft constraint (such as energy consumption exceeding the standard ≤ 15%), and the green instruction corresponds to the full compliance state, and this quantitative grading improves the transparency of system decision-making.

[0045] The second invention purpose of the present application is achieved by the following technical scheme:

[0046] A home intelligent recommendation system based on cross-modal fusion, comprising:

[0047] A data collection module for collecting user modal data, environment modal data and device behavior modal data in real time;

[0048] A data processing module for converting the environment modal data into quantum state amplitude features through a quantum encoder, and obtaining user behavior phase features through a time sequence feature extractor;

[0049] A cross-modal fusion module for cross-modal fusion of the quantum state amplitude features and the behavior phase features based on quantum entanglement operation to generate a dynamic fusion vector;

[0050] An instruction recommendation module for inputting the dynamic fusion vector into a hybrid decision engine, the hybrid decision engine outputting a recommended instruction set, and verifying the physical feasibility of the recommended instruction set based on a device state knowledge graph.

[0051] The execution module is configured to generate a final recommendation instruction according to the verification result and perform the final recommendation instruction.

[0052] The third purpose of the present application is achieved by the following technical solution:

[0053] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the above-mentioned home intelligent recommendation method based on cross-modal fusion when executing the computer program.

[0054] The fourth purpose of the present application is achieved by the following technical solution:

[0055] A computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of the above-mentioned home intelligent recommendation method based on cross-modal fusion.

[0056] In summary, the present application includes at least one of the following beneficial technical effects:

[0057] 1. The present application converts environmental data into quantum state amplitude features (such as temperature sudden rise mapping to ground state probability 0.95), and encodes user behavior sequence into phase features (such as operation delay marking as phase lag), and establishes a dynamic binding channel between dual carriers through quantum entanglement operation, taking the example of a solitary old person falling down: when the gravity sensor detects 9.8 m / s 2 impact (environmental amplitude mutation) and user stationary timeout (behavior phase lag) occurs, the system generates a strong correlation fusion state, automatically triggers an alarm and increases the heating of the floor heating, improves the response speed and reduces the false alarm rate, and solves the problem of separation of environmental data and user behavior in the field of smart home through the quantum amplitude-phase entanglement fusion mechanism;

[0058] 2. The present application dynamically adjusts the behavior phase through the environmental amplitude feature: when a heavy rain (amplitude continuous step) is detected, the mutation suppression signal triggers a 120° large-angle rotation, rotating the user's "open window" instruction phase from the ventilation demand zone to the safety protection zone, and at the same time enhancing the entanglement strength to force the binding of the "close window + dehumidification" instruction; In the energy optimization scene, after the system identifies the power peak (amplitude fluctuation mark), the high-energy-consumption behavior instruction is executed in reverse rotation to reduce the weight. This quantum cascade mechanism of environment-behavior improves the resource utilization efficiency and safety protection strength;

[0059] 3. Through the neural-symbol hybrid decision engine and the three-dimensional knowledge graph verification system, the physical feasibility of the recommended instructions is realized, and the traditional smart home system often ignores the reality constraints, resulting in dangerous operation, such as recommending "air conditioner 16℃ strong cooling + oven start" to cause circuit overload, the initial instruction output by the decision engine in the application needs to be verified by the knowledge graph three times: electrical compatibility (load capacity verification), spatial accessibility (mechanical arm path obstacle avoidance), and safety compliance (safety regulation review), which effectively reduces the device failure rate and reduces household electrical safety accidents. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 is a flowchart of an embodiment of the home intelligent recommendation method and system based on cross-modal fusion of the application;

[0061] Figure 2 is an implementation flowchart of step S30 in an embodiment of the home intelligent recommendation method and system based on cross-modal fusion of the application;

[0062] Figure 3 is an implementation flowchart of step S40 in an embodiment of the home intelligent recommendation method and system based on cross-modal fusion of the application;

[0063] Figure 4 is a schematic diagram of the home intelligent recommendation system based on cross-modal fusion of the application;

[0064] Figure 5 is a principle block diagram of a computer device of the application. DETAILED DESCRIPTION

[0065] The application will be further described in detail below with reference to the accompanying drawings.

[0066] In an embodiment, as shown in Figures 1-3 , the application discloses a home intelligent recommendation method based on cross-modal fusion, which specifically includes the following steps:

[0067] S10: Real-time collection of user modal data, environment modal data and device behavior modal data;

[0068] S20: converting the environment modal data into quantum state amplitude features through a quantum encoder, and obtaining user behavior phase features through a time sequence feature extractor;

[0069] S30: cross-modal fusion of the quantum state amplitude features and the behavior phase features based on quantum entanglement operation to generate a dynamic fusion vector;

[0070] S40: inputting the dynamic fusion vector into a hybrid decision engine, the hybrid decision engine outputting a recommended instruction set, and verifying the physical feasibility of the recommended instruction set based on a device state knowledge graph.

[0071] S50: generating final recommendation instructions according to the verification result and executing.

[0072] In the embodiment, the quantum entanglement operation reduces the environment and behavior data fusion delay, is suitable for the smart home millisecond level response scene, and improves the error instruction interception rate through the three-dimensional verification of the device state knowledge graph. For example, when a gas leakage (environment mutation) and a user leaving home (behavior anomaly) occur in the kitchen at the same time, the quantum fusion module generates a "safety risk" dynamic fusion vector within 0.2 seconds, the decision engine outputs the "close the gas valve + start ventilation" instruction, and the knowledge graph verification circuit executes immediately after the load is safe, thereby solving the problem of cross-modal data fragmentation and burst scene response delay.

[0073] In an embodiment, step S20 specifically includes the following steps:

[0074] S21: receiving a continuous physical quantity data stream from an environment sensor, including temperature, light intensity and device running state, and discretizing the physical quantity value into an environment quantum state probability distribution interval that can be processed by quantum computing;

[0075] S22: capturing a user operation event sequence, including a voice instruction trigger time point and a device interaction action interval, constructing a behavior time axis, and extracting user behavior phase features.

[0076] Specifically, the temperature value is mapped to the probability value of the quantum bit being in the ground state, the light intensity is converted into the proportion of the |0> state in the quantum superposition state, and the device switch state is coded as the direction identifier of the quantum rotation operation; the user operation event sequence is captured, including the voice instruction trigger time point and the device interaction action interval, the behavior time axis is constructed, the operation frequency distribution pattern is analyzed through the sliding window, the key phase features of the behavior pattern are extracted, the high-frequency operation cluster is identified as the phase leading feature, the operation delay event is converted into the phase lag quantity, and the periodicity of the behavior periodicity pattern is embodied as the periodicity of the phase angle.

[0077] In the embodiment, when the user turns on the reading lamp at 22:00 for three times in succession (behavior phase feature: high-frequency stable operation), the light sensor detects that the environmental brightness is > 90 lux (quantum amplitude feature: strong light probability 0.92).

[0078] In an embodiment, step S30 specifically includes the following steps:

[0079] S31: loading the quantum state amplitude feature to a first quantum feature carrier, loading the behavior phase feature to a second quantum feature carrier, performing a quantum entanglement initialization operation on the two carriers, and establishing a quantum state correlation channel;

[0080] S32: Monitor the real-time changes of the behavior phase feature, trigger the phase rotation operation, and dynamically adjust the entanglement strength based on the environmental amplitude feature;

[0081] S33: Take the quantum state amplitude feature of the first quantum feature carrier as the control bit, perform conditional rotation on the behavior phase feature of the second quantum feature carrier, and generate a fusion quantum state;

[0082] S34: Project the fusion quantum state to the classical data space, verify the bimodal carrying capacity of the vector, and generate a dynamic fusion vector carrying the quantum state amplitude feature and the behavior phase feature.

[0083] Specifically, the quantum state amplitude feature is loaded into the first quantum feature carrier, the behavior phase feature is loaded into the second quantum feature carrier, quantum entanglement initialization operation is performed on the double carriers to establish a quantum state correlation channel; the real-time changes of the behavior phase feature are monitored to trigger the phase rotation operation, specifically, when a high-frequency operation cluster is detected, the second carrier is driven to perform forward phase rotation, and when an operation delay event occurs, the second carrier is driven to perform reverse phase compensation; the entanglement strength is dynamically adjusted based on the environmental amplitude feature, specifically, when the environmental data is stable, a weak entangled state is maintained to preserve the modal independence, and when the environment mutates, a strong entangled state is started to force synchronous update of the bimodal feature; the quantum state amplitude feature of the first quantum feature carrier is taken as the control bit to perform conditional rotation on the behavior phase feature of the second quantum feature carrier to generate a fusion quantum state, the characteristics of which include: the amplitude component inherits the statistical distribution of environmental data, the phase component reflects the behavior timing change rule, and the entanglement correlation degree represents the environmental-behavior matching degree; the amplitude component is mapped to the numerical strength of the feature vector, the phase component is converted to the direction attribute of the feature vector, it is checked whether the amplitude strength retains the environmental mutation mark, it is confirmed whether the direction attribute contains the behavior period pattern, and a dynamic fusion vector carrying the environmental-behavior joint feature is output.

[0084] In this embodiment, when the window vibration (environmental mutation: amplitude feature abnormally enhanced) and the user's mobile phone moving away from the house (behavior feature: phase lagging) are monitored, the strong entangled state binds the double features, the controlled door rotates the behavior phase to the warning interval with the window vibration amplitude as the control bit, the output vector carries the "high amplitude + negative phase" feature, the decision layer analyzes it as an intrusion risk, and the window is closed and an alarm is pushed.

[0085] In an embodiment, step S32 specifically includes the following steps:

[0086] S321: Continuously receive the user behavior timing feature stream, and calculate the phase offset of adjacent time windows;

[0087] S322: When a positive mutation pattern is identified, a phase accelerator is started to perform forward rotation to improve the priority of the feature vector in decision-making;

[0088] S323: When a negative drift pattern is identified, trigger the phase compensator to perform reverse rotation, reduce the confidence weight of the feature vector;

[0089] S324: Real-time analysis of the environmental amplitude characteristics, adjust the entanglement energy level through the quantum coupler, when the amplitude fluctuation is below the preset threshold, maintain low entanglement strength, and preserve the independence of each modal feature;

[0090] S325: When detecting continuous amplitude steps, switch to high entanglement strength, and force synchronous update of the dual-modal features.

[0091] Specifically, a real-time phase tracking channel is established to continuously receive user behavior time sequence feature flow, calculate the phase offset of adjacent time windows, and classify the phase change pattern: positive mutation pattern: identify the operation frequency jump event; negative drift pattern: detect operation delay or interruption event; periodic oscillation pattern: capture regular behavior cycle characteristics. According to the change pattern, activate the rotation mechanism, when a positive mutation pattern is identified, start the phase accelerator to perform forward rotation, and increase the priority of the feature vector in decision-making; when a negative drift pattern is identified, trigger the phase compensator to perform reverse rotation, and reduce the confidence weight of the feature vector, wherein the rotation angle is positively correlated with the phase offset, and the maximum rotation amplitude is limited by the physical hardware constraint. Real-time analysis of the environmental amplitude characteristics, calculate the amplitude fluctuation variance, when the amplitude fluctuation is below the preset threshold, mark it as a steady-state environment, identify the amplitude step change point, when continuous amplitude steps are detected, mark it as a transient environment. Establish the environment-entanglement strength mapping rule, maintain low entanglement strength in the steady-state environment, and preserve the independence of each modal feature; switch to high entanglement strength in the transient environment, and force synchronous update of the dual-modal features, wherein the entanglement energy level is adjusted through the quantum coupler, and the strength level is positively correlated with the amplitude mutation amplitude.

[0092] In this embodiment, when the user leaves home (negative drift pattern), the phase compensator performs reverse rotation to reduce the "user at home" feature weight, while detecting a 5°C room temperature drop (amplitude step), the entanglement strength switches to high energy level to force the binding of temperature mutation and user departure features, and the output fusion vector triggers the "energy saving mode": turn off the air conditioner and start the low temperature protection.

[0093] In an embodiment, step S33 specifically includes the following steps:

[0094] S331: Extract the environmental data stability index and burst event marker strength information in the quantum state amplitude characteristics, and generate a quantum control signal bound to the quantum state amplitude characteristics;

[0095] S332: input the quantum control signal into the second quantum feature carrier interface, activate the rotation mechanism according to the control signal type, trigger the fine phase calibration for the steady state control signal, and activate the large-angle phase reset for the mutation suppression signal;

[0096] S333: execute the rotation operation, enhance the behavior feature weight through forward rotation, and suppress abnormal behavior interference through reverse rotation;

[0097] S334: maintain quantum entanglement synchronously during the execution of the rotation operation, generate a fusion quantum state containing a main feature layer, a modulation layer, and a correlation layer, the main feature layer inherits the environmental amplitude distribution, the modulation layer carries the phase information after rotation, and the correlation layer records the interaction trajectory of the double carriers.

[0098] Specifically, the synchronous maintenance of quantum entanglement during rotation includes: transmitting amplitude feature changes through a quantum bus, and updating phase rotation parameters in real time. After generating the fusion quantum state, execute quantum state characteristic detection: verify whether the main feature layer retains the environmental mutation mark, and detect whether the modulation layer reflects the behavior pattern change; generate a vector through a quantum-classical converter: the amplitude component is mapped to a numerical feature, the phase component is converted into a direction vector, and the correlation layer information is encoded into a feature confidence.

[0099] In this embodiment, when detecting heavy rain weather (mutation suppression signal: amplitude feature fluctuates sharply), the control signal triggers a large-angle phase reset, rotates the user's "timed light-off" behavior phase to an emergency mode, and generates a fusion state containing a main feature layer: a heavy rain red warning mark; a modulation layer: a light keep-on instruction; a correlation layer: an environment-behavior strong coupling identifier, and an output vector triggers the whole house lighting system to extend the opening.

[0100] In an embodiment, step S40 specifically includes the following steps:

[0101] S41: extract the environmental state code containing the abnormal event mark in the dynamic fusion vector and the user intention probability distribution containing each behavior mode confidence, and generate a double-channel input data stream;

[0102] S42: input the environmental state code into the context perception module, identify the current home scene type, and predict the environmental evolution trend;

[0103] S43: input the user intention probability distribution into the intention reasoning module, analyze the explicit operation demand of the user, mine the implicit behavior preference of the user, and output a recommended instruction set with confidence;

[0104] S44: load the recommended instruction set into the symbolic rule verifier, link the device state knowledge graph to perform three-dimensional verification, generate an instruction feasibility mark, and the three-dimensional verification includes electrical compatibility, spatial accessibility, and safety compliance verification.

[0105] In this embodiment, the abnormal event is marked as a sudden abnormality in environmental data (e.g., sudden temperature change, smoke alarm), the behavior pattern confidence is a probabilistic quantification of the user's operation intention (e.g., "movie watching mode" probability 85%), and the dual-channel input is a parallel processing channel that separates environmental and user data. The home scene type includes security, comfort, energy saving, etc., the environmental evolution trend is the prediction of physical quantity changes, such as temperature rise rate, the embedded AI chip (e.g., NVIDIA Jetson) runs a lightweight neural network, input: "abnormal type": "PM2.5 exceeds standard", "value": 150 μg / m 3 ; output: scene type = health warning, trend = continuous deterioration (slope + 12% / min). The user's explicit operation requirement includes the user's direct instruction (e.g., "set to 26°C"), and the implicit behavior preference is the user's historical habit derivation (e.g., automatic turn off the light at 22:00). The electrical compatibility is the matching degree of device power consumption / circuit load, which is marked red if it exceeds the limit; the spatial accessibility is the feasibility of physical path (e.g., robot arm movement), which is marked yellow if there is an obstacle; and the safety compliance is the degree of violation of safety regulations, which is marked red if it violates the ban.

[0106] In an embodiment, the instruction feasibility marker includes green instructions that pass all constraints, yellow instructions that violate soft constraints, and red instructions that violate hard constraints, and step S50 specifically includes the steps of:

[0107] S51: Perform hierarchical processing according to the instruction feasibility marker, green instructions are directly entered into the execution queue, yellow instructions trigger the constraint optimizer, and red instructions start the safety override engine;

[0108] S52: Output the final recommended instruction with a correction record and execute.

[0109] In this embodiment, the initial recommendation for the (smart morning wake-up scene) is: main instruction: brew coffee + toaster; alternative instruction: play news. Knowledge graph verification: electrical dimension: total power of breakfast appliance group exceeds limit (yellow marker); safety dimension: coffee machine water tank is empty (red marker). Conflict resolution: power optimization: delay toaster start by 10 minutes; safety replacement: coffee machine -> instant hot water dispenser. Final execution sequence: "instruction": "heat water", "source": "safety replacement"; "instruction": "play news", "source": "original alternative"; "instruction": "toaster", "source": "delayed execution".

[0110] It should be understood that the size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0111] In an embodiment, a home intelligent recommendation system based on cross-modal fusion is provided, which corresponds to the home intelligent recommendation method based on cross-modal fusion in the above embodiment. As shown in Figure 4 The system comprises:

[0112] A data collection module is configured to collect user modal data, environment modal data and device behavior modal data in real time.

[0113] A data processing module is configured to convert the environment modal data into quantum state amplitude features through a quantum encoder, and obtain user behavior phase features through a time sequence feature extractor.

[0114] A cross-modal fusion module is configured to perform cross-modal fusion on the quantum state amplitude features and the behavior phase features based on a quantum entanglement operation to generate a dynamic fusion vector.

[0115] An instruction recommendation module is configured to input the dynamic fusion vector into a hybrid decision engine, and output a recommended instruction set. The physical feasibility of the recommended instruction set is verified based on a device state knowledge graph.

[0116] An execution module is configured to generate a final recommended instruction according to the verification result and execute the final recommended instruction.

[0117] The specific limitations of the home intelligent recommendation system based on cross-modal fusion can be referred to the limitations of the home intelligent recommendation method based on cross-modal fusion in the above, which will not be repeated here. Each module in the home intelligent recommendation system based on cross-modal fusion can be realized by software, hardware and their combinations in whole or in part. Each module can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory in the computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0118] In an embodiment, a computer device is provided, which can be a server. The internal structure diagram of the computer device can be as shown in Figure 5 The computer device comprises a processor, a memory, a network interface and a database connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is configured to communicate with an external terminal through a network connection. The computer program is executed by the processor to implement a home intelligent recommendation method based on cross-modal fusion.

[0119] In one embodiment, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor implementing a home intelligent recommendation method based on cross-modal fusion when executing the computer program.

[0120] In one embodiment, a computer readable storage medium is provided, having stored thereon a computer program, the computer program being executable by a processor to implement a home intelligent recommendation method based on cross-modal fusion.

[0121] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing related hardware, and the computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. Any reference to memory, storage, database or other medium used in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0122] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is exemplified, and in actual application, the above-mentioned functions can be completed by several functional units and modules according to needs, that is, the internal structure of the device is divided into several functional units or modules to complete all or part of the above-mentioned functions.

[0123] The above-described embodiments are only used to illustrate the technical solutions of the present application, but not limit them; although the present application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A home intelligent recommendation method based on cross-modal fusion, characterized in that, The method comprises the steps of: collecting user modal data, environment modal data and device behavior modal data in real time; transforming the environment modal data into quantum state amplitude features through a quantum encoder, and obtaining user behavior phase features through a time sequence feature extractor; cross-modal fusion of the quantum state amplitude features and the behavior phase features based on quantum entanglement operations to generate a dynamic fusion vector; inputting the dynamic fusion vector into a hybrid decision engine, and outputting a recommended instruction set by the hybrid decision engine, and verifying the physical feasibility of the recommended instruction set based on a device state knowledge graph; generating a final recommended instruction according to the verification result and executing the final recommended instruction; the step of cross-modal fusion of the quantum state amplitude features and the behavior phase features based on quantum entanglement operations to generate a dynamic fusion vector comprises the steps of: loading the quantum state amplitude features into a first quantum feature carrier, loading the behavior phase features into a second quantum feature carrier, performing quantum entanglement initialization operations on the two carriers to establish a quantum state correlation channel; monitoring real-time changes of the behavior phase features, triggering a phase rotation operation, and dynamically adjusting the entanglement strength based on the environment amplitude features; using the quantum state amplitude features of the first quantum feature carrier as a control bit, performing conditional rotation on the behavior phase features of the second quantum feature carrier to generate a fusion quantum state; projecting the fusion quantum state to a classical data space to verify the dual-modal carrying capacity of the vector, and generating a dynamic fusion vector carrying the quantum state amplitude features and the behavior phase features; the step of monitoring real-time changes of the behavior phase features, triggering a phase rotation operation, and dynamically adjusting the entanglement strength based on the environment amplitude features comprises the steps of: continuously receiving user behavior time sequence feature streams, and calculating the phase offset of adjacent time windows; when a positive mutation pattern is identified, starting a phase accelerator to perform positive rotation to improve the priority of the feature vector in decision-making; when a negative drift pattern is identified, triggering a phase compensator to perform inverse rotation to reduce the confidence weight of the feature vector; analyzing the environment amplitude features in real time, adjusting the entanglement energy level through a quantum coupler, when the amplitude fluctuation is below a preset threshold, maintaining low entanglement strength and preserving the independence of each modal feature; when a continuous amplitude step is detected, switching to high entanglement strength and forcing synchronous update of dual-modal features; the step of using the quantum state amplitude features of the first quantum feature carrier as a control bit, performing conditional rotation on the behavior phase features of the second quantum feature carrier to generate a fusion quantum state comprises the steps of: extracting environment data stability indicators and burst event marker strength information from the quantum state amplitude features to generate quantum control signals bound to the quantum state amplitude features; inputting the quantum control signals into a second quantum feature carrier interface, activating the rotation mechanism according to the control signal type, triggering micro-amplitude phase calibration for stable control signals, and activating large-angle phase reset for mutation suppression signals; performing rotation operations to enhance behavior feature weights through positive rotation and suppress abnormal behavior interference through inverse rotation; In the process of performing the rotation operation, the quantum entanglement is maintained synchronously, a fused quantum state including a main feature layer, a modulation layer and a correlation layer is generated, the main feature layer inherits the environmental amplitude distribution, the modulation layer carries the phase information after rotation, and the correlation layer records the interaction trajectory of the two carriers.

2. The home intelligent recommendation method based on cross-modal fusion according to claim 1, characterized in that, The step of converting the environmental modal data into quantum state amplitude features through a quantum encoder, and simultaneously acquiring user behavior phase features through a time sequence feature extractor, specifically includes the following steps: Receiving continuous physical quantity data streams from environmental sensors, including temperature, light intensity and device running state, and discretizing the physical quantity values into environmental quantum state probability distribution intervals that can be processed by quantum computing; Capturing user operation event sequences, including voice instruction trigger time points and device interaction action intervals, constructing a behavior time axis, and extracting user behavior phase features. 3.The home intelligent recommendation method based on cross-modal fusion according to claim 1, characterized in that, The step of inputting the dynamic fusion vector into a hybrid decision engine, the hybrid decision engine outputting a recommended instruction set, and verifying the physical feasibility of the recommended instruction set based on a device state knowledge graph, specifically includes the following steps: Extracting environmental state encodings containing abnormal event markers and user intent probability distributions containing behavior mode confidences from the dynamic fusion vector, and generating a dual-channel input data stream; Inputting the environmental state encodings into a context perception module to identify the current home scene type and predict the environmental evolution trend; Inputting the user intent probability distribution into an intent reasoning module to analyze the user's explicit operation demand and mine the user's implicit behavior preference, and outputting a recommended instruction set with confidence; Loading the recommended instruction set into a symbolic rule verifier, and performing three-dimensional verification with the device state knowledge graph to generate instruction feasibility markers, the three-dimensional verification including electrical compatibility, spatial accessibility and safety compliance verification.

4. The home intelligent recommendation method based on cross-modal fusion according to claim 3, characterized in that, The instruction feasibility markers include green instructions that pass all constraints, yellow instructions that violate soft constraints, and red instructions that violate hard constraints, and the step of generating a final recommended instruction according to the verification result and executing it, specifically includes the following steps: According to the instruction feasibility markers, perform hierarchical processing, green instructions directly enter the execution queue, yellow instructions trigger the constraint optimizer, and red instructions start the safety override engine; Output the final recommended instruction with a correction record and execute it.

5. A home intelligent recommendation system based on cross-modal fusion, configured to perform the home intelligent recommendation method based on cross-modal fusion according to any one of claims 1 to 4. Comprise: A data collection module for collecting user modal data, environmental modal data and device behavior modal data in real time; A data processing module for converting the environmental modal data into quantum state amplitude features through a quantum encoder, and simultaneously acquiring user behavior phase features through a time sequence feature extractor; A cross-modal fusion module for cross-modal fusion of the quantum state amplitude features and behavior phase features based on quantum entanglement operation to generate a dynamic fusion vector; An instruction recommendation module for inputting the dynamic fusion vector into a hybrid decision engine, the hybrid decision engine outputting a recommended instruction set, and verifying the physical feasibility of the recommended instruction set based on a device state knowledge graph; An execution module for generating a final recommended instruction according to the verification result and executing it.

6. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The computer program is executed by the processor to implement the steps of the method for smart home recommendation based on cross-modal fusion according to any one of claims 1 to 4.

7. A computer-readable storage medium storing a computer program, wherein the computer program comprises the following steps of: receiving a request for a resource from a client; determining whether the client is authorized to access the resource; and if the client is authorized to access the resource, providing the resource to the client. The computer program is executed by the processor to implement the steps of the method for smart home recommendation based on cross-modal fusion according to any one of claims 1 to 4.

Images