Smart home Internet of Things interaction method

Abstract

The invention relates to the technical field of smart home Internet of Things, and discloses a smart home Internet of Things interaction method, which comprises the steps of dynamically regulating and controlling indoor light brightness based on a user work and rest rule and ambient light change so as to match user experience requirements, and collecting real-time air quality and historical data for analysis, the operation mode of the air purifier is intelligently regulated and controlled to guarantee breathing health, energy consumption distribution is optimized according to behavior habits of family members and household appliance use conditions, an energy-saving control instruction is generated, and the refrigerating or heating power of an air conditioner is flexibly adjusted in combination with temperature and humidity sensor data and seasonal characteristics to improve the indoor comfort degree. Through the scheme of the embodiment of the invention, the problem of how to dynamically regulate and control the indoor light brightness according to the work and rest rule of the user and the ambient light change so as to solve the problem of inconsistent user experience can be solved, and the breathing health of the user is effectively protected through statistical analysis of a real-time air quality monitoring result and past historical data.

Description

Technical Field

[0001] The present invention relates to the technical field of smart home Internet of Things, and in particular to a smart home Internet of Things interaction method. Background Art

[0002] A smart home Internet of Things interaction method aims to provide users with a convenient, efficient and comfortable home experience through the interconnection and data sharing of smart devices.

[0003] However, this method still faces a series of key problems that need to be solved: First, how to dynamically adjust the brightness of indoor lights according to the user's daily routine and changes in ambient light to improve the consistency of the lighting experience; second, how to intelligently adjust the air purifier's operating mode based on real-time air quality and historical data analysis to protect the user's respiratory health; third, it is necessary to optimize the relationship between family members' behavioral habits and home appliance usage to avoid energy waste; fourth, for multiple voice commands received by smart speakers, how to accurately coordinate the task response sequence of multiple devices according to priority rules to ensure the orderly operation of the system; finally, it is necessary to flexibly control the cooling or heating power of the air conditioner based on the temperature and humidity data collected by the sensor and combined with seasonal characteristics to improve the problem of indoor comfort deviation. Solving these problems will greatly enhance the intelligence level and user experience of the smart home system. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0005] A smart home Internet of Things interaction method, comprising:

[0006] Dynamically adjust indoor lighting brightness based on user routines and ambient light changes to match user experience needs. Real-time air quality and historical data are collected and analyzed, and the operating mode of the air purifier is intelligently adjusted to ensure respiratory health. Energy consumption distribution is optimized and energy-saving control instructions are generated based on family members' behavior habits and appliance usage. The cooling or heating power of the air conditioner is flexibly adjusted based on temperature and humidity sensor data and seasonal characteristics to improve indoor comfort.

[0007] The dynamic regulation of indoor light brightness based on the user's work and rest schedule and environmental light changes further includes: taking the average daily sleep time S of the user as a reference parameter. Judging the possible state of the user according to the current time T and S: if TS < 0, the rest mode is executed, and the formula is Light_Intensity = Brightness_Max * exp(a*(TS)^2), where Brightness_Max is the maximum brightness of the light, and a is the sensitivity coefficient; collecting the light sensor data Env_Light in the daytime environment and judging whether it is lower than the threshold Th: If Env_Light < Th, Light_Step = Step_Default * (1 - Env_Light / Th), and the light brightness is smoothly transitioned through a progressive adjustment strategy to avoid glare.

[0008] Preferably, the operation mode of the intelligent air purifier includes:

[0009] Collecting the real-time air quality index AQ_Index and the average air quality value AVG_INDEX in the recent 24 hours;

[0010] Setting the quality standard interval Q_STANDARD = (Q_Low, Q_High) as the good state range, and judging by the formula If AQ_Index > AVG_INDEX, Run_Speed = Run_Slow × AQ_Index / Q_High^b, where b is the acceleration ratio parameter;

[0011] When the sensor detects that the particulate matter concentration exceeds the preset warning level ALARM_LEVEL, it enters the emergency purification mode;

[0012] Generating a daily report and pushing it to the user's mobile phone to prompt the air quality analysis result and improvement suggestions.

[0013] Preferably, the analysis of the impact of family members' behavior habits on energy consumption distribution includes:

[0014] Statistical distribution curve of the appliance startup time P in historical data to extract the peak interval;

[0015] Calculating the total power consumption Power_Sum in each time period and the number of members Member-count in that period, and obtaining the load distribution coefficient Member-count_Allocate = k * Power_Sum / (Member_Count + β), where k and β are system optimization factors;

[0016] Combining the peak electricity consumption warning notice to preferentially reduce the power demand of non-core equipment;

[0017] Form an energy-saving target feedback loop to verify whether the improvement efficiency meets expectations after a fixed period.

[0018] Preferably, the smart speaker receives multi-source tasks T_List and classifies them by weight according to the source device type to form a basic score Priority_Basic;

[0019] For the same type of instruction group I_List, use the formula `Priority_Adjust = Priority_Basic(Distance_Centroid^γ+Time_Dif_Std×γ)` to adjust the weight level, where Distance_Centroid is the center distance deviation and Time_Dif_Std is the mean difference parameter;

[0020] When a conflict occurs, the one with the highest score will be activated first, otherwise the power will be evenly divided or executed in turns;

[0021] Save the execution history so that the priority rule set can be quickly called in subsequent similar scenarios.

[0022] Preferably, the temperature and humidity sensor and air conditioning control matching algorithm includes:

[0023] Read the current temperature and humidity combination H_Com and map it to a seasonal correction value `Season_Modify = α × H_Com^(δ / T_Outer)`, where T_Outer is the outside temperature and δ is a seasonal adjustment term.

[0024] Determine the comfort limit range Comf_Limit = (Cmin, Cmax), and activate the automatic compensation mechanism when it exceeds the limit;

[0025] For refined control of cooling power R, the relationship `R_Ajusted = Base_Pwr × (T_SetSensor_Read) + ζ × Temp_RisingRate` is obtained based on the target temperature T_set;

[0026] Provides multiple operation panels for users to manually switch between different work preference presets.

[0027] Preferably, introducing external weather forecast data to assist in air conditioning adjustment includes:

[0028] Integrate the weather forecast Weather_Data for the next 72 hours and select the significant fluctuation segments to mark the key nodes K_List;

[0029] The impact degree is calculated using the calibration formula parameter `Calib_Factor=f(Weather_Data_Match)=∑(W_TmpTarget_Tmp)*(Δt^λ)` based on the historical meteorological comparison database;

[0030] Enhance the system standby response level one hour before extreme weather is about to occur;

[0031] Record the actual results and compare them with the initial strategy to evaluate the error rate and store the revised version of the model for next use.

[0032] Preferably, customizing more personalized lighting effect logic based on user's personal habits includes:

[0033] Define the personalized indicator matrix Profile_mat and initialize the vector length Vector_length;

[0034] The module `Profile_Upt = Ψ(Vector_length × Profilemat + New_Data_Input * σ(Tune_Level) / μ)` is updated by layer-by-layer superposition formula to capture subtle preference changes;

[0035] Set the number of learning iterations N_limit to check if there is a persistent deviation so that training can be restarted;

[0036] Syncing to other sub-devices ensures that the unified theme style continues without disrupting the overall coordination.

[0037] Preferably, the process of enhancing the accuracy of air quality analysis includes:

[0038] Get the long-term PM5 data stream PM_List and perform dimensionality reduction transformation Dataspace_reduction;

[0039] Use the dynamic window function Dynamic_Window to locate the short-term anomaly point `Peak_Detect = Dataspace_Reduction × [log(Variances) / ω+Threshold]^ρ` to trigger the early warning signal;

[0040] Re-aggregate the information obtained from the fine-grained monitoring network to draw a risk distribution heat map;

[0041] Suggestions for local improvements are made for high-risk areas, such as opening ventilation valves or adding activated carbon adsorption layers.

[0042] Preferably, the household appliance energy consumption planning method includes:

[0043] Introduce sub-item metering chips to track the output power details of each channel Energy_Detail;

[0044] Use multiple linear regression to predict the trend of the next week, assuming that the model parameter fit must be higher than the threshold Min_Fit_Rate;

[0045] Implement dynamic pricing simulation Cost_Plan = `Sum(Base_Cost×Predict_Energy×Peak_Flag)` to estimate financial expenditure;

[0046] Develop a stage cost management checklist for regular review and timely budget adjustment.

[0047] Preferred new strategies to improve voice interaction efficiency include:

[0048] Establish user intention clustering Cluster_Group and reduce redundant computing overhead Cluster_Reduce;

[0049] Apply deep learning neural network DL_network to extract the association probability of high-frequency phrases Feature_Words;

[0050] Design a fuzzy decision engine to evaluate the answer scores of similar questions Rank_Score = `(Confidence_Value×Word_Relevance×Context_Similarity+Ω)^η` to find the best matching result;

[0051] Recording the conversation path facilitates future retrieval of similar cases and speeds up feedback.

[0052] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0053] 1. Dynamically adjust indoor lighting brightness based on the user's daily routine and ambient light changes to match user experience needs. Real-time air quality and historical data are collected and analyzed, and the air purifier's operating mode is intelligently adjusted to ensure respiratory health. Energy consumption distribution is optimized and energy-saving control instructions are generated based on family members' behavior habits and appliance usage. Air conditioning cooling or heating power is flexibly adjusted based on temperature and humidity sensor data and seasonal characteristics to improve indoor comfort. The solutions of the disclosed embodiments can address the issue of inconsistent user experience by dynamically adjusting indoor lighting brightness based on the user's daily routine and ambient light changes.

[0054] 2. Combining sensor data (such as light intensity) with the user's daily routine, the system dynamically adjusts indoor lighting brightness in real time. The system pre-defines different scenes based on the user's sleep, work, or leisure time periods, and adjusts the light color and brightness based on external light changes, ensuring the lighting environment is always optimal to suit the needs of different time periods.

[0055] 3. Based on the real-time air quality monitoring results and statistical analysis of historical data (such as PM2.5 concentration fluctuation trends), the air purifier's operating mode (such as strong mode, silent mode or automatic mode) is intelligently adjusted. This method can not only quickly respond to the current deterioration of indoor air conditions, but also predictively adjust the filtration effect, thereby effectively protecting the user's respiratory health and avoiding the risks caused by long-term inefficient operation or ignoring potential pollution threats. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A schematic diagram of the invention process;

[0057] Figure 2 A schematic diagram of the invention efficiency feedback process;

[0058] Figure 3 Schematic diagram of the invention record and synchronization process;

[0059] Figure 4 Flowchart of a new strategy to improve voice interaction efficiency. DETAILED DESCRIPTION

[0060] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0061] See also Figure 1 - Figure 4 , each step and specific operation process of a smart home Internet of Things interaction method of the present invention;

[0062] First, to dynamically adjust indoor lighting based on the user's daily routines and ambient light levels, the system uses a human body sensor to capture the activity patterns of family members. It then uses a light sensor to collect real-time light intensity inside and outside the room. The system then combines this collected data with user-defined preferences to develop an appropriate lighting strategy and implement control. Throughout this process, the system leverages artificial intelligence algorithms to continuously learn from the user's usage habits and lighting requirements over different time periods. For example, during the morning wake-up period, the lighting can gradually brighten from dim to bright, simulating a natural sunrise to enhance the user's emotional comfort. At night, it automatically switches to a soft, warm light mode to reduce visual fatigue and promote good sleep.

[0063] The second step is to collect real-time air quality and analyze it with historical data to intelligently adjust the working status of the air purifier to protect the user's respiratory health. The system has a built-in high-precision gas detector that can monitor particulate matter (PM2.5 / PM10), harmful gas concentration and humidity level in the air in real time. These measured values ​​are then compared with the previously accumulated historical big data to calculate whether there is potential danger in the current indoor and outdoor environment. Once an excess condition is found, the purification equipment will be notified in time to enhance the filtration efficiency or switch the mode to the optimal configuration for specific scenarios such as the night silent mode. In one embodiment, if the outdoor pollution is serious one day and there are elderly people or children at home, the smart home will quickly switch to the high-efficiency filtration mode to ensure that vulnerable groups have a clean air supply.

[0064] Third, to assess family members' behavioral habits and overall appliance operation to reduce energy loss, comprehensive tracking and recording of the operating status of every electrical appliance is employed. Power monitoring units are installed to collect detailed information on the daily start-up and shutdown frequencies and power consumption of all major electrical components, such as refrigerators, washing machines, and televisions. Mathematical modeling is then developed based on each resident's daily activity patterns. This generates a predictive graph to identify areas at risk of overspending and recommends energy-saving measures. For example, if a family habitually uses their gaming console for hours on weekends but only occasionally on other days, this cyclical pattern can be used to mitigate power losses during idle periods.

[0065] Next, a dual-feedback control scheme was designed to flexibly control the air conditioner's output power based on quantitative parameters provided by temperature and humidity sensor nodes and the seasonal variations unique to the location, thereby improving user experience. This approach includes, but is not limited to, deploying a sufficient number of distributed temperature and humidity sensing modules throughout the residential area to form a dense measurement network; integrating this with a cloud-based center to store long-term, accumulated statistical data on macro-meteorological elements to construct a comprehensive evaluation framework; setting thresholds based on seasonal temperature ranges to delineate comfort zones, and incorporating localized lifestyle details to provide customized decision-making guidance. For example, in tropical summer climates, if the detected indoor temperature exceeds a preset target threshold, a powerful cooling function is immediately activated to prevent excessively high temperatures from causing strain on the body and unnecessary power consumption.

[0066] The final key function involves handling simultaneous voice requests from multiple sources, queuing logic and prioritizing mechanisms to ensure effective communication without causing confusion and a degraded user experience. This section emphasizes the importance of establishing a hierarchical structure that clearly distinguishes between general query commands and emergency alarm processing tasks. For example, when a fire or smoke alarm arrives, music playback and other common leisure activities should be interrupted to focus on safety precautions and evacuation guidance, protecting lives and property from disaster. Ultimately, the entire interactive process must work together to maximize the expected benefits of smart home living and meet the actual needs of consumers, continuously evolving and innovating.

[0067] The present invention further dynamically controls the brightness of indoor lights based on the user's daily routine and changes in ambient light, and achieves the effect of dynamically controlling the lights through several key steps, so that users can feel comfortable in different scenarios.

[0068] First, consider the user's average daily sleep time, S, as a reference parameter. This average daily sleep time can be calculated by analyzing long-term monitoring data from smart home devices and expressed in hours. For example, if the user typically sleeps 8 hours per night, S could be 8.

[0069] Secondly, the possible state is judged based on the current time T and the user's average daily sleep time S. If TS is less than 0 (where TS represents the difference between the current time T and the user's expected sleep end time), the current time period may be in rest mode. At this time, the light brightness calculation formula is: Light_Intensity = Brightness_Max*exp(a*(TS)^2)&;In this formula, Brightness_Max&;represents the maximum brightness of the light (ranging from 0 to 1 or a percentage value, such as 0 to 100), a& is the sensitivity coefficient, the range is negative, and the recommended optimal value is adjusted according to the environment and user sensitivity. By introducing an exponential function, it is ensured that the light gradually decreases as the rest period approaches to simulate the natural sunset environment and help users enter a deep rest state.

[0070] In the daytime environment, collect the ambient light sensor data Env_Light and determine whether it is lower than the threshold Th (the threshold is an empirical value or a set value). If &; Env_Light < Th &; is satisfied, then execute the adjustment logic Light_Step = Step_Default * (1 - Env_Light / Th) &; This formula aims to dynamically adjust the change rate of the indoor lights according to the actual external light conditions, making the lamp adjustment more precise and gentle. Among them, &; Step_Default &; is a predefined reference adjustment amount, and the rest of the formula reflects the influence weight of the external ambient light on the adjustment speed.

[0071] Finally, the gradual adjustment of the lighting strategy avoids discomfort caused by sudden light changes through smooth transitions. This means that even when the calculation results change rapidly, the actual illumination will be adjusted at a stable speed to meet the comfort requirements. In one embodiment, specifically, assume that it is 3 am currently and the TS value is approximately -5 hours. Then, according to the formula, it can be known that the lights in the room will output at a lower brightness and maintain a steady dimming state at this time; on the other hand, in the daytime, when the indoor light is dim due to strong sunlight, the brightness is automatically increased by a certain amount, and at the same time, visual interference is ensured to be avoided according to the above step logic.

[0072] In summary, this intelligent method can provide the most suitable lighting effect more in line with people's biological rhythms and environmental conditions, improving the quality of people's living environment and physical and mental health levels.

[0073] Furthermore, for the operation mode of the intelligent control air purifier of the present invention, the specific steps are as follows: First step, collect the real-time air index (AQ_Index) and the average air quality value in the last 24 hours (AVG_INDEX). Second step, set the quality standard range Q_STANDARD = (Q_Low, Q_High), and adjust the operation speed of the purifier when the formula AQ_Index > AVG_INDEX and Run_Speed = Run_Slow × AQ_Index / Q_High^b is satisfied. Among them, the acceleration ratio parameter b is the key variable affecting the formula result. Third step, when the sensor detects that the particulate matter concentration exceeds the warning level (ALARM_LEVEL), enter the emergency purification mode. Fourth step, generate a daily report containing the air quality analysis results and improvement suggestions, and push it to the user's mobile phone.

[0074] The acquisition of the real-time air index AQ_Index and the average air quality AVG_INDEX in the last 24 hours is completed through air sensor devices installed in the smart home network. This is to make the control closer to the comparison between real-time data and long-term trends, and provide a reference benchmark for subsequent algorithms. The quality standard interval Q_STANDARD includes the upper limit Q_High and the lower limit Q_Low of the good state. The purpose is to determine the environmental target range to ensure that the living space is always in healthy air conditions. The value range of the acceleration ratio parameter b is usually (1,5]. The best choice is set to 3 according to the test environment. This value can effectively adjust the balance of the purifier's workload and response time.

[0075] In one embodiment, assume that the real-time air quality index (AQ_Index) in a room is recorded at 70 μg / m³, and the average value (AVG_INDEX) over the last 24 hours has remained at 60 μg / m³. The known quality standard interval is set to (50 μg / m³, 80 μg / m³). According to the formula, when AQ_Index exceeds the average value of 60 μg / m³ and the acceleration ratio parameter b is set to 3, the operating speed should be a multiple of the normal minimum speed, indicating that the machine will gradually increase the wind speed until the air returns to a good range.

[0076] Finally, when the sensor data exceeds the preset particulate matter concentration alarm level (for example, if PM2.5 exceeds the safe value of 35 micrograms per cubic meter and reaches 50 micrograms per cubic meter), the system will initiate an emergency response procedure and perform a strong purification operation. At the same time, a detailed assessment of the current environmental monitoring status is generated daily, along with feasible improvement suggestions, which are sent to the end user for review and confirmation to continuously improve the health of the home. Therefore, the above steps work closely together to ensure that the indoor air quality in the smart home reaches or approaches the ideal clean state, improving the comfort and safety of the user's living environment.

[0077] The analysis of the impact of family members' behavioral habits on energy consumption distribution in the present invention is divided into four main steps: first, historical data is collected to obtain the distribution curve of the appliance startup time P and extract the peak interval; second, the load distribution coefficient formula is obtained by calculating the total power consumption in each time period and the number of family members in that time period; third, the power demand of non-core equipment is reduced based on priority management during peak power consumption; and fourth, the energy-saving target feedback loop is used to evaluate and verify the effectiveness of improvement measures.

[0078] The distribution curve of the appliance startup time P in the historical data is statistically analyzed to extract the peak interval. This is to divide the historical data points into time periods, draw a trend chart of the power startup behavior, and locate the peak power consumption period corresponding to the frequent use period. The purpose of this step is to determine the time range of high power load to facilitate the targeted implementation of the optimization strategy. For example, in a specific scenario, if a family has multiple home appliances (such as microwave ovens, vacuum cleaners, and ovens) started at the same time between 7 and 9 am on weekdays, this period can be defined as the startup peak interval.

[0079] Calculate the total power consumption Power_Sum and the number of members in each time period Member-count, and form the load distribution coefficient Member-count_Allocate = k*Power_Sum / (Member_Count+β)&; The quantitative distribution of energy consumption among users is achieved through the formula. Among them, the parameter Power_Sum represents the total cumulative power consumption of all household appliances in a specific time period, and the value is usually in the range of several hundred watt-hours to several kilowatt-hours; the parameter Member_Count represents the actual number of family members at the current moment; the parameter k is used to adjust the load proportional factor to balance the individual energy consumption contribution, and the recommended range is 0.8 to 1.2; the parameter β is an optimization factor set to avoid instability caused by a denominator that is too small or zero. The default value is about 0.5 and can be dynamically adjusted according to actual usage effects. The meaning of the above formula is to determine the degree of load balancing based on a comprehensive consideration of the overall power consumption level and member participation. This setting method takes into account both the use of equipment and the impact of human activities, ensuring a more scientific calculation basis.

[0080] Integrating peak power demand warnings to prioritize reducing the power demands of non-core devices is a strategy that intervenes when the system detects an impending or already-entered high-load state. For example, a signal could be sent before a predicted surge in washing machine operations around 11:00 PM, potentially increasing the overall grid load. This could allow residents to postpone or suspend these tasks in advance to avoid unnecessary expenses.

[0081] Finally, to ensure that the initial designs are truly effective, the results of energy-saving improvements are periodically reviewed. This energy-saving feedback loop involves collecting relevant statistical data after a certain period of time and comparing it with the initial plan to determine if the target has been met. If it falls short, the corresponding parameter configuration plan is revised and redeployed based on the gap until the energy-saving and emission-reduction standards are met.

[0082] The voice command priority response of the present invention includes:

[0083] First, the smart speaker receives multi-source tasks and classifies them by weighted source device type, obtaining an initial Priority_Basic score. Next, an adjustment formula is applied to optimize the scoring system for groups of commands of the same type. The adjusted weighting is: Priority_Adjust = Priority_Basic × (Distance_Centroid^γ + Time_Dif_Std × γ), where Distance_Centroid represents the distance deviation from the center point, ranging from zero to ten, with smaller values ​​indicating closer to the ideal state; Time_Dif_Std represents the standard deviation of the time distribution, ranging from zero to five; and γ, a sensitivity adjustment parameter, ranges from 0.5 to 1.5. Setting γ to around 1.2 achieves optimal performance. If multiple commands conflict, the one with the highest weight is activated first. Otherwise, power allocation is evenly distributed or tasks are scheduled in rotation to ensure balance. The execution information and results of each task are stored as a reference database for subsequent similar situations, enabling rapid activation of the optimal decision-making mechanism.

[0084] The first step emphasizes the impact of different input devices on the importance of tasks. By distinguishing the attributes of tasks coming from, for example, refrigerators, TVs, or other smart home devices, a preliminary basic weight assessment for each type of instruction can be established, allowing the system to understand the default urgency level of requests issued by each device. The second step introduces mathematical adjustment logic to ensure that even multiple requests in the same category can be reasonably reallocated in the ranking position based on real-time conditions. That is, the original judgment value is corrected by using a formula to dynamically balance the importance of geographical proximity and uniform distribution of time periods. The final processing rules clarify the course of action when facing competitive situations and the significance of reusing historical data analysis.

[0085] In one embodiment, the above process operates in practice. Consider three voice commands: one from the living room TV screen controller requesting to turn on a movie channel, one from the bedroom bedside lamp to set a timer wake-up service, and one from the kitchen interface to activate a breakfast preheater. The first step assigns preliminary scores based on the usage characteristics of each of these three devices. For example, assuming viewing entertainment is relatively unimportant, it's given a score of 25; a nighttime alarm, considered a necessity, receives a score of 50; and cooking, directly related to healthy eating, receives a higher priority of 70. The final score is then calculated based on the geographically dispersed nature of the participants and the differences in the time intervals they submit. For example, the kitchen equipment may be slightly farther from the core family activity area, while the other two may be more or less close. Furthermore, there may be a slight difference in the set times. The corresponding numbers are then substituted according to the aforementioned variable definition rules to obtain a precise score comparison. The breakfast task, once again, requires the most immediate response, so its function is performed first. The subsequent two actions are then appropriately scheduled, either alternating or performing them simultaneously to minimize energy consumption. This demonstrates the practical application value of the complete process, and as more and more similar scenarios are accumulated, the system can automatically learn certain fixed patterns to improve the speed and accuracy of future responses.

[0086] The temperature and humidity sensor and air conditioning control matching algorithm of the present invention include: first reading the temperature and humidity combination H_Com of the current environment, and mapping it to a seasonal correction value &;Season_Modify=α×H_Com^(δ / T_Outer)&;wherein, α is a constant parameter related to the system sensitivity, and the default optimal value is 0.9 to 1.1; H_Com is the currently detected temperature and humidity combination (expressed as a dimensionless numerical value, such as a comprehensive score with a value range of [0,1] after standardization); T_Outer represents the outside temperature in degrees Celsius; δ is a seasonal adjustment item, preferably -0.5 to -1.5 in winter, and 0.5 to 1.5 in summer. The setting of this formula is intended to take into account the changes in the internal and external environments, so as to optimize the adaptability of the adjustment. Due to the complex interactive influence between temperature and humidity and the outside temperature, the use of this form of mathematical relationship can be more in line with the actual changes in environmental conditions.

[0087] The comfort limit range, Comf_Limit = (Cmin, Cmax), is then determined. Here, Cmin and Cmax define the minimum and maximum comfortable indoor temperature ranges, respectively, typically preset to Cmin = 18°C ​​and Cmax = 26°C. If the sensor reading falls outside this range, an automatic compensation mechanism is immediately activated to adjust. For example, if the temperature is too high or too low, the air conditioner will quickly cool or heat the room. Setting a clear comfort range effectively improves the user experience.

[0088] The cooling power R is then finely controlled, establishing the following relationship based on the user-set target temperature T_set and the actual sensor temperature Sensor_Read: R_Adjusted = Base_Pwr × (T_Set - Sensor_Read) + ζ × Temp_RisingRate. Base_Pwr specifies the base output power of the air conditioner (e.g., 1kW), the ζ parameter reflects the system's influence on the heating rate, and a default range of 0.3 to 0.7 is recommended. Temp_RisingRate indicates the specific temperature rise per minute, expressed in degrees Celsius per minute. This relationship aims to combine the differences between the target and the current situation while taking into account the ambient temperature change rate, thereby enabling a more dynamic and flexible allocation of energy consumption.

[0089] Finally, the system should provide a variety of operation panels, allowing users to manually switch between different operating preferences. In one embodiment, if a resident prefers to reduce air conditioning noise at night, they can select the energy-saving and silent mode. Specifically, this mode not only reduces the wind speed setting but also adjusts the weight of sensitive area detection to avoid unnecessary operational burden caused by minor fluctuations.

[0090] The present invention introduces external weather forecast data to assist in air conditioning adjustment, including: integrating the weather forecast Weather_Data for the next 72 hours, selecting time segments with significant temperature fluctuations and marking them as key nodes K_List; by matching these key nodes with data in the historical meteorological comparison database, calculating the calibration formula parameter Calib_Factor = f(Weather_Data_Match) = ∑(W_Tmp-Target_Tmp)*(Δt^λ), and evaluating the impact of external weather on the air conditioning operation strategy based on this value. When extreme weather is predicted, the standby response level of the air conditioning system is increased one hour in advance. Subsequently, the deviation between the actual adjustment result and the estimated strategy is recorded, and the correction model is used to optimize the next execution plan.

[0091] First, call the weather service interface provided by the third-party platform to obtain the weather forecast data (Weather_Data) for the next 72 hours, ensuring that it covers variables such as humidity, wind speed, and key temperature variables. Next, analyze whether there are significant temperature fluctuations or trend turning points within this time period, and mark these as the key time node set K_List. For example, if the temperature difference between the 36th and 48th hours exceeds the threshold of 5°C, each is added to the K_List. This screening process aims to identify periods of external condition transitions that may affect indoor thermal comfort.

[0092] To further quantify the actual effects of these fluctuations, a database of historically similar situations was used in conjunction with a calibration formula to calculate the impact coefficient. The parameters are defined as follows: W_Tmp represents the average outdoor temperature at the corresponding time point in the weather forecast, Target_Tmp is the target constant room temperature reference value, and Δt represents the time span from the current moment to the occurrence of the event in hours. The exponential term λ is typically set between 0.8 and 1 to simulate the degree to which different delay effects on the final indoor environment are attenuated, with an optimal value of approximately 0.9. The cumulative structure of the entire expression reflects the overall impact factor, Calib_Factor, that combines the contributions of all candidate moments. The logic behind this approach is that changes closer to implementation will have more immediate and noticeable consequences, while allowing for flexible adjustment of the weight distribution ratio to meet the needs of specific hardware configurations.

[0093] In one example, if the system determines that the temperature will drop significantly tomorrow afternoon (from 25°C to 15°C), it will preemptively switch the smart air conditioning mode to energy-saving pre-cooling mode and activate fan circulation to maintain a stable stratified air distribution within an hour before the temperature drops. The measured improvement in user satisfaction after actual application can verify the accuracy of the original prediction and the room for optimization, and accordingly update the basic model parameter combination for the next similar scenario application.

[0094] After completing all steps in the process, the entire data, including but not limited to the difference in energy consumption before and after setup and changes in user experience scores, must be integrated into a long-term learning sample and stored as a centralized backup. This provides a rich and reliable source of empirical reference material for future algorithm iterations and upgrades. This repetitive cycle will lead to a higher level of intelligence for the overall smart home IoT interaction system.

[0095] The present invention's logic for customizing lighting effects to individual user preferences includes defining a personalized indicator matrix and initializing its vector length; updating a module to capture changes in user preferences using a formula; setting a learning iteration count to detect persistent deviations and adjust the training state; and applying the synchronized results to other sub-devices to maintain overall theme consistency. Each step is described in detail below.

[0096] First, define the personalized indicator matrix Profile_mat and initialize the vector length Vector_length. This operation generates a data infrastructure based on each user's unique behavioral habits. Profile_mat serves as the primary record matrix, dynamically storing various user-related parameters such as usage duration, on / off frequency, and color and brightness preferences for specific time periods. Initializing the vector length sets the initial values ​​for the matrix dimensions. In smart home scenarios, the default range can be flexibly adjusted between 5 and 100, with the optimal initial value determined based on data complexity.

[0097] Next, the layered formula Profile_Upt = Ψ(Vector_length × Profile_mat + New_Data_Input * σ(Tune_Level) / μ) is used to capture subtle changes in user preferences in real time. The parameter New_Data_Input refers to the most recently input behavioral data, such as the user's newly selected brightness adjustment value; Tune_Level is the adjustment level factor (in the range [1,10]), indicating the desired responsiveness to user input in different situations; μ is the standard deviation, ensuring result stability, and its best practice value is a constant of around 2 to 3 within the experimental range; σ(Tune_Level) allows for further fine-tuning of sensitivity. The significance of this formula lies in continuously optimizing prediction accuracy based on existing and newly added data, thereby gradually bringing the lighting effect logic closer to the ideal model that meets user needs.

[0098] To prevent the system from favoring short-term preferences and affecting long-term adaptability, we introduced a control for the number of learning iterations, N_limit, to periodically monitor for significant deviations. The range of N_limit can be selected based on practical needs. For example, in one embodiment, N_limit can be set to 50 iterations. Each time the preset limit is reached, a new round of comprehensive reset analysis is triggered, ensuring that the model continuously learns accurate information and avoids the accumulation of errors that could lead to further deviations.

[0099] The final step is to synchronize adjustments to other sub-devices to ensure global style coordination and unity. The key to this step is to push and share the results of the previously precisely calibrated lighting effect algorithm to all related sub-nodes. Specifically, once the master bedroom lighting is customized, the corresponding settings will be immediately transmitted to the lamps or smart decorative elements in the adjacent areas, and the color and light and dark transition strategies of these components will be automatically fine-tuned to match the atmosphere of the main room. For example, when the master bedroom is adjusted to a warm orange-yellow low light source at night, the bathroom and corridor lights will also softly switch to a similar style, providing a consistent and comfortable living experience.

[0100] The process of enhancing the precision of air quality analysis of the present invention is as follows:

[0101] The first step is to acquire a long-term PM5 data stream and perform a dimensionality reduction transformation on the data to reduce computational effort and complexity. In this step, a collection of PM5 data continuously transmitted by a home IoT sensor network (denoted as PM_List) is collected and mapped into a lower-dimensional space to simplify data processing. This dimensionality reduction operation uses a method called Dataspace_reduction, which aims to eliminate noise and redundant information while retaining key features.

[0102] The second step uses the dynamic window function Dynamic_Window to precisely locate short-term peaks in particulate matter concentration fluctuations and generate corresponding warning signals. This formula is expressed as Peak_Detect = Dataspace_Reduction × [log(Variances) / ω + Threshold]^ρ. Variances represents the rate of change of the current detection value relative to the baseline; a larger value indicates an increase in suspended particulate matter in the air. ω is a control factor that determines the window size scaling ratio. The default recommended setting is between 0.1 and 2, adjusted based on the application scenario, with a typical value of 1 being used to ensure stability. Threshold is used to filter out small disturbances. The setting should be reasonably selected based on historical data analysis, typically within the range of ±1σ from the mean. The exponent ρ is used to further optimize the sensitivity curve, typically selecting a moderate value of approximately 1.5 within a range of 0 to 3 to achieve a good balance between sensitivity and false alarm probability. This formula accounts for short-term data variability and the influence of background noise, providing a more accurate means of identifying anomalies.

[0103] Next, a reaggregation operation is performed, integrating data returned by multiple nodes within the fine-grained monitoring network to create a spatially representative heat map of risk levels, visualizing which areas are potentially at risk due to high pollution levels. This heat map is constructed based on the distribution characteristics of the previously extracted valid and compressed feature vectors, and is derived through statistical modeling.

[0104] For the areas of special concern identified above, practical improvement plans are proposed, including automated equipment linkage mechanisms to respond and adjust indoor air quality, such as opening or closing designated smart window ventilation systems, adjusting the speed of air purifiers, or prompting users to add corresponding chemical and physical purification media materials, such as adding an activated carbon adsorption layer, and other hardware intervention strategies to achieve overall optimization of residential environmental hygiene management.

[0105] In one specific embodiment, consider a smart home environment equipped with a series of IoT-connected PM5 monitoring units. If, during a specific period, high activity in certain rooms causes significant fluctuations in PM5 values, triggering a formula to determine if the PM5 level exceeds the standard, a visual heat map alert is instantly displayed on the terminal interface to inform the user of the situation. Simultaneously, the system can automatically respond according to a predefined algorithm, for example, by directing the operation of associated exhaust fans to alleviate localized issues and providing recommendations for installing additional protective equipment to enhance filtration efficiency and ensure a consistently healthy living environment.

[0106] This invention further improves the household appliance energy consumption planning method as follows: First, a sub-item metering chip is introduced to track the detailed power output of each channel. This sub-item metering chip installed on the device collects the specific power consumption data of each household appliance during operation, forming a complete detailed table Energy_Detail. The purpose is to accurately understand the actual power consumption of each household appliance, providing basic data for subsequent optimization and adjustment.

[0107] Using multivariate linear regression to predict trends for the next week, this step uses historical electricity consumption data as sample input and builds a multivariate model to estimate future energy consumption. The parameter fit in the prediction model must exceed a threshold, Min_Fit_Rate. For example, Min_Fit_Rate can be set to 0.8 or above to ensure the most accurate and reliable prediction results. Min_Fit_Rate indicates how well the model's predictions match the actual situation. If the fit falls below the threshold, the prediction is unfeasible and the model needs to be rebuilt.

[0108] The dynamic pricing simulation formula Cost_Plan = Sum(Base_Cost×Predict_Energy×Peak_Flag) is then implemented to evaluate financial expenditure. Base_Cost represents the basic unit cost of electricity price; Predict_Energy is the electricity consumption of the next week predicted from the previous step, usually in kilowatt-hours (kWh); Peak_Flag is the price factor mark for peak or valley time, with a value range of {1 (peak period), 0.7 (valley period)}, and the optimal value is determined by the actual grid pricing strategy. The calculation principle of this formula is to dynamically reflect the overall electricity cost based on the electricity price and the characteristics of the electricity consumption period combined with the predicted value. For example, in one embodiment, the Base_Cost of a certain area is 0.5 yuan per kilowatt-hour. If a specific device is estimated to consume 20 kilowatt-hours of electricity during peak hours, the cost is: Cost = Base_Cost×Predict_Energy×1=0.5×20×1=10 yuan.

[0109] Finally, a cost management checklist is developed for regular review and budget adjustments. This requires the system to generate a detailed cost breakdown for decision makers to review regularly, identifying anomalies or potential optimization opportunities. For example, in a specific implementation scenario, if a three-month data review reveals significantly lower nighttime washing machine usage, energy costs could be reduced by encouraging nighttime washing and maintenance operations with electricity discounts.

[0110] This invention introduces a new strategy for improving voice interaction efficiency. This strategy includes four key steps: first, clustering user intent to reduce redundant computational overhead. Second, a deep learning neural network is used to automatically extract high-frequency phrases and their associated probabilities. Third, a fuzzy decision engine is designed to evaluate the scores of responses to similar questions based on a specific formula to identify the optimal match. Finally, conversation paths are recorded to facilitate future retrieval of similar cases, accelerating feedback.

[0111] Clustering user intent and reducing redundant computation involves classifying similar user intent into different clusters and groups. This involves leveraging historical or real-time data analysis to optimize computational requirements and avoid unnecessary repetition. This reduces repeated reasoning on the same problem and conserves system resources. For example, in a smart home, if multiple users frequently request to adjust the living room temperature, these commands can be grouped into a temperature adjustment cluster to process similar queries uniformly.

[0112] Applying a deep learning neural network (DL_network) to extract features and analyze association probabilities for high-frequency phrases involves using advanced AI algorithms to learn language patterns from a large corpus and identify commonly used expressions and the potential connections between them. This technology can effectively enhance natural language understanding capabilities. In one embodiment, if a user repeatedly mentions commands such as setting the air conditioner to 26 degrees, the system automatically marks these phrases as a high-weighted feature combination for subsequent judgment.

[0113] A fuzzy decision engine is designed to evaluate the quality of each possible response and select the final answer using the formula Rank_Score = (Confidence_Value × Word_Relevance × Context_Similarity + Ω)^η. The confidence parameter, Confidence_Value, represents the certainty of the candidate response and typically falls within the range [0, 1]; the closer to 1, the better. The word relevance coefficient, Word_Relevance, reflects the semantic consistency between the word and the current input, also ranging from zero to zero. Higher scores indicate a closer fit to the user's question. Context_Similarity measures the consistency between the preceding and following contexts and falls within the same closed interval, with higher scores indicating a stronger connection. The compensation term, Ω, smooths the overall score variation and is set to a fixed small positive value, such as 0.1, to ensure that no single factor causes an overall low score. The adjustment exponent, η, influences the amplification and reduction effects of the combined factors. It is recommended to use a positive value within the range but slightly greater than 1, such as 1.5, to ensure an overall increasing trend while not overshooting individual characteristics, thus maintaining a reasonable equilibrium. The purpose of such careful design is to accurately and efficiently locate the most appropriate answers to provide visitors with satisfactory service standards and maximize customer satisfaction levels.

[0114] Recording and archiving complete conversation sequences helps speed up the resolution process when the same topic is inquired about again in the future. Specifically, each complete question-and-answer experience, including the cause and effect, is saved as a database entry. When faced with the same or very similar situation again, previous solutions can be quickly recalled, reducing the cost of recalculation and analysis and significantly improving operational efficiency. For example, when a resident repeatedly inquires about the bedroom light switch, the previously saved interaction data allows direct reference to past successful examples without the need to spend additional time building and analyzing new solutions, achieving rapid response goals and completing established tasks. The process is smooth and convenient, the operation is simple and intuitive, and the user experience is excellent, which enhances brand value and awareness, continuously grows, and the market share steadily increases.

[0115] The present invention provides a smart home IoT interaction method that enhances the user experience in a smart home environment through a series of intelligent sensing, data processing, and dynamic control mechanisms. The specific steps and technical solutions are as follows:

[0116] 1. Dynamic Lighting Adjustment: To address inconsistent user experiences, this method combines sensor data (such as light intensity) with the user's daily routine to dynamically adjust indoor lighting brightness in real time. The system pre-defines different scenes based on the user's sleep, work, or leisure time periods. Based on these scenes and external light changes, it adjusts the lighting color and brightness, ensuring the lighting environment is always optimal to suit the needs of different time periods.

[0117] 2. Intelligent Air Quality Control: Based on real-time air quality monitoring results and statistical analysis of historical data (such as PM2.5 concentration fluctuation trends), the air purifier's operating mode (such as high-power, quiet, or automatic) is intelligently adjusted. This method not only quickly responds to current indoor air deterioration, but also predictively adjusts filtration effectiveness, effectively protecting users' respiratory health and avoiding the risks associated with prolonged inefficient operation or ignoring potential pollution threats.

[0118] 3. Energy Consumption Distribution Optimization: By analyzing family members' behavior and appliance usage frequency data, an energy consumption optimization model is established. For example, when appliances are not frequently used but in standby mode during certain periods, the system automatically generates energy-saving control instructions to limit these devices to full standby mode. During periods of high usage, power output distribution among various appliance loads is balanced to maximize overall efficiency, significantly alleviating energy waste.

[0119] 4. Voice command priority allocation: Although this part is not explicitly mentioned in the patent claims, as an extended solution, a multi-level voice parsing engine can be introduced to interpret multi-source command input requests and allocate actions to each smart terminal based on logical judgment (such as security emergency level, functional urgency, or initiator authority, etc. to define the priority), thereby avoiding the risk of conflicts caused by task overlap.

[0120] 5. Seasonally Optimized Air Conditioning: Air conditioning operating parameters, such as the heating / cooling capacity adjustment ratio, are set based on real-time readings from built-in temperature and humidity sensors and the region's specific climate patterns. This ensures optimal temperature and humidity levels for the human body in both extreme cold and heat. These measures ensure comfort while minimizing energy loss due to improper settings, perfectly adapting to the complex changes in indoor and outdoor conditions during the seasonal cycle, which place greater demands on air conditioning management.

[0121] In summary, this method makes full use of advanced sensing technology and machine learning algorithms to achieve refined and personalized home service delivery, truly achieving the dual goals of improving convenience and saving resources.

[0122] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and alterations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A smart home Internet of Things interaction method, characterized in that: Including: Based on the user's daily routine and changes in environmental light, dynamically adjust the indoor light brightness to match the user experience requirements, collect real-time air quality and historical data for analysis, intelligently adjust the operating mode of the air purifier to ensure respiratory health, optimize the energy consumption distribution according to the behavior habits of family members and the usage of household appliances, and generate energy-saving control instructions. Combine the temperature and humidity sensor data and seasonal characteristics to flexibly adjust the air-conditioning cooling or heating power to improve indoor comfort; The dynamic adjustment of indoor light brightness based on the user's daily routine and changes in environmental light further includes: taking the average daily sleep time S of the user as a reference parameter; Judge the possible state of the user according to the current time T and S: if TS < 0, then execute the rest mode, and the formula is Light_Intensity = Brightness_Max * exp(a * (TS)^2), where Brightness_Max is the maximum light brightness and a is the sensitivity coefficient; collect the light sensor data Env_Light in the daytime environment and judge whether it is lower than the threshold Th: If Env_Light < Th, Light_Step = Step_Default * (1 - Env_Light / Th). Smoothly transition the light brightness through a progressive adjustment strategy to avoid glare.

2. A smart home Internet of Things interaction method according to claim 1, characterized in that: The intelligent adjustment of the operating mode of the air purifier includes: Collect the real-time air quality index AQ_Index and the average air quality value AVG_INDEX in the last 24 hours; Set the quality standard interval Q_STANDARD = (Q_Low, Q_High) as the good state range, and judge by the formula: If AQ_Index > AVG_INDEX, Run_Speed = Run_Slow × AQ_Index / Q_High^b, where b is the acceleration ratio parameter; When the sensor detects that the particulate matter concentration exceeds the preset warning level ALARM_LEVEL, enter the emergency purification mode; Generate a daily report and push it to the user's mobile phone to prompt the air quality analysis results and improvement suggestions.

3. The method for interacting with the smart home Internet of Things according to claim 1, wherein: In the analysis of the impact of the behavior habits of family members on the energy consumption distribution, it includes: Statistical distribution curve of the appliance startup time P in the historical data to extract the peak interval; Calculate the total power consumption Power_Sum in each time period and the number of members Member_count in that period, and obtain the load distribution coefficient Member-count_Allocate = k * Power_Sum / (Member_Count + β), where k and β are system optimization factors; Combine the peak electricity consumption warning notice to preferentially reduce the power demand of non-core equipment; Form an energy-saving target feedback loop to verify whether the improvement efficiency meets the expectation after a fixed period.

4. A smart home Internet of Things interaction method according to claim 1, wherein The smart speaker receives multi-source tasks T_List, classifies them weighted according to the source device type, and forms a basic score Priority_Basic; For the same type of instruction group I_List, use the formula `Priority_Adjust = Priority_Basic(Distance_Centroid^γ+Time_Dif_Std×γ)` to adjust the weight level, where Distance_Centroid is the center distance deviation and Time_Dif_Std is the mean difference parameter; When a conflict occurs, the one with the highest score will be activated first, otherwise the power will be evenly divided or executed in turns; Save the execution history so that the priority rule set can be quickly called in subsequent similar scenarios.

5. The method for interacting with the smart home Internet of Things according to claim 1, wherein: The matching algorithm between the temperature and humidity sensor and the air conditioning control includes: Read the current temperature and humidity combination H_Com and map it to a seasonal correction value `Season_Modify = α × H_Com^(δ / T_Outer)`, where T_Outer is the outside temperature and δ is the seasonal adjustment term. Determine the comfort limit range Comf_Limit = (Cmin, Cmax), and activate the automatic compensation mechanism when it exceeds the limit; For refined control of cooling power R, based on the target temperature T_set, the relationship `R_Ajusted=Base_Pwr×(T_SetSensor_Read)+ζ×Temp_RisingRate` is obtained. Provides multiple operation panels for users to manually switch between different work preference presets.

6. The method for interacting with the smart home Internet of Things according to claim 1, wherein: Introducing external weather forecast data to assist in air conditioning adjustments includes: Integrate the weather forecast Weather_Data for the next 72 hours and select the significant fluctuation segments to mark the key nodes K_List; The impact degree is calculated using the calibration formula parameter `Calib_Factor=f(Weather_Data_Match)=∑(W_TmpTarget_Tmp)*(Δt^λ)` based on the historical meteorological comparison database; Enhance the system standby response level one hour before extreme weather is about to occur; Record the actual results and compare them with the initial strategy to evaluate the error rate and store the revised version of the model for next use.

7. The smart home Internet of Things interaction method according to claim 1, characterized in that: Customizing more personalized lighting effect logic based on user habits includes: Define the personalized indicator matrix Profile_mat and initialize the vector length Vector_length; The module `Profile_Upt = Ψ(Vector_length × Profilemat + New_Data_Input * σ(Tune_Level) / μ)` is updated by layer-by-layer superposition formula to capture subtle preference changes; Set the number of learning iterations N_limit to check if there is a persistent deviation so that training can be restarted; Syncing to other sub-devices ensures that the unified theme style continues without disrupting the overall coordination.

8. The method for interacting with the smart home Internet of Things according to claim 1, wherein: The process of enhancing the accuracy of air quality analysis includes: Get the long-term PM5 data stream PM_List and perform dimensionality reduction transformation Dataspace_reduction; Use the dynamic window function Dynamic_Window to locate the short-term anomaly point `Peak_Detect = Dataspace_Reduction × [log(Variances) / ω+Threshold]^ρ` to trigger the early warning signal; Re-aggregate the information obtained from the fine-grained monitoring network to draw a risk distribution heat map; Suggestions for local improvements are made for high-risk areas, such as opening ventilation valves or adding activated carbon adsorption layers.

9. The method for interacting with the smart home Internet of Things according to claim 1, wherein: Home appliance energy consumption planning methods include: Introduce sub-item metering chips to track the output power details of each channel Energy_Detail; Use multiple linear regression to predict the trend of the next week, assuming that the model parameter fit must be higher than the threshold Min_Fit_Rate; Implement dynamic pricing simulation Cost_Plan = `Sum(Base_Cost×Predict_Energy×Peak_Flag)` to estimate financial expenditure; Develop a stage cost management checklist for regular review and timely budget adjustment.

10. The smart home Internet of Things interaction method according to claim 1, characterized in that: New strategies to improve the efficiency of voice interaction include: Establish user intention clustering Cluster_Group and reduce redundant computing overhead Cluster_Reduce; Apply deep learning neural network DL_network to extract the association probability of high-frequency phrases Feature_Words; Design a fuzzy decision engine to evaluate the answer scores of similar questions Rank_Score = `(Confidence_Value×Word_Relevance×Context_Similarity+Ω)^η` to find the best matching result; Recording the conversation path facilitates future retrieval of similar cases and speeds up feedback.

Images