Intelligent fragrance regulation system and method based on environmental perception and emotion analysis

Abstract

This invention discloses an intelligent fragrance control system and method based on environmental perception and emotion analysis, which has the following advantages: 1) It realizes multimodal and cross-sensor environmental and emotion perception, ensuring more comprehensive perception and higher robustness; 2) It systematically incorporates the mechanism of aromatic substances and the dose-time effect of inhalation into decision-making, improving the scientific nature and safety of fragrance strategies; 3) It adopts an online / offline learning mechanism based on time-series feedback to achieve adaptive optimization of fragrance control, improving the effect of emotion regulation and atmosphere creation; 4) It supports scenario-based and personalized configuration (such as meetings, families, treatments, etc.), taking into account privacy protection and user preferences, and providing controllable and verifiable fragrance services; 5) Through real-time data analysis and decision feedback, it enhances the system's responsiveness to dynamic environmental and emotional changes, ensuring the personalization and continuity of the fragrance experience.

Description

Technical Field

[0001] This invention relates to the field of intelligent aromatherapy control technology, specifically to an intelligent aromatherapy control system and method based on environmental perception and emotion analysis. Background Technology

[0002] Currently, the application of intelligent fragrance control technology is still in its early stages. Traditional fragrance release systems typically rely on manually set buttons, timers, or simple scene presets to release essential oils. This singular control method results in a lack of real-time perception and dynamic adjustment to environmental changes and user emotions, making it unable to effectively adapt to complex usage scenarios such as meetings, multi-member family environments, or medical settings. In these situations, users' needs and emotional states may change frequently, and traditional fragrance systems struggle to provide a personalized and continuous fragrance experience.

[0003] While some smart fragrance solutions have emerged on the market, these solutions are often limited to using a single sensor or a single modality for human emotion determination, such as assessing emotions solely through facial expressions or voice recognition. This limitation prevents the system from fully understanding and responding to the user's true emotional state, making it difficult to achieve personalized and contextualized fragrance effects. Furthermore, existing systems lack a comprehensive understanding of the mechanisms of action of aromatic substances and their interactions with environmental factors such as air quality, light, temperature, and humidity. This includes a lack of systematic application of research on the effects of chemical components on physiology and subjective perception, as well as the effects of inhalation dosage on time series. This lack of knowledge leads to instability and safety issues in fragrance effects, potentially causing discomfort or negative emotions during use, further reducing user satisfaction with the fragrance experience.

[0004] Building an intelligent fragrance control system capable of effectively responding to environmental changes and user emotions is crucial. Simply relying on traditional button settings, timers, or single-sensor emotion judgment is insufficient to meet the personalized needs of complex scenarios. A more reasonable approach should integrate multiple environmental parameters (such as temperature, humidity, and air quality), user emotion data, and the interaction of fragrance components. This involves fusing sensor technology, emotion analysis algorithms, and environmental monitoring systems, employing multi-layered real-time monitoring, data analysis, and intelligent adjustment strategies to comprehensively evaluate the effectiveness and safety of fragrance release. The multi-dimensional perception and intelligent control approach is necessary in several ways: first, it compensates for the shortcomings of traditional fragrance systems in dynamic response capabilities, improving the system's adaptability to environmental changes; second, it enhances the understanding and prediction of the complex relationship between fragrance effects and user emotions; and third, it provides quantifiable evidence for fragrance formulation optimization and user experience improvement, thereby promoting the modernization, standardization, and personalized application of intelligent fragrance control technology. Summary of the Invention

[0005] One technical problem this application aims to solve is to overcome the shortcomings of the above-mentioned related technologies and provide an intelligent fragrance control system and method based on environmental perception and emotion analysis. By perceiving environmental variables and emotional states through multimodal sensing, it makes intelligent formulation and dosage decisions based on the mechanism of action of aromatic substances and the knowledge base of olfactory effects, and optimizes the fragrance release strategy through a closed-loop real-time feedback and continuous learning mechanism, thereby achieving intelligent, controllable and personalized adjustment of environmental atmosphere and emotions, and enhancing the fragrance experience.

[0006] The technical solution adopted by this intelligent fragrance control system to solve the technical problem is as follows: an intelligent fragrance control system based on environmental perception and emotion analysis, comprising: Environmental sensing module: Connects to various environmental sensors to collect and monitor environmental conditions in real time and acquire environmental parameters; Emotion perception module: Connects to multiple modal sensors to comprehensively perceive and analyze the emotions and states of people present, and acquire emotion data; Aroma Knowledge Base Module: Integrates the chemical composition, mechanism of action, olfactory dose-time response, synergistic or antagonistic relationships, safe dose limits, and allergy / contraindication information of aromatic substances to construct a structured knowledge representation to support aroma regulation decisions; Intelligent Decision Module: Constructs a fragrance regulation decision model or rule base, comprehensively analyzes real-time collected environmental parameters and emotional data, combines an aroma knowledge base, uses a time-series-based strategy optimizer to weigh multiple objectives, generates fragrance regulation strategies, and outputs control commands; the objectives include, but are not limited to, emotion regulation effect, indoor air quality, safety, and energy consumption. Fragrance release module: Connects to the fragrance release control system, supports the mixing of different formulas and controlled dosage release; according to the received control instructions, it implements precise fragrance release through a multi-channel essential oil / fragrance storage, atomization / vaporization / heating release mechanism that can be controlled by channels; Feedback and Learning Module: Based on monitored environmental parameters and emotional data, the module analyzes short-term and long-term effects of fragrance release and time-series data to update the fragrance control decision model or rule base online or offline, thereby optimizing the fragrance control strategy.

[0007] Preferably, the strategy optimizer is based on a machine learning model with a reward function, a reinforcement learning model, a Bayesian optimization method, or a hybrid strategy based on rules and historical experience. It uses a rule engine to select suitable fragrance ingredients by preset rules, calculates a score for each fragrance ingredient, and selects the fragrance ingredient with the highest score as the basis for the formula.

[0008] Preferably, the intelligent decision-making module further includes: Emotion / Environment Fusion Analysis Module: Denoises, cleans, extracts and aligns features from collected environmental parameters and emotion data, and performs multimodal fusion to obtain the current environment-emotion representation; Scene recognition module: Based on the current environment-emotion representation fused from multimodal fusion, identify the current scene type and set control targets and constraints; Formula generation module: Based on the identified scene type and the set control objectives, and using the aroma knowledge base and strategy optimizer, select one or more fragrance ingredients to combine and generate the final fragrance formula. Dosage / Schedule Planning Module: Based on the generated fragrance formula, calculate the release dosage and timing of each essential oil / fragrance agent and output control commands.

[0009] Preferably, the fragrance release control system is based on a proportional control algorithm, specifically including: Multi-channel essential oil compartment, each channel holds one type of essential oil or fragrance; Precision metering pumps or micro-nebulizers are used to mix different essential oils or fragrances in proportion and ensure accurate release of fragrance dosage; The fan / airflow guiding module is responsible for guiding the fragrance airflow. It uses a fluid dynamics model to design the fan speed and airflow direction to ensure that the fragrance is evenly distributed in the indoor environment. The airflow distribution valve is used to control the release direction and intensity of fragrance to adapt to different scenarios and user needs. It uses a PID control algorithm to adjust the valve opening to achieve precise airflow control. The communication module supports receiving commands locally or in the cloud, and transmitting status information, remaining amount and actual release dose data in real time, as well as receiving user feedback. The fragrance release control system also adjusts the fragrance release strategy in real time based on user feedback using simple scoring rules.

[0010] The technical solution adopted by this intelligent fragrance control method to solve the technical problem is as follows: an intelligent fragrance control method, which adopts the intelligent fragrance control system based on environmental perception and emotion analysis as described above, specifically including the following steps: Step S100 Environment and Emotion Perception: Collect environmental parameters in real time through the environment perception module, and collect emotion data in real time through the emotion perception module; Step S200: Data preprocessing and multimodal fusion: The collected environmental parameters and emotion data are denoised, cleaned, feature extracted and aligned, and multimodal fusion is performed to obtain the current environment-emotion representation; Step S300 Scene recognition and target setting: Determine the scene category based on historical context and current environment-emotion representation, and determine the control target and constraints for this operation; Step S400: Formula and Dosage Calculation: Based on the aroma knowledge base and strategy optimizer, generate the fragrance formula and release dosage / timing, and output control instructions; Step S500 Execution and Release: The control command is sent to the fragrance release module and executed; during the execution process, the fragrance can be released in stages according to a plan or adjusted in real time. Step S600 Feedback Collection and Model Update: Continuously collect environmental and emotional responses after release, evaluate the regulation effect, and update and optimize the fragrance regulation decision model or rule base based on the evaluation results to form a closed-loop learning.

[0011] Preferably, step S200, data preprocessing and multimodal fusion, specifically includes: Step S210: Denoise the collected environmental parameters and emotional data to improve data quality; specifically including, Step S211: Apply a low-pass filter or median filter to remove high-frequency noise; Step S212: Smooth the surface using a moving average filter; Step S213: Clean the collected environmental parameters and emotion data, including the following operations: Missing value handling: Identify and fill in missing data using mean imputation, linear interpolation or other suitable methods; Outlier detection: Detect and remove outliers using statistical analysis methods to ensure data quality and reliability; Step S214: Standardize the data processed in step S213 to eliminate the influence of different units and dimensions. Through min-max normalization or Z-score standardization, transform the values ​​of all environmental parameters to the same range to facilitate subsequent analysis. Step S220: Data from different environmental sensors are fused using weighted averaging and principal component analysis techniques to generate a comprehensive environmental status assessment index, providing more accurate environmental information; specifically including, Step S221: Calculate a confidence score scoreParam for each environmental parameter, scoreParam∈[0, 1]. This confidence score is based on the following factors: sensor performance, data acquisition frequency, and reliability of historical records; using a weighted scoring algorithm: in, Scoring for each environmental parameter, The corresponding weights; Step S222: Output a table containing standardized and fused environmental parameters to provide a real-time environmental status report; this data will be used by the intelligent decision-making unit to support the generation and optimization of fragrance control strategies; Step S230: Extract key features from video and audio data, including: facial features: extract key points of facial expressions and calculate emotion scores using a sentiment analysis model; speech features: analyze the timbre, rhythm, and tone of speech to identify speech emotions and keywords; specifically including, Step S231: Facial feature extraction: Use a deep learning model to extract facial expression features. The sentiment score is calculated by a trained sentiment analysis model, and the probability distribution of the sentiment category is output. Step S232: Speech feature extraction: Analyze the timbre, rhythm, and tone of the speech to identify emotions and keywords; use Mel-frequency cepstral coefficients (MFCCs) to extract audio features; Step S233: Analyze the physiological data provided by the wearable device, including: Heart rate and HRV: assess the relationship between emotional state and heart rate variability, use time-domain analysis and frequency-domain analysis to assess heart rate variability and determine emotional fluctuations; Skin conductance response: monitor changes in emotional state by calculating skin conductance response, analyze skin conductance data, and detect the correlation between the participant's physiological response and emotional state; Step S234: Perform localized preprocessing of privacy-sensitive information at the data acquisition layer, extract and store only localized facial features, avoid transmitting raw video data to protect user privacy. Facial feature extraction algorithms can be used to store only feature point coordinates instead of images. Step S235: Output a comprehensive emotion analysis report. The emotion data includes each participant's emotion score, emotion state change trend, and overall emotion assessment. This data will support the subsequent intelligent decision-making module to formulate corresponding fragrance control strategies. Step S240: Fusing data from different sensors to generate a comprehensive environment-emotion representation; specifically including, Step S241: Perform feature fusion using weighted average method and principal component analysis; Step S242: Assign weights to different modalities using an attention mechanism; Step S250: Generate the current environment-emotion representation and output the fused data for subsequent scene recognition and target setting.

[0012] Preferably, step S300, scene recognition and target setting, specifically includes: Step S310: Based on the collected environmental parameters and emotion data, identify the current scene type; specifically including, Step S311: Select key features of environmental parameters and sentiment data from the multimodal fusion results; Step S312: Use Support Vector Machine (SVM) and Decision Tree to classify the scene. Classify the current features using the trained model and find the best splitting hyperplane by maximizing the margin. Step S313: Scene determination: Based on the classification results, determine the category of the current scene as meeting, family gathering, sleep / rest, or treatment / rehabilitation; Step S320: Based on the identified scene category, set the control objectives and constraints; specifically including, Step S321: Set the expected emotional state according to the scene type; Step S322: Ensure that the release dose of the fragrance meets safety standards; Step S323: Set the fragrance release time according to the scene type and event duration; Step S324: Adjust the fragrance formula and dosage based on user feedback and preference settings; Step S330: Integrate the identified scene categories and the set control targets and constraints to provide a basis for subsequent fragrance formulation and dosage calculation.

[0013] Preferably, step S400, formulation and dosage calculation, specifically includes: Step S410: Based on the scene recognition results and control objectives, select appropriate essential oils or fragrances from the aroma knowledge base; specifically including, Step S411: Extract fragrance ingredients and their characteristics that match the current scenario and objective from the fragrance knowledge base, including chemical components, subjective effects and safe dosage; Step S412: Select suitable fragrance ingredients according to preset rules, calculate a score for each fragrance ingredient, and select the ingredient with the highest score as the basis for the formula: in, , , It's weight. , and These are the effectiveness, safety, and user preference ratings of the fragrance ingredients; Step S413: Select one or more fragrance ingredients to combine and form the final fragrance formula; Step S420: Based on the generated fragrance formula, calculate the release dosage and timing of each essential oil / fragrance agent; specifically including, Step S421: Calculate the release dose of each ingredient based on the proportions of the ingredients in the fragrance formula; assuming the total release dose is... The proportion of each ingredient in the formula is: The dosage of each component is then calculated as follows: in, It is the first Dosage of essential oils; Step S422: Ensure the calculated dose does not exceed the safe range, and verify using the following formula: Step S423: Based on the scene requirements and control objectives, set the release sequence of the fragrance and adopt a time-segmented release strategy: in, It is the incremental time, set to an appropriate value that is adjusted according to the needs of the scenario; Step S430: Integrate the generated fragrance formula and its dosage and release timing information into an output object.

[0014] Preferably, the execution and release of step S500 specifically includes: Step S510: Multi-channel essential oil compartment, each channel stores one type of essential oil or fragrance to facilitate flexible blending of various fragrances; Step S520: A precision metering pump or micro-needle is used to mix different essential oils in proportion and ensure accurate release of fragrance dosage; a proportioning control algorithm is used to calculate the required release of fragrance dosage to ensure mixing according to the formula. in, It is the first Dosage of the fragrance, It is the total release. That is the total amount of fragrance. It is the first The proportions of various fragrances in the formula; specifically including, Step S521: The fan / airflow guiding module is responsible for guiding the fragrance airflow. A fluid dynamics model is used to design the fan speed and airflow direction to achieve uniform distribution; Bernoulli's equation is used to calculate the airflow velocity. in, It's pressure. It is air density. It is the airflow speed. It is gravitational acceleration. It is height; Step S522: The airflow distribution valve is used to control the release direction and intensity of the fragrance to adapt to different scenarios and user needs. A PID control algorithm is used to adjust the valve opening to achieve precise airflow control. in, It is a control signal. It is the error between the expected output and the actual output. , and These are proportional, integral, and differential gains, respectively. Step S530: The device supports receiving commands locally or in the cloud, and can transmit status information, remaining amount and actual release dose data in real time; Step S540: Monitor environmental and user feedback during the release process; If the security index is lower than the preset threshold, the release strategy will be automatically adjusted or the release will be stopped. Step S550: Use a simple rating system to collect user feedback during the release process and adjust the fragrance release strategy in real time; in, It's about adjusting sensitivity. This is the user's current feedback. It is the expected satisfaction level.

[0015] Preferably, step S600, which involves feedback acquisition and model updating, specifically includes: Step S610: By analyzing the feedback data, identify the effectiveness of fragrance control and user satisfaction, continuously optimize the fragrance release strategy and improve the user experience; the feedback data includes, but is not limited to, mood change curves, changes in environmental parameters, explicit user evaluations, and whether a safety alarm is triggered; Step S620: Use the collected data to train and improve the machine learning model to enhance the system's adaptability and intelligence level, and ensure that it can provide the best fragrance control solution in different environments and scenarios. Step S630: A time-series-based policy optimizer is used to weigh multiple objectives and generate a fragrance modulation strategy; specifically, it includes, Step S631: Optimize the fragrance control strategy using a reinforcement learning model, defining the state, action, and reward function: in, This is the current state. This is the current action. It is the reward received. It is a discount factor; Step S632: Optimize the fragrance release strategy using Bayesian optimization method, select the optimal parameter configuration to maximize the objective function, and establish the prior distribution of the objective function through a probability model; Step S633: Combine historical data and user feedback to establish a rule-based strategy, and generate control strategies through a rule engine; Step S640: Feed the evaluation results and user feedback into the decision model to form a cyclical update mechanism, enabling the system to continuously optimize the fragrance control strategy over time.

[0016] Compared with related technologies, this intelligent fragrance control system and method have the following advantages: 1) Achieve multimodal, cross-sensor environmental and emotional perception, ensuring more comprehensive and robust perception; 2) Systematically incorporate the mechanisms of aromatic substances and the dose-time effect of inhalation into decision-making, enhancing the scientific rigor and safety of fragrance strategies; 3) Employ an online / offline learning mechanism based on time-series feedback to achieve adaptive optimization of fragrance control, improving the effectiveness of emotion regulation and atmosphere creation; 4) Support scenario-based and personalized configurations (such as meetings, families, and treatments), balancing privacy protection and user preferences, providing controllable and verifiable fragrance services; 5) Enhance the system's responsiveness to dynamic environmental and emotional changes through real-time data analysis and decision feedback, ensuring the personalization and continuity of the fragrance experience. Attached Figure Description

[0017] Figure 1 This is a structural diagram of the intelligent fragrance control system of this application.

[0018] Figure 2 This is a flowchart of the intelligent fragrance control method of this application.

[0019] Figure 3 This is a control diagram of the fragrance release module and fragrance release control system of this application.

[0020] Figure 4 This is a schematic diagram of the data structure of the aroma knowledge base in this application.

[0021] Figure 5 This is a flowchart of the fragrance release strategy update cycle based on time-series feedback in this application. Detailed Implementation

[0022] First, those skilled in the art should understand that these embodiments are merely used to explain the technical principles of the embodiments of this application and are not intended to limit the scope of protection of the embodiments of this application. Those skilled in the art can make adjustments as needed to adapt to specific application scenarios.

[0023] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0024] This invention designs a method such as Figure 1 The intelligent fragrance control system based on environmental perception and emotion analysis shown includes: Environmental sensing module: Connects to various environmental sensors to collect and monitor environmental conditions in real time, acquiring various environmental parameters, including air quality (such as PM2.5, CO2, VOC), temperature, humidity, light intensity, and noise level; Emotion perception module: Connected to multiple modal sensors, it comprehensively perceives and analyzes the emotions and states of people present, and acquires emotional data; the modal sensors include video (facial expressions, posture, number of people and distribution), audio (voice emotion, volume, keywords), physiological signals (heart rate, heart rate variability, skin conductance, and electroencephalogram signals collected by wearable devices) and other modal sensors; Aroma Knowledge Base Module: Integrates the chemical composition, mechanism of action, olfactory dose-time response, synergistic or antagonistic relationships, safe dose limits, and allergy / contraindication information of aromatic substances to construct a structured knowledge representation to support aroma regulation decisions; Intelligent Decision Module: Constructs a fragrance regulation decision model or rule base, comprehensively analyzes real-time collected environmental parameters and emotional data, combines an aroma knowledge base, and utilizes a time-series-based strategy optimizer (e.g., machine learning model with reward function, reinforcement learning, Bayesian optimization, or a hybrid strategy based on rules and historical experience) to weigh multiple objectives and generate fragrance regulation strategies, outputting control commands; the objectives include, but are not limited to, emotion regulation effect, indoor air quality, safety, and energy consumption; includes an emotion / environment fusion analysis module, a scene recognition module, a formula generation module, and a dosage / scheduling planning module; Fragrance release module: Connected to the fragrance release control system, it supports the mixing of different formulas and controlled dosage release; according to control instructions, it implements precise fragrance release through a multi-channel essential oil / fragrance storage, atomization / vaporization / heating release mechanism that can be controlled by channels; Feedback and Learning Module: Based on the collected environmental parameters and emotional data after fragrance release, the module analyzes short-term and long-term effects of fragrance release (such as changes in the environment and changes in people's emotions) and time-series data to update the fragrance control decision model or rule base online or offline, thereby optimizing the fragrance control strategy.

[0025] Preferably, the strategy optimizer is based on a machine learning model with a reward function, a reinforcement learning model, a Bayesian optimization method, or a hybrid strategy based on rules and historical experience. It uses a rule engine to select suitable fragrance ingredients by preset rules, calculates a score for each fragrance ingredient, and selects the fragrance ingredient with the highest score as the basis for the formula.

[0026] Preferably, the intelligent decision-making module further includes: Emotion / Environment Fusion Analysis Module: Denoises, cleans, extracts and aligns features from collected environmental parameters and emotion data, and performs multimodal fusion to obtain the current environment-emotion representation; Scene recognition module: Based on the current environment-emotion representation fused from multimodal fusion, identify the current scene type and set control targets and constraints; Formula generation module: Based on the identified scene type and the set control objectives, and using the aroma knowledge base and strategy optimizer, select one or more fragrance ingredients to combine and generate the final fragrance formula. Dosage / Schedule Planning Module: Based on the generated fragrance formula, calculate the release dosage and timing of each essential oil / fragrance agent and output control commands.

[0027] Preferably, the fragrance release control system is based on a proportional control algorithm, specifically including: Multi-channel essential oil compartment, each channel holds one type of essential oil or fragrance; Precision metering pumps or micro-nebulizers are used to mix different essential oils or fragrances in proportion and ensure accurate release of fragrance dosage; The fan / airflow guiding module is responsible for guiding the fragrance airflow. It uses a fluid dynamics model to design the fan speed and airflow direction to ensure that the fragrance is evenly distributed in the indoor environment. The airflow distribution valve is used to control the release direction and intensity of fragrance to adapt to different scenarios and user needs. It uses a PID control algorithm to adjust the valve opening to achieve precise airflow control. The communication module supports receiving commands locally or in the cloud, and transmitting status information, remaining amount and actual release dose data in real time, as well as receiving user feedback. The fragrance release control system also adjusts the fragrance release strategy in real time based on user feedback using simple scoring rules.

[0028] The intelligent fragrance control method based on environmental perception and emotion analysis, implemented using the aforementioned intelligent fragrance control system, includes the following steps: Step S100: Environment and Emotion Perception: Collect real-time data through the environment perception module and the emotion perception module.

[0029] Step S110: Utilize multiple sensors, including an air quality sensor module (comprising PM2.5, CO2, and VOC sensors), temperature and humidity sensors, light intensity sensors, and noise sensors, to collect and monitor environmental parameters in real time. These sensors can be placed in different locations indoors or integrated into a comprehensive environmental monitor to ensure comprehensive coverage and accurate monitoring. This unit provides real-time feedback on environmental conditions, offering necessary data support for subsequent fragrance control decisions.

[0030] Step S120: Utilizing multiple sensor technologies, including a camera (for face detection, pose estimation, and people counting), a microphone array (supporting voice activity detection, voice emotion recognition, and keyword detection), a wearable device interface (reading physiological data such as heart rate, heart rate variability (HRV), conductance of skin, and sleep status), and an optional EEG acquisition device interface (for acquiring electroencephalogram (EEG) data), this unit aims to monitor and analyze the emotional state of those present in real time.

[0031] Step S200: Data preprocessing and multimodal fusion: The collected data is denoised, cleaned, feature extracted and aligned, and multimodal fusion is performed to obtain the current environment-emotion representation.

[0032] Step S200 specifically includes: Step S210: Denoise the collected environmental parameters and emotional data to improve data quality.

[0033] Step S211: First, apply a low-pass filter or a median filter to remove high-frequency noise.

[0034] in, It is the denoised signal. It is the original signal. It refers to the window size.

[0035] Step S212: Smooth the surface using a moving average filter.

[0036] in, It refers to the window size.

[0037] Step S213: Clean the collected environmental and emotional data, including the following operations: Missing value handling: Identify and impute missing data, using mean imputation, linear interpolation, or other suitable methods. Outlier detection: Detect and remove outliers using statistical analysis methods (such as Z-score, IQR) to ensure data quality and reliability.

[0038] Step S214: Standardize the processed data to eliminate the influence of different units and dimensions. Transform all environmental parameter values ​​to the same range (e.g., [0, 1]) using min-max normalization or Z-score standardization to facilitate subsequent analysis.

[0039] Step S220: The data from different environmental sensors are fused, and a comprehensive environmental status assessment index is generated using weighted averaging and principal component analysis (PCA) techniques to provide more accurate environmental information.

[0040] 1) Weighted Average: in As weight, These are the measured values ​​from each sensor.

[0041] 2) Principal Component Analysis (PCA): This involves decomposing the covariance matrix C into eigenvalues ​​to obtain the principal components. Selecting from previous selections by eigenvalue sorting Each feature vector forms a new feature space.

[0042] Step S221: Calculate a confidence score (scoreParam ∈ [0, 1]) for each environmental parameter, based on factors such as sensor performance (e.g., accuracy, stability), data acquisition frequency, and reliability of historical records. A weighted scoring algorithm can be used. in, Scoring for each parameter, The corresponding weights.

[0043] Step S222: Finally, output a table containing standardized and fused environmental parameters, providing a real-time environmental status report. This data will be used by the intelligent decision-making unit to support the generation and optimization of fragrance control strategies.

[0044] Step S230: Extract key features from video and audio data, including: Facial features: Extract key points of facial expressions and calculate emotion scores using a sentiment analysis model. Speech features: Analyze the timbre, rhythm, and tone of speech to identify speech emotions and keywords.

[0045] Step S231: Facial Feature Extraction: Deep learning models such as Convolutional Neural Networks (CNNs) are used to extract facial expression features. Sentiment scores can be calculated using a trained sentiment analysis model, outputting the probability distribution of sentiment categories. in, From the input image Features extracted from and These are model parameters.

[0046] Step S232: Speech Feature Extraction: Analyze the timbre, rhythm, and pitch of the speech to identify emotion and keywords. Mel-frequency cepstral coefficients (MFCCs) can be used to extract audio features, calculated as follows: in, It is the spectrum of the audio signal.

[0047] Step S233: Analyze the physiological data provided by the wearable device, including: Heart rate and HRV: assess the relationship between emotional state and heart rate variability to determine emotional fluctuations. Skin conductance response: analyze skin conductance data to detect the correlation between the participant's physiological response and emotional state.

[0048] 1) Heart Rate and HRV Analysis: Time-domain and frequency-domain analyses are used to assess heart rate variability. The standard deviation formula in time-domain analysis is: in, It is the interval between adjacent heartbeats. It is the average heartbeat interval.

[0049] 2) Skin conductance response analysis: Changes in emotional state are monitored by calculating the skin conductance response (SCR). The formula for calculating SCR is: in, It's a change in skin conductivity. It is the baseline conductivity.

[0050] Step S234: Perform localized preprocessing of privacy-sensitive information at the data acquisition layer, extracting and storing only localized facial features to avoid transmitting raw video data and protect user privacy. A facial feature extraction algorithm can be used, storing only feature point coordinates instead of the image. The feature extraction formula can be expressed as: in, It is the extracted feature set. These are the coordinates of each key point.

[0051] Step S235: Output a comprehensive emotion analysis report, including each participant's emotion score, emotion state change trends, and overall emotion assessment. This data will support the subsequent intelligent decision-making module in developing corresponding fragrance control strategies.

[0052] Step S240: Fuse data from different sensors to generate a comprehensive environmental-emotional representation.

[0053] Step S241: Perform feature fusion using weighted average and principal component analysis (PCA).

[0054] 1) Weighted average method: in, These are the characteristics after fusion. It's weight. These are the characteristics of each modality.

[0055] 2) Principal component analysis: Principal components are obtained through eigenvalue decomposition.

[0056] Step S242: Assign weights to different modalities using an attention mechanism.

[0057] in, yes The feature scoring function.

[0058] Step S250: Generate the current environment-emotion representation and output the fused data for subsequent scene recognition and target setting.

[0059] Step S300: Scene recognition and goal setting step: Determine the scene category (such as meeting, family gathering, sleep / rest, treatment / rehabilitation) based on historical context and current representation, and determine the current regulation goals and constraints (emotional goals, safety constraints, duration, personnel preferences, etc.).

[0060] Step S300 specifically includes: Step S310: Identify the current scene type based on the collected environmental and emotional data.

[0061] Step S311: Feature selection: Select key features from the multimodal fusion results, such as environmental parameters (temperature, humidity, noise level) and emotional state (emotional score, emotional change rate).

[0062] Step S312: Classify the scene using Support Vector Machine (SVM) and Decision Tree.

[0063] 1) Support Vector Machine (SVM): This model classifies current features using a pre-trained model. The decision function of an SVM can be expressed as: in, It is a weight vector. It is an eigenvector. This is the bias term. The optimal separating hyperplane is found by maximizing the margin.

[0064] 2) Decision Tree: Using a decision tree model for scene classification, the basic process of constructing a decision tree is as follows: in, It's entropy. It is a feature The Class of samples.

[0065] Step S313: Scene determination: Based on the classification results, determine the category of the current scene, such as meeting, family gathering, sleep / rest, treatment / rehabilitation.

[0066] Step S320: Set control targets and constraints based on the identified scene categories.

[0067] Step S321: Set the expected emotional state according to the type of scenario. For example, in a meeting scenario, the goal is to improve the participants' attention and concentration; in a family gathering, the goal is to create a relaxed and pleasant atmosphere.

[0068] Step S322: Ensure the fragrance release dose meets safety standards. The following formula can be used: in, This is the safe dose. Recommended Dose is the dose recommended based on the aroma knowledge base. User Preference is the dose set by the user according to their personal preferences.

[0069] Step S323: Set the duration of fragrance release based on the scene type and activity duration. For example, in a treatment / rehabilitation scene, continuous fragrance release may be necessary to support the therapeutic effect. The duration can be expressed using the following formula: Where StartTime is the startup time and SessionLength is the duration; Step S324: Adjust the fragrance formula and dosage based on user feedback and preference settings. For example, consider user preferences using the following logic: Among them, AdjustedFormula is the target dose, BaseFormula is the basic dose, and UserPreferenceAdjustment is the preference setting; Step S330: Integrate the identified scene categories and the set control targets and constraints to provide a basis for subsequent fragrance formulation and dosage calculation.

[0070] Step S400: Formula and Dosage Calculation: Based on the aroma knowledge base and strategy optimizer, generate a fragrance formula (select one or more essential oils / fragrances and their proportions) and release dosage / timing; Step S400 specifically includes: Step S410: Based on the scene recognition results and control objectives, select appropriate essential oils or fragrances from the aroma knowledge base.

[0071] Step S411: Extract fragrance ingredients and their characteristics that match the current scenario and objective from the fragrance knowledge base, including chemical components, subjective effects, and safe dosage.

[0072] Step S412: Select suitable fragrance ingredients according to preset rules (such as user preferences, scene type), calculate a score for each fragrance ingredient, and select the ingredient with the highest score as the formula basis: in, , , It's weight. , and These are the effectiveness, safety, and user preference ratings of the fragrance ingredients.

[0073] Step S413: Select one or more fragrance ingredients to combine and form the final fragrance formula.

[0074] Step S420: Calculate the release dosage and timing of each essential oil / fragrance based on the generated fragrance formula.

[0075] Step S421: Calculate the release dose of each ingredient based on the proportions of the ingredients in the fragrance formula. Assume the total release dose is... The proportion of each ingredient in the formula is: The dosage of each component is then calculated as follows: in, It is the first Dosage of essential oils.

[0076] Step S422: Ensure the calculated dose does not exceed the safe range, and verify using the following formula: Step S423: Based on the scene requirements and control objectives, set the release sequence of the fragrance. A time-segmented release strategy can be adopted: ReleaseTiming is the release duration. It is the incremental time, set to an appropriate value that is adjusted according to the needs of the scenario.

[0077] Step S430: Integrate the generated fragrance formula and its dosage and release timing information into an output object.

[0078] Step S500: Execution and Release: The control command is sent to the fragrance release unit and executed; during the execution process, the fragrance can be released in stages according to the plan or adjusted in real time; Step S500 specifically includes: Step S510: Multi-channel essential oil compartment, each channel holds one type of essential oil or fragrance, to facilitate flexible blending of various fragrances.

[0079] Step S520: A precision metering pump or micro-needle is used to mix different essential oils in proportion and ensure accurate release of fragrance dosage. A proportioning control algorithm is used to calculate the desired fragrance dosage, ensuring mixing is performed according to the formula. in, It is the first Dosage of the fragrance, It is the total release. That is the total amount of fragrance. It is the first The proportion of each fragrance in the formula.

[0080] Step S521: The fan / airflow guiding module is responsible for guiding the fragrance airflow to ensure that the fragrance is evenly distributed in the indoor environment. A fluid dynamics model is used to design the fan speed and airflow direction to achieve uniform distribution. For example, Bernoulli's equation is used to calculate the airflow velocity: in, It's pressure. It is air density. It is the airflow speed. It is gravitational acceleration. It's about altitude.

[0081] Step S522: The airflow distribution valve is used to control the release direction and intensity of the fragrance to adapt to different scenarios and user needs. A PID control algorithm is used to adjust the valve opening to achieve precise airflow control. in, It is a control signal. It is the error between the expected output and the actual output. , and These are proportional, integral, and differential gains, respectively.

[0082] Step S530: The device supports receiving commands locally or in the cloud and can transmit status information, remaining amount and actual release dose data in real time to ensure the transparency and controllability of the system.

[0083] Step S540: Monitor the environment and user feedback during the release process to ensure safety and effectiveness.

[0084] If the security index is lower than the preset threshold, the release strategy will be automatically adjusted or the release will be stopped.

[0085] Step S550: Collect user feedback during the release process using a simple rating system (e.g., 1 to 5 points) and adjust the fragrance release strategy in real time.

[0086] in, It's about adjusting sensitivity. This is the user's current feedback. It is the expected satisfaction level.

[0087] Step S600: Feedback Collection and Model Update Step – Continuously collect environmental and emotional responses after release, evaluate the regulation effect, and update and optimize the decision-making model or knowledge base based on the evaluation results to form a closed-loop learning.

[0088] Step S600 specifically includes: Step S610: By analyzing these feedback data (emotional change curves, changes in environmental parameters, explicit user evaluations (such as APP feedback), and whether security alarms are triggered, etc.), the system can identify the effectiveness of fragrance control and user satisfaction, thereby continuously optimizing the fragrance release strategy and improving the user experience.

[0089] 1) Data Analysis Methods: Descriptive statistics and regression analysis can be used to evaluate the effectiveness of fragrance control. For example, by calculating the mean and standard deviation of user satisfaction: in, It is a user satisfaction rating. It refers to the number of samples.

[0090] 2) Sentiment Change Curve Analysis: Time series analysis methods (such as autoregressive moving average models, ARIMA) can be used to model sentiment changes. in, It is time The emotional value, It is a constant term. It is a coefficient. It is a white noise term.

[0091] Step S620: Use the collected data to train and improve the machine learning model to enhance the system's adaptability and intelligence, ensuring that it can provide the best fragrance control solution in different environments and scenarios.

[0092] 1) Machine Learning Model Training: The model is trained using supervised learning algorithms (such as Random Forest, Support Vector Machine, or Deep Learning). Taking Random Forest as an example, the basic training process is as follows: Feature selection: Select features related to fragrance control (such as environmental parameters and user feedback).

[0093] Model training: using the training set To train the model: in, It is a decision tree.

[0094] 2) Model Evaluation: Use cross-validation and performance metrics (such as accuracy, F1 score, mean squared error) to evaluate the model's performance. in, This is a real example. It is a true negative example. It is a false positive example. It is a false negative.

[0095] 3) Online learning: This allows for the implementation of online learning algorithms (such as incremental learning) to continuously update model weights while the system is running. in, These are model parameters. It's the learning rate. It is the loss function.

[0096] Step S630: A time-series-based policy optimizer is used to weigh multiple objectives (mood regulation effect, indoor air quality, safety, energy consumption, etc.) and generate a fragrance regulation strategy.

[0097] Step S631: Optimize the fragrance modulation strategy using a reinforcement learning model (such as Q-learning), defining the state, action, and reward function: in, This is the current state. This is the current action. It is the reward received. It is a discount factor.

[0098] Step S632: Optimize the fragrance release strategy using a Bayesian optimization method, selecting the optimal parameter configuration to maximize the objective function: The prior distribution of the objective function is established using a probabilistic model (such as a Gaussian process).

[0099] Step S633: Combine historical data and user feedback to establish a rule-based strategy, and generate control strategies through a rule engine.

[0100] Step S640: Feed the evaluation results and user feedback into the decision model to form a cyclical update mechanism, enabling the system to continuously optimize the fragrance control strategy over time.

[0101] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An intelligent fragrance control system based on environmental perception and emotion analysis, characterized in that, include: Environmental sensing module: Connects to various environmental sensors to collect and monitor environmental conditions in real time and acquire environmental parameters; Emotion perception module: Connects to multiple modal sensors to comprehensively perceive and analyze the emotions and states of people present, and acquire emotion data; Aroma Knowledge Base Module: Integrates the chemical composition, mechanism of action, olfactory dose-time response, synergistic or antagonistic relationships, safe dose limits, and allergy / contraindication information of aromatic substances to construct a structured knowledge representation to support aroma regulation decisions; Intelligent Decision Module: Constructs a fragrance regulation decision model or rule base, comprehensively analyzes real-time collected environmental parameters and emotional data, combines an aroma knowledge base, uses a time-series-based strategy optimizer to weigh multiple objectives and generate fragrance regulation strategies, and outputs control commands; the objectives include, but are not limited to, emotion regulation effect, indoor air quality, safety and energy consumption; Fragrance release module: Connects to the fragrance release control system, supports the mixing of different formulas and controlled dosage release; according to the received control instructions, it implements precise fragrance release through a multi-channel essential oil / fragrance storage, atomization / vaporization / heating release mechanism that can be controlled by channels; Feedback and Learning Module: Based on monitored environmental parameters and emotional data, the module analyzes short-term and long-term effects of fragrance release and time-series data to update the fragrance control decision model or rule base online or offline, thereby optimizing the fragrance control strategy.

2. The intelligent fragrance control system based on environmental perception and emotion analysis according to claim 1, characterized in that, The strategy optimizer is based on a machine learning model with a reward function, a reinforcement learning model, a Bayesian optimization method, or a hybrid strategy based on rules and historical experience. It uses a rule engine to select suitable fragrance ingredients according to preset rules, calculates a score for each fragrance ingredient, and selects the fragrance ingredient with the highest score as the basis for the formula.

3. The intelligent fragrance control system based on environmental perception and emotion analysis according to claim 2, characterized in that, The intelligent decision-making module also includes: Emotion / Environment Fusion Analysis Module: Denoises, cleans, extracts and aligns features from collected environmental parameters and emotion data, and performs multimodal fusion to obtain the current environment-emotion representation; Scene recognition module: Based on the current environment-emotion representation fused from multimodal fusion, identify the current scene type and set control targets and constraints; Formula generation module: Based on the identified scene type and the set control objectives, and using the aroma knowledge base and strategy optimizer, select one or more fragrance ingredients to combine and generate the final fragrance formula. Dosage / Schedule Planning Module: Based on the generated fragrance formula, calculate the release dosage and timing of each essential oil / fragrance agent and output control commands.

4. The intelligent fragrance control system based on environmental perception and emotion analysis according to any one of claims 1 to 3, characterized in that, The fragrance release control system is based on a proportional control algorithm and specifically includes: Multi-channel essential oil compartment, each channel holds one type of essential oil or fragrance; Precision metering pumps or micro-nebulizers are used to mix different essential oils or fragrances in proportion and ensure accurate release of fragrance dosage; The fan / airflow guiding module is responsible for guiding the fragrance airflow. It uses a fluid dynamics model to design the fan speed and airflow direction to ensure that the fragrance is evenly distributed in the indoor environment. The airflow distribution valve is used to control the release direction and intensity of fragrance to adapt to different scenarios and user needs. It uses a PID control algorithm to adjust the valve opening to achieve precise airflow control. The communication module supports receiving commands locally or in the cloud, and transmitting status information, remaining amount and actual release dose data in real time, as well as receiving user feedback. The fragrance release control system also adjusts the fragrance release strategy in real time based on user feedback using simple scoring rules.

5. A smart fragrance control method, employing the smart fragrance control system based on environmental perception and emotion analysis as described in any one of claims 1 to 4, characterized in that, Specifically, the steps include the following: Step S100 Environment and Emotion Perception: Collect environmental parameters in real time through the environment perception module, and collect emotion data in real time through the emotion perception module; Step S200: Data preprocessing and multimodal fusion: The collected environmental parameters and emotion data are denoised, cleaned, feature extracted and aligned, and multimodal fusion is performed to obtain the current environment-emotion representation; Step S300 Scene recognition and target setting: Determine the scene category based on historical context and current environment-emotion representation, and determine the control target and constraints for this operation; Step S400: Formula and Dosage Calculation: Based on the aroma knowledge base and strategy optimizer, generate the fragrance formula and release dosage / timing, and output control instructions; Step S500 Execution and Release: The control command is sent to the fragrance release module and executed; During execution, the system can be released in stages according to a plan or adjusted in real time. Step S600 Feedback Collection and Model Update: Continuously collect environmental and emotional responses after release, evaluate the regulation effect, and update and optimize the fragrance regulation decision model or rule base based on the evaluation results to form a closed-loop learning.

6. The intelligent fragrance control method according to claim 5, characterized in that, The data preprocessing and multimodal fusion step S200 specifically includes: Step S210: Denoise the collected environmental parameters and emotion data; specifically including, Step S211: Apply a low-pass filter or median filter to remove high-frequency noise; Step S212: Smooth the surface using a moving average filter; Step S213: Clean the collected environmental parameters and emotion data, including the following operations: Missing value handling: Identify and fill in missing data using mean interpolation, linear interpolation or other suitable methods; Outlier detection: Detect and remove outliers using statistical analysis methods; Step S214: Standardize the data processed in step S213 by using min-max normalization or Z-score normalization to transform the values ​​of all environmental parameters to the same range. Step S220: Data from different environmental sensors are fused using weighted averaging and principal component analysis techniques to generate a comprehensive environmental status assessment index; specifically including, Step S221: Calculate a confidence score scoreParam for each environmental parameter, scoreParam∈[0, 1]. This confidence score is based on the following factors: sensor performance, data acquisition frequency, and reliability of historical records; using a weighted scoring algorithm: ； in, Scoring for each environmental parameter, The corresponding weights; Step S222: Output a table containing standardized and fused environmental parameters, providing a real-time environmental status report; Step S230: Extract key features from video and audio data, including: facial features: extract key points of facial expressions and calculate emotion scores using a sentiment analysis model; speech features: analyze the timbre, rhythm, and tone of speech to identify speech emotions and keywords; specifically including, Step S231: Facial feature extraction: Use a deep learning model to extract facial expression features. The sentiment score is calculated by a trained sentiment analysis model, and the probability distribution of the sentiment category is output. Step S232: Speech feature extraction: Analyze the timbre, rhythm, and tone of the speech to identify emotions and keywords; use Mel-frequency cepstral coefficients (MFCCs) to extract audio features; Step S233: Analyze the physiological data provided by the wearable device, including: heart rate and HRV: assess the relationship between emotional state and heart rate variability to determine emotional fluctuations; skin conductance response: analyze skin conductance data to detect the correlation between the participant's physiological response and emotional state; Step S234: Perform localized preprocessing of privacy-sensitive information at the data acquisition layer, extracting and storing only localized facial features; Step S235: Output a comprehensive emotion analysis report, including emotion data such as each participant's emotion score, emotion state change trend, and overall emotion assessment; Step S240: Fusing data from different sensors to generate a comprehensive environment-emotion representation; specifically including, Step S241: Perform feature fusion using weighted average method and principal component analysis; Step S242: Assign weights to different modalities using an attention mechanism; Step S250: Generate current environment-emotion representation.

7. The intelligent fragrance control method according to claim 5, characterized in that, The step S300, scene recognition and target setting, specifically includes: Step S310: Based on the collected environmental parameters and emotion data, identify the current scene type; specifically including, Step S311: Select key features of environmental parameters and sentiment data from the multimodal fusion results; Step S312: Classify the scene using Support Vector Machine (SVM) and Decision Tree; Step S313: Scene determination: Based on the classification results, determine the category of the current scene as meeting, family gathering, sleep / rest, or treatment / rehabilitation; Step S320: Based on the identified scene category, set the control objectives and constraints; specifically including, Step S321: Set the expected emotional state according to the scene type; Step S322: Ensure that the release dose of the fragrance meets safety standards; Step S323: Set the fragrance release time according to the scene type and event duration; Step S324: Adjust the fragrance formula and dosage based on user feedback and preference settings; Step S330: Integrate the identified scene categories and the set control objectives and constraints.

8. The intelligent fragrance control method according to claim 5, characterized in that, The step S400, formulation and dosage calculation, specifically includes: Step S410: Based on the scene recognition results and control objectives, select appropriate essential oils or fragrances from the aroma knowledge base; specifically including, Step S411: Extract fragrance ingredients and their characteristics that match the current scenario and objective from the fragrance knowledge base, including chemical components, subjective effects and safe dosage; Step S412: Select suitable fragrance ingredients according to preset rules, calculate a score for each fragrance ingredient, and select the ingredient with the highest score as the basis for the formula: ； in, , , It's weight. , and These are the effectiveness, safety, and user preference ratings of the fragrance ingredients; Step S413: Select one or more fragrance ingredients to combine and form the final fragrance formula; Step S420: Based on the generated fragrance formula, calculate the release dosage and timing of each essential oil / fragrance agent; specifically including, Step S421: Calculate the release dose of each ingredient based on the proportions of the ingredients in the fragrance formula; assuming the total release dose is... The proportion of each ingredient in the formula is: The dosage of each component is then calculated as follows: ； in, It is the first Dosage of essential oils; Step S422: Ensure the calculated dose does not exceed the safe range, and verify using the following formula: ； Step S423: Based on the scene requirements and control objectives, set the release sequence of the fragrance and adopt a time-segmented release strategy: ； in, It is the incremental time, set to an appropriate value that is adjusted according to the needs of the scenario; Step S430: Integrate the generated fragrance formula and its dosage and release timing information into an output object.

9. The intelligent fragrance control method according to claim 5, characterized in that, The execution and release of step S500 specifically includes: Step S510: Multi-channel essential oil compartment, each channel holds one type of essential oil or fragrance, so as to facilitate the flexible blending of various fragrances; Step S520: A precision metering pump or micro-needle is used to mix different essential oils in proportion and ensure accurate release of fragrance dosage; a proportioning control algorithm is used to calculate the required release of fragrance dosage to ensure mixing according to the formula. ； in, It is the first Dosage of the fragrance, It is the total release. That is the total amount of fragrance. It is the first The proportions of various fragrances in the formula; specifically including, Step S521: The fan / airflow guiding module is responsible for guiding the fragrance airflow. A fluid dynamics model is used to design the fan speed and airflow direction to achieve uniform distribution; Bernoulli's equation is used to calculate the airflow velocity. ； in, It's pressure. It is air density. It is the airflow speed. It is gravitational acceleration. It is height; Step S522: The airflow distribution valve is used to control the release direction and intensity of the fragrance to adapt to different scenarios and user needs. A PID control algorithm is used to adjust the valve opening to achieve precise airflow control. ； in, It is a control signal. It is the error between the expected output and the actual output. , and These are proportional, integral, and differential gains, respectively. Step S530: The device supports receiving commands locally or in the cloud, and can transmit status information, remaining amount and actual release dose data in real time; Step S540: Monitor environmental and user feedback during the release process; ； If the security index is lower than the preset threshold, the release strategy will be automatically adjusted or the release will be stopped. Step S550: Use a simple rating system to collect user feedback during the release process and adjust the fragrance release strategy in real time; ； in, It's about adjusting sensitivity. This is the user's current feedback. It is the expected satisfaction level.

10. The intelligent fragrance control method according to claim 5, characterized in that, The step S600 feedback acquisition and model update specifically includes: Step S610: By analyzing the feedback data, identify the effectiveness of fragrance control and user satisfaction, continuously optimize the fragrance release strategy and improve the user experience; the feedback data includes, but is not limited to, mood change curves, changes in environmental parameters, explicit user evaluations, and whether a safety alarm is triggered; Step S620: Use the collected data to train and improve the machine learning model to enhance the system's adaptability and intelligence level, and ensure that it can provide the best fragrance control solution in different environments and scenarios. Step S630: A time-series-based policy optimizer is used to weigh multiple objectives and generate a fragrance modulation strategy; specifically, it includes, Step S631: Optimize the fragrance control strategy using a reinforcement learning model, defining the state, action, and reward function: ； in, This is the current state. This is the current action. It is the reward received. It is a discount factor; Step S632: Optimize the fragrance release strategy using Bayesian optimization method, select the optimal parameter configuration to maximize the objective function, and establish the prior distribution of the objective function through a probability model; Step S633: Combine historical data and user feedback to establish a rule-based strategy, and generate control strategies through a rule engine; Step S640: Feed the evaluation results and user feedback into the decision model to form a cyclical update mechanism, enabling the system to continuously optimize the fragrance control strategy over time.

Images