Data acquisition and analysis method of intelligent sofa

Abstract

The invention provides a data acquisition and analysis method of an intelligent sofa, which comprises the following steps: arranging an electrostatic sensor array on the intelligent sofa, and acquiring electrostatic signal intensity and polarity distribution, human body charge and environmental humidity data of different areas on the surface of the sofa in real time to obtain electrostatic distribution data; under different air quality conditions including pollutant concentration and particulate matter content, according to the electrostatic characteristic parameters, a mapping relation model of the electrostatic characteristic parameters and the air quality is trained through a support vector machine algorithm; inputting sofa static signal intensity collected in real time into the trained mapping relation model of static characteristics and air quality to obtain changes of the static characteristics, and evaluating indoor air quality in real time; if the evaluation result shows that the air quality is smaller than the air quality threshold value, an air purification device of the intelligent sofa is triggered, an air purifier is started, and the ventilation amount of a fresh air system is increased so as to improve the indoor air environment.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a data collection and analysis method for a smart sofa. Background Art

[0002] Smart sofas play an important role in people's daily lives. They not only provide a comfortable resting experience, but also become an important carrier for home environment monitoring. Different sitting postures of people on the sofa will lead to differences in static electricity accumulation. This phenomenon raises an interesting technical question: Can the air quality at home be analyzed by collecting static electricity data on the sofa? Static electricity is caused by the uneven distribution of charge on the surface of objects, and factors such as dust and pollutants in the air will affect the accumulation of static electricity. Sofas of different materials and shapes have different static electricity accumulation. In addition, the human body itself also carries a certain charge, and charge transfer occurs when it comes into contact with the sofa. Considering these factors, how to accurately collect and distinguish the static electricity signals on the sofa and associate them with air quality is a complex technical challenge. The collection of static electricity data requires highly sensitive sensors. At the same time, the data processing and analysis algorithms also need to consider many interference factors, such as human movement and environmental humidity. How to extract effective static electricity features in a noisy background and build a reliable air quality assessment model to achieve a more comprehensive and detailed environmental perception and provide new possibilities for people to create a healthier and more comfortable living environment is a technical problem that needs to be solved urgently. Summary of the invention

[0003] The present invention provides a data collection and analysis method for a smart sofa, which mainly includes:

[0004] By setting up an electrostatic sensor array on the smart sofa, the electrostatic signal intensity and polarity distribution, human body charge, and environmental humidity data of different areas on the sofa surface are collected in real time to obtain electrostatic distribution data;

[0005] Based on the static electricity distribution data, according to different sofa materials and shapes, a static electricity accumulation model is established, and static electricity measurement equipment is used to conduct accumulation experiments under different conditions to obtain static electricity distribution measurement data, and the static electricity measurement data is normalized;

[0006] Based on the normalized electrostatic measurement data, wavelet transform and frequency domain analysis are used for signal processing to extract electrostatic characteristic parameters used to reflect the data change trend, obtain air quality indicators, and use correlation analysis methods to obtain the correlation strength between different electrostatic characteristic parameters and air quality indicators, and screen out electrostatic characteristic parameters associated with air quality;

[0007] Under different air quality conditions, including pollutant concentration and particulate matter content, the mapping relationship model between electrostatic characteristic parameters and air quality is trained through the support vector machine algorithm based on the electrostatic characteristic parameters;

[0008] The real-time collected sofa electrostatic signal strength is input into the trained mapping relationship model between electrostatic features and air quality to obtain the changes in electrostatic features and evaluate the indoor air quality in real time.

[0009] If the evaluation result shows that the air quality is lower than the air quality threshold, the air purification device of the smart sofa will be triggered to turn on the air purifier and increase the ventilation volume of the fresh air system to improve the indoor air environment.

[0010] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0011] The present invention discloses a data collection and analysis method for a smart sofa. The present invention sets a high-sensitivity electrostatic sensor array on the smart sofa to collect the electrostatic signal intensity and polarity distribution, human body charge, and environmental humidity data in different areas of the sofa surface in real time to obtain electrostatic distribution data. Combined with the geometric shape data of the sofa, the detailed features of the fold seams on the sofa surface are extracted, the fold seam area is judged, the electrostatic accumulation risk coefficient of the area is determined, and an electrostatic accumulation model is established. Wavelet transform and frequency domain analysis signal processing are used to extract electrostatic characteristic parameters. Through a machine learning algorithm, a mapping relationship model between electrostatic characteristic parameters and air quality is established to achieve real-time evaluation of indoor air quality. When the air quality is lower than a threshold, the air purification device is automatically turned on to improve the indoor air environment. Through the association between electrostatic signals and air quality, the present invention realizes intelligent air quality monitoring and regulation, and improves the comfort and health level of the indoor environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 The present invention is a flow chart of a data collection and analysis method for a smart sofa. DETAILED DESCRIPTION

[0013] The following will describe the technical solutions in the embodiments of the present invention in detail in conjunction with the accompanying drawings in the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention.

[0014] like Figure 1 In this embodiment, a data collection and analysis method for a smart sofa may specifically include:

[0015] Step S101, by setting an electrostatic sensor array on the smart sofa, the electrostatic signal strength and polarity distribution, human body charge, and environmental humidity data of different areas on the sofa surface are collected in real time to obtain electrostatic distribution data.

[0016] An electrostatic sensor array is set on the surface of the sofa, and the electrostatic sensor array is arranged in a grid shape. An electrostatic sensor is placed at each grid point to obtain electrostatic signal distribution data; the electrostatic signal distribution data is filtered and denoised by wavelet transform to obtain filtered electrostatic signal data; the electric field intensity distribution on the sofa surface is calculated according to the filtered electrostatic signal data, and the electrostatic accumulation situation in the human body contact area is identified by analyzing the charge density to obtain electrostatic accumulation area data; the electrostatic accumulation area data is classified by using a K-means clustering algorithm to identify different types of electrostatic interference sources and obtain electrostatic interference source classification results; and the electrostatic safety status is evaluated according to the electrostatic interference source classification results.

[0017] For example, an electrostatic sensor array is set on the sofa surface according to the electrostatic sensor sensitivity characteristics, arranged in a grid, with an electrostatic sensor placed at each grid point and a sensor spacing of 10 cm. The sensor detection threshold is determined from the electrostatic signal intensity range, the sensor response time is adjusted by the signal sampling accuracy and the real-time acquisition frequency, and the sensor array density layout is optimized by the data transmission bandwidth to obtain a high-precision electrostatic signal distribution map. Based on the initial measurement results, the sensor sensitivity is dynamically adjusted using the gradient descent method.

[0018] Wavelet transform is used to filter and reduce noise of the collected electrostatic signal intensity and polarity distribution, and the signal feature extraction method is used to separate the human body charge and environmental electrostatic interference. The weighted average method is used to integrate the electrostatic distribution data of different areas on the sofa surface to generate a global electrostatic distribution map. The electrostatic distribution map is analyzed in time series, and the moving average method is used to identify the trend of static accumulation.

[0019] The electric field intensity distribution on the sofa surface is calculated based on the static distribution diagram, and the static electricity accumulation in the human body contact area is identified through the charge density analysis algorithm. The electrostatic force is calculated using Coulomb's law, and the electrostatic discharge risk is predicted based on the surface charge density distribution. The static electricity accumulation trend is determined based on the time series analysis of the static electricity distribution data, and the static electricity safety status assessment result of the sofa surface is obtained.

[0020] The K-means clustering algorithm is used to classify the electrostatic distribution data and identify different types of electrostatic interference sources. The spatial correlation analysis of the electrostatic distribution data is used to optimize the sensor array layout, calculate the Pearson correlation coefficient of adjacent sensor data, and adjust the sensor density. The sensor sampling frequency is automatically adjusted based on the electrostatic safety status assessment results. When the electrostatic accumulation trend is obvious, the sampling frequency is increased, and vice versa, the sampling frequency is reduced to save energy. The data compression rate is dynamically adjusted according to the data transmission bandwidth to optimize the transmission efficiency while ensuring data quality.

[0021] The electrostatic sensor array on the sofa surface is arranged in a grid with a spacing of 10 cm, and an electrostatic sensor with a sensitivity of 0.1pC is placed at each grid point. The sensor detection threshold is set to 0.5pC, the signal sampling accuracy is 16 bits, and the real-time acquisition frequency is 100Hz. The density layout of the sensor array is optimized by using a data transmission bandwidth of 10Mbps to generate a high-precision electrostatic signal distribution map with a resolution of 100x50. The initial measurement results show that the electrostatic signal intensity fluctuates in the range of 0.2-0.8pC. The sensor sensitivity is dynamically adjusted using the gradient descent method with a learning rate of 0.01, and the adjustment is made every 100 samples. The collected electrostatic signal is decomposed by 5 layers using the db4 wavelet transform, and the soft threshold is set to 0.05 for data denoising. By comparing the correlation of adjacent sensor signals, the human body charge characteristics are extracted and separated from the environmental electrostatic interference. The weighted average method is used to integrate the electrostatic distribution data of different regions, and the weight is set according to the signal strength to generate a global electrostatic distribution map of 300x150 pixels. The electrostatic distribution map is analyzed in time series, and the moving average method with a 10-minute window is used to identify the trend of static electricity accumulation. The electric field intensity distribution on the sofa surface is calculated based on the electrostatic distribution diagram, and the electric field intensity range is 0.1-2kV / m. By setting the charge density threshold to 1nC / cm 2 , identify the static electricity accumulation in the human body contact area. Use Coulomb's law to calculate the electrostatic force, combined with 10nC / m 2 The surface charge density distribution is used to predict the risk of electrostatic discharge. When the calculated electrostatic force exceeds 0.1mN, it is judged as a high-risk area. Based on the judgment standard that the electrostatic accumulation rate exceeds 0.5nC / min within 10 minutes, the electrostatic safety status assessment result of the sofa surface is obtained. The K-means clustering algorithm is used to classify the electrostatic distribution data, and the K value is set to 3 to identify three electrostatic interference sources: human body, fabric friction and environment. The Pearson correlation coefficient of adjacent sensor data is calculated, and the sensor density in the area is increased when the coefficient is lower than 0.5. According to the evaluation results of the electrostatic safety status, the sampling frequency is increased to 200Hz when the electrostatic accumulation trend is obvious, and vice versa, it is reduced to 50Hz to save energy. The data transmission bandwidth is dynamically adjusted. When the bandwidth utilization rate exceeds 80%, the data compression ratio is increased from 1:1 to 2:1, optimizing the transmission efficiency while ensuring data quality.

[0022] Step S102, based on the static electricity distribution data, according to different sofa materials and shapes, establish a static electricity accumulation model, use static electricity measurement equipment to perform accumulation experiments under different conditions, obtain static electricity distribution measurement data, and normalize the static electricity measurement data.

[0023] A clustering algorithm is used to classify sofa materials to obtain leather, fabric, and suede material categories; the corresponding resistivity range is determined according to the material category to obtain material characteristic parameters; a static electricity accumulation model is constructed by combining the material characteristic parameters with the three-dimensional shape data of the sofa; an experimental scheme is designed according to the static electricity accumulation model, and a static electricity accumulation experiment is carried out under different conditions specified in the experimental scheme using static electricity measurement equipment; data cleaning is performed on the static electricity measurement data obtained from the static electricity accumulation experiment, and the missing data in the static electricity measurement data is supplemented by the cubic spline interpolation method; the static electricity measurement data is unified into the range of 0-1 by the maximum and minimum value normalization method to obtain normalized static electricity distribution data; and a static electricity accumulation prediction model is established based on the normalized static electricity distribution data.

[0024] Exemplarily, the sofa material is classified according to the electrostatic distribution data, and the K-means clustering algorithm is used to divide the material into leather, cloth, suede and other categories, and the resistivity range is determined for each material. By measuring the surface charge density of the material, the electrostatic characteristic parameters, including the charge decay time constant and the maximum charge accumulation, are calculated in combination with the resistivity. Combined with the three-dimensional shape data of the sofa, an electrostatic accumulation model is constructed, which takes into account the material characteristics, geometry and environmental factors. Based on the constructed electrostatic accumulation model, the experimental scheme is designed and the electrostatic measurement equipment is used to carry out accumulation experiments under different conditions. A constant temperature and humidity chamber is used to control the ambient temperature in the range of 20-30°C and the relative humidity varies between 30-70%. Human body contact is simulated by a standard friction device, and the friction intensity is controlled within the range of 0.1-1N. The electric field strength on the sofa surface is measured using an electrostatic field strength meter, ranging from 0-100kV / m. The electrostatic voltage values ​​at different positions are recorded using an electrostatic voltmeter, and the measurement range is 0-10kV. The electrostatic leakage current is measured using an electrostatic ammeter, and the measurement range is 0-100μA. Each set of experimental conditions was measured 5 times to obtain a complete static distribution measurement data set. The static measurement data was cleaned, and the outlier threshold was set to the mean ± 3 times the standard deviation, and the data points outside this range were removed. The cubic spline interpolation method was used to fill in the missing data to ensure the continuity and smoothness of the data. The static data of different dimensions were unified into the range of 0-1 by the maximum and minimum normalization method. The calculation formula was X normalization = (X-Xmin) / (Xmax-Xmin), where X was the original data, and Xmin and Xmax were the minimum and maximum values ​​of this type of data, respectively. A standardized static distribution data set was obtained, which contained information such as material type, shape characteristics, environmental conditions, static intensity and distribution. Based on the normalized static distribution data, a static accumulation prediction model was established using the random forest regression algorithm. The input variables include material type, shape characteristics, environmental conditions and other factors, and the output variables are static intensity and distribution. The model accuracy was evaluated using ten-fold cross validation, and the root mean square error (RMSE) and determination coefficient (R2 ) as the evaluation metric. The hyperparameters of random forest were optimized by grid search, including the number of trees, ranging from 50 to 500, and the maximum depth, ranging from 5 to 20. The model with the smallest RMSE and R 2 The maximum parameter combination is used to obtain the final electrostatic accumulation model. This model can predict the electrostatic accumulation of sofas of different materials and shapes under various environmental conditions. During the electrostatic distribution data collection process, the K-means clustering algorithm is used to classify 100 sofa material samples, and the cluster number K is set to 5. After 50 iterations, five categories of leather, fabric, suede, synthetic leather and composite materials are obtained. For each material, the four-probe method is used to measure the resistivity. For example, the resistivity range of leather materials is 10^6-10^9Ω·m. The surface charge density of the material is measured using an electrostatic coulometer with a value range of 0.1-10nC / cm 2. Combining resistivity and charge density, the charge decay time constant τ=ρε is calculated, where ρ is resistivity, ε is dielectric constant, and τ ranges from 0.1 to 100 seconds. The sofa shape data is obtained by a 3D scanner with a resolution of 1 mm. The constructed static accumulation model takes into account material properties, geometric shape and environmental factors, and the finite element method is used for numerical simulation. The experimental scheme is designed based on the model, and a constant temperature and humidity chamber is used to control the environment. The temperature range is 20-30℃, the relative humidity is 30-70%, and the step size is 2℃ and 5%. The standard friction device simulates human contact, with a friction force range of 0.1-1N and a step size of 0.1N. The electrostatic field strength meter has a measurement range of 0-100kV / m and an accuracy of ±1%. The electrostatic voltmeter has a measurement range of 0-10kV and a resolution of 1V. The electrostatic ammeter has a measurement range of 0-100μA and an accuracy of ±0.1%. The measurement is repeated 5 times for each group of conditions, and a total of 10,000 data points are obtained. During the data cleaning process, the outlier threshold was set to μ±3σ, μ was the mean value, σ was the standard deviation, and about 300 outliers were removed. The cubic spline interpolation method was used to fill in 50 missing data points. After normalization, a standardized static distribution data set of 9750 valid data points was obtained. The static accumulation prediction model was established using the random forest regression algorithm. The input variables included 5 categories of material type, 10 parameters of shape characteristics, and 3 variables of environmental conditions. The output variables were 5 parameters of static intensity and distribution. The materials can be synthetic fiber materials, such as polyester fibers, natural fiber materials, such as cotton or wool, leather materials, including artificial leather or genuine leather, plastic or polymer materials, such as polyvinyl chloride (PVC), wood or wood composite materials. The shape features may be: total length is the overall length of the sofa, total width is the overall width of the sofa, total height is the overall height of the sofa, seat area is the area of ​​the seat part, seat depth is the distance from the front to the back of the seat, seat height is the height from the seat surface to the ground, armrest height is the height from the armrest to the seat surface, armrest width is the width of the armrest, backrest angle is the angle between the backrest and the seat surface, and backrest height is the height of the backrest. Environmental conditions include the temperature, relative humidity and atmospheric pressure of the environment. The electrostatic strength and distribution may include the maximum electrostatic voltage is the maximum electrostatic voltage that may appear on the surface of the sofa, the average electrostatic voltage is the average electrostatic voltage on the surface of the sofa, the electrostatic energy is the electrostatic energy accumulated on the surface of the sofa, the electrostatic field strength is the electrostatic field strength around the sofa, and the charge distribution is the charge distribution on the surface of the sofa, such as the charge density in different parts. The model accuracy was evaluated by ten-fold cross validation, with an RMSE of 0.15 and a determination coefficient R 2 The grid search method was used to optimize the hyperparameters, and the number of trees was finally determined to be 300, the maximum depth was 12, and the prediction accuracy of the electrostatic accumulation model obtained on the test set reached 92%.

[0025] According to the geometric shape data of the sofa, the detailed features of the pleats and seams on the sofa surface are extracted, the pleat and seam areas are judged, and the electrostatic accumulation risk coefficient of the area is determined. Combined with the material and electrostatic distribution data of different sofas, a static electricity accumulation model is established.

[0026] According to the three-dimensional scanning data of the sofa, the curvature distribution of the sofa surface is calculated by the discrete Gaussian curvature calculation method to obtain the surface wrinkle area. For the surface wrinkle area, the seam line features are extracted to determine the precise position and range of the wrinkle seam area. The seam depth and width of the wrinkle seam area are measured by the structured light three-dimensional scanning method, and the geometric and electrical characteristic model of the wrinkle seam area is established. Electrostatic field simulation is performed to obtain the electric field intensity distribution of the wrinkle seam area. If the electric field intensity distribution exceeds the preset threshold, it is determined that there is a risk of static electricity accumulation in the wrinkle seam area. The electrostatic characteristic parameters of the sofa material are extracted from the pre-established material database, including dielectric constant, surface resistivity and charge decay time constant. Combined with the charge density and electric field intensity information in the electrostatic distribution data, a basic model of static electricity accumulation combined with material characteristics is constructed. A multi-layer perceptron neural network is used to construct a static electricity accumulation prediction model, the input layer of the neural network contains material parameters, geometric features and environmental factors, and the output layer represents the amount of static electricity accumulation in different areas. The model parameters of the neural network are optimized by the back propagation algorithm to obtain the final static electricity accumulation model.

[0027] For example, based on the three-dimensional scanning data of the sofa, the discrete Gaussian curvature calculation method is used to calculate the curvature distribution of the sofa surface, and the surface wrinkle area is identified by setting the curvature threshold to 0.1cm^(-1). The Canny edge detection algorithm is used to extract the seam line features, and the double threshold is set to a low threshold of 50 and a high threshold of 150. Combining the curvature and edge information, the precise position and range of the wrinkle seam area are determined through the expansion and corrosion operations in the morphological operation. For the identified wrinkle seam area, the seam depth and width are measured using the structured light three-dimensional scanning technology, and the scanning resolution is set to 0.1mm. Combined with the material resistivity data, the geometric and electrical characteristics model of the wrinkle seam area is established. The electrostatic field simulation is performed by COMSOL Mu lt iphys ics software, the grid size is set to 0.5mm, and the MUMPS direct solver is selected as the solver to calculate the electric field intensity distribution in the area. According to the maximum value and gradient of the electric field intensity, the electrostatic aggregation risk coefficient of the wrinkle seam area is obtained. Based on the sofa material classification results, the corresponding electrostatic characteristic parameters are extracted from the pre-established material database, including dielectric constant range 1-10, surface resistivity, unit Ω / sq, range 10^6-10^14 and charge decay time constant, range 0.1-100s. Combined with the charge density (unit C / m^2) and electric field strength (unit V / m) information in the electrostatic distribution data, a basic model of electrostatic accumulation considering material characteristics is constructed. The model uses an exponential decay function to describe the change of electrostatic charge over time. The electrostatic accumulation risk coefficient of the wrinkled seam area is used as a weight factor and fused with the basic model of electrostatic accumulation. A multi-layer perceptron neural network is used to construct a static accumulation prediction model. The network structure is three-layer, with 10 neurons in the input layer, including 3 material parameters, 4 geometric features, and 3 environmental factors. The number of neurons in the hidden layer is 20, and the number of neurons in the output layer is 5, which represents the amount of static electricity accumulation in different areas. The output of the basic model of static electricity accumulation is used as an additional input feature of the neural network. The model parameters were optimized by back propagation algorithm, the learning rate was set to 0.01, and the number of iterations was 1000 to obtain the final static electricity accumulation model. In practical applications, the sofa was first scanned using a high-precision 3D laser scanner to obtain point cloud data with a resolution of 0.5mm. The point cloud data was processed using the discrete Gaussian curvature calculation method to calculate the curvature value of each point, and the curvature threshold was set to 0.1cm^(-1) to identify the surface wrinkle area. The Canny edge detection algorithm was then applied with the low threshold set to 50 and the high threshold set to 150 to extract the seam line features. Through the morphological expansion and corrosion operations of 3x3 pixels, the wrinkle area and the seam features were fused to accurately locate the wrinkle seam area. Using structured light 3D scanning technology, the seam depth and width were measured at a resolution of 0.1mm, and data such as the depth range of 1-5mm and the width range of 2-8mm were obtained.The geometric data was combined with the pre-measured material resistivity, such as leather 10^9Ω·m, to establish an electrostatic field model in COMSOL Multiphysics software. The grid size was set to 0.5mm, and the electric field distribution was calculated using the MUMPS direct solver. The maximum electric field strength was 5kV / m, and the electrostatic accumulation risk factor was calculated to be 0.8. Electrostatic characteristic parameters were extracted from the material database, such as the dielectric constant of leather was 2.5, the surface resistivity was 10^11Ω / sq, and the charge decay time constant was 5s. Combined with the measured charge density 2x10^-6C / m^2 and the electric field strength data, a basic model of electrostatic accumulation was constructed. The exponential decay function Q(t)=Q_0*e^(-t / τ) was used to describe the change of electrostatic charge over time. Q(t) is the amount of electrostatic charge remaining at time t, Q0 is the initial moment, which refers to the amount of electrostatic charge at t=0, t is time, the time elapsed from the initial moment, and τ is the time constant, which indicates the time required for the charge to decay to 1 / e (about 36.8%) of the initial value. The time constant τ reflects the speed of electrostatic charge decay. The larger the τ, the slower the charge decays, and vice versa. Finally, a three-layer multi-layer perceptron neural network is constructed. The 10 neurons in the input layer include 3 material parameters, including dielectric constant, surface resistivity, and decay time constant, 4 geometric features, including curvature, depth, width, and area, 3 environmental factors, including temperature, humidity, and number of frictions, 20 neurons in the hidden layer, and 5 neurons in the output layer represent the amount of electrostatic accumulation in different regions. The output of the basic model is used as an additional input feature, and the parameters are optimized using the back propagation algorithm with a learning rate of 0.01 and 1000 iterations. The final model achieves a prediction accuracy of 95% on the test set.

[0028] Step S103, based on the normalized electrostatic measurement data, wavelet transform and frequency domain analysis are used to perform signal processing, electrostatic characteristic parameters used to reflect the data change trend are extracted, air quality indicators are obtained, and correlation analysis methods are used to obtain the correlation strength between different electrostatic characteristic parameters and air quality indicators, and electrostatic characteristic parameters associated with air quality are screened out.

[0029] The normalized electrostatic measurement data is subjected to wavelet transform, and denoised by soft threshold method; the denoised electrostatic data is subjected to fast Fourier transform to obtain frequency domain representation, calculate power spectrum density, and extract frequency domain feature parameters, including main frequency, frequency band energy ratio, and harmonic distortion. For the frequency domain feature parameters, an electrostatic feature parameter set is formed in combination with time domain features to establish the association between electrostatic features and air quality. The synchronously collected air quality data is obtained, the air quality data is cleaned and preprocessed, the moving median method is used to remove outliers, the linear interpolation method is used to fill missing values, and the normalized air quality index data set is obtained by z-score standardization. The Pearson correlation coefficient is used to calculate the correlation between the electrostatic feature parameters and the normalized air quality index. If the correlation coefficient is greater than the preset threshold, the electrostatic feature parameter subset that is strongly correlated with the air quality index is screened out, and the principal component analysis method is used for dimensionality reduction. The first preset principal components whose cumulative contribution rate reaches the preset value are selected, and the electrostatic feature parameters for air quality assessment are obtained by linear combination.

[0030] Exemplarily, the normalized electrostatic measurement data is subjected to wavelet transform, and the db4 wavelet basis function and the 5-level decomposition are selected based on the frequency characteristics and data length of the electrostatic signal. Denoising is performed by the soft threshold method, and the threshold is set to σ√(2logN), where σ is the noise standard deviation, N is the data length, logN is the natural logarithm, and the base of logN is e. After eliminating high-frequency noise, the reconstructed signal obtains a smoothed electrostatic data sequence. The smoothed electrostatic data sequence is subjected to a fast Fourier transform to obtain a frequency domain representation and calculate the power spectral density. The frequency domain characteristic parameters are extracted, including the main frequency, the frequency band energy ratio, and the harmonic distortion. The main frequency is determined by finding the frequency corresponding to the maximum value of the power spectral density, the frequency band energy ratio is obtained by calculating the ratio of the energy in a specific frequency band to the total energy, and the harmonic distortion is obtained by calculating the ratio of the harmonic component to the fundamental component. Combined with time domain features such as mean, variance, skewness, and kurtosis, an electrostatic characteristic parameter set is formed. In order to establish the association between electrostatic characteristics and air quality, air quality data is collected synchronously. The data of air quality indicators such as PM2.5, PM10, ozone, and sulfur dioxide were collected through air quality monitoring equipment. The original air quality data was cleaned and preprocessed, and the moving median method was used to remove outliers, and the linear interpolation method was used to fill in missing values. The processed data was z-score standardized to obtain a normalized air quality index data set with a mean of 0 and a standard deviation of 1. The Pearson correlation coefficient was used to calculate the correlation between the electrostatic characteristic parameters and the air quality indicators. The correlation coefficient threshold was set to 0.7, and the subset of electrostatic characteristic parameters that were strongly correlated with the air quality indicators was screened out. The principal component analysis method was used for dimensionality reduction. First, the screened features were standardized, the covariance matrix between the features was calculated, and the eigenvalue decomposition was performed. The first few principal components with a cumulative contribution rate of 95% were selected, and the key electrostatic characteristic parameters for air quality assessment were finally obtained through linear combination. In practical applications, the 10,000 collected electrostatic measurement data points were first normalized to unify the numerical range to between 0 and 1. Then, the db4 wavelet basis function was applied to the normalized data for 5-level wavelet decomposition to obtain wavelet coefficients of different scales. Denoising was performed by soft threshold method, and the threshold was set to 0.05×√(2log10000)≈0.3, which effectively removed high-frequency noise. After reconstruction, a smoothed electrostatic data sequence was obtained, and the number of data points remained unchanged. A 1024-point fast Fourier transform was performed on the smoothed sequence to calculate the power spectral density. The frequency domain feature parameters were extracted, such as the main frequency was found to be around 50Hz, the 0-100Hz band energy ratio accounted for about 80% of the total energy, and the harmonic distortion was about 5%. Combined with time domain features such as mean 0.5, variance 0.04, skewness 0.1, and kurtosis 3.2, a 15-dimensional electrostatic feature parameter set was formed.Air quality data were collected synchronously, including PM2.5, with a range of 0-500μg / m³, PM10, with a range of 0-600μg / m³, ozone, with a range of 0-500ppb, sulfur dioxide, with a range of 0-500ppb, etc. The sampling interval was 5 minutes, with a total of 24 hours of data. The moving median method was used with a window size of 5 to remove outliers, and the linear interpolation method was used to fill in about 2% of missing values. The normalized air quality index data set was obtained by z-score standardization. The Pearson correlation coefficient between 15 electrostatic characteristic parameters and 4 air quality indicators was calculated, and 8 strongly correlated electrostatic characteristic parameters were screened out by setting a threshold of 0.7. The principal component analysis was performed on these 8 features, and the top 3 principal components with a cumulative contribution rate of 95% were selected, and finally 3 key electrostatic characteristic parameters were obtained for air quality assessment.

[0031] Step S104, under different air quality conditions, including pollutant concentration and particulate matter content, according to the electrostatic characteristic parameters, a mapping relationship model between the electrostatic characteristic parameters and the air quality is trained by a support vector machine algorithm.

[0032] The electrostatic characteristic parameters and the corresponding pollutant concentration and particulate matter content data under different air quality conditions are obtained to construct a training data set; the training data set is subjected to Z-score standardization, and the electrostatic characteristic parameters after standardization are screened by a recursive feature elimination method to determine an optimal feature combination; according to the optimal feature combination, the kernel function type, regularization parameter C and kernel function parameter γ of a support vector machine are optimized by a grid search method to obtain an optimal parameter combination; using the optimal feature combination and the optimal parameter combination, a support vector machine algorithm is used to train a mapping relationship model between the electrostatic characteristic parameters and the air quality; corresponding sub-models are trained for different air quality conditions, and the sub-models are trained based on the mapping relationship model to improve the adaptability of the sub-models under various pollution levels.

[0033] Exemplarily, a training data set is constructed based on the electrostatic characteristic parameters under different air quality conditions and the corresponding pollutant concentration and particulate matter content data. The data is Z-score standardized, the mean μ and standard deviation σ of each feature are calculated, and the data is converted to a standard normal distribution with a mean of 0 and a standard deviation of 1 by (x-μ) / σ, which is suitable for data of different dimensions and distributions. The electrostatic characteristic parameters are screened by recursive feature elimination. The number of features is gradually increased from 1 to the total number of features, and each feature subset is cross-validated 5-fold to calculate the root mean square error. When the increase in the number of features results in a performance improvement of less than 1%, the number is set as the feature number threshold. Select the optimal feature combination to reduce model complexity and improve generalization ability. Based on the optimal feature combination screened out, the kernel function type, regularization parameter C and kernel function parameter γ of the support vector machine are optimized using the grid search method. Set the parameter search range. The kernel function types include linear kernel, polynomial kernel and radial basis kernel function RBF. The range of C is

[0034] [0.1,1,10,100], and the range of γ is [0.001,0.01,0.1,1]. A 5-fold cross validation was used to evaluate the performance of each group of parameters and select the optimal parameter combination. Based on the optimized parameters and feature combinations, the support vector machine algorithm was used to train the mapping relationship model between electrostatic feature parameters and air quality. The sequential minimum optimization algorithm was used to solve the model parameters, and the maximum number of iterations was set to 1000 and the convergence threshold was set to 1e-3. The model performance was evaluated by 10-fold cross validation, and the root mean square error and determination coefficient were calculated to determine the prediction accuracy and fit of the model. For different air quality conditions, the corresponding sub-models were trained separately to improve the adaptability of the model under various pollution levels. In practical applications, 1000 groups of data under different air quality conditions were first collected, each group containing 20 electrostatic feature parameters and corresponding PM2.5, PM10, ozone, and sulfur dioxide concentrations. The data was normalized by Z-score, such as the mean of the original PM2.5 data was 50μg / m 3 , with a standard deviation of 20 μg / m 3 , after standardization, it becomes a distribution with mean 0 and standard deviation 1. The recursive feature elimination method starts with 1 feature and gradually increases to 20, calculating the root mean square error of 5-fold cross validation each time. When it increases to 15 features, the performance improvement drops to 0.8%, which is lower than the 1% threshold, so 15 features are selected as the optimal combination. The support vector machine parameters are optimized by grid search method, trying linear kernel, quadratic polynomial kernel and RBF kernel, C value is searched in [0.1,1,10,100], γ value is in

[0035] The search was performed in [0.001, 0.01, 0.1, 1]. After 5-fold cross validation, the RBF kernel, C = 10, γ = 0.01 were finally selected as the optimal parameters. The support vector machine model was trained using these parameters and features, and the sequential minimum optimization algorithm reached a convergence threshold of 1e-3 after 732 iterations. The 10-fold cross validation results showed that the root mean square error of PM2.5 prediction was 3.2 μg / m 3 , coefficient of determination R 2 The root mean square error of PM10 prediction is 5.1 μg / m 3 , coefficient of determination R 2 is 0.89, the root mean square error of ozone prediction is 4.8 ppb, R 2 is 0.87, the root mean square error of sulfur dioxide prediction is 3.5ppb, R 2 The sub-models were trained for the three conditions of air quality index (AQI) less than 100, 100-200, and greater than 200, and the prediction accuracy on their respective data sets was 5-8% higher than that of the entire model.

[0036] Step S105, inputting the sofa electrostatic signal strength collected in real time into the trained mapping relationship model between electrostatic features and air quality, obtaining the change of electrostatic features, and evaluating the indoor air quality in real time.

[0037] The electrostatic signal sequence of the sofa surface collected by the electrostatic sensor array is obtained, and the electrostatic signal sequence contains preset data points. The electrostatic signal sequence is preprocessed, and the median filtering algorithm is used to remove the burst noise, and the signal is denoised by wavelet transform. The electrostatic characteristic parameters are extracted from the electrostatic signal after signal denoising, and the time domain features, frequency domain features and time-frequency domain features are calculated to obtain a multidimensional feature vector. If the multidimensional feature vector is not standardized, the multidimensional feature vector is standardized so that all feature values ​​are scaled to the [-1, 1] interval. The standardized multidimensional electrostatic feature vector is input into the pre-trained support vector machine model to predict the PM2.5, PM10, ozone and sulfur dioxide concentrations in the current indoor air. The single air quality index is calculated according to the predicted pollutant concentration, and the maximum value is taken as the comprehensive air quality index. The air quality level interval to which the comprehensive air quality index belongs is determined to determine the current indoor air quality level.

[0038] Exemplarily, the electrostatic signal intensity of the sofa surface is collected in real time by a high-precision electrostatic sensor array. According to the changing characteristics of the electrostatic signal, the sampling frequency is set to 100 Hz, and the acquisition time window is 10 seconds, and the original electrostatic signal sequence containing 1000 data points is obtained. The sliding window method is used to update the data every 1 second to maintain the real-time and continuity of the data. The original electrostatic signal sequence is preprocessed, and the median filter algorithm is used to remove the sudden noise. Through experimental comparison, the filter window size is determined to be 5. Then the signal is denoised by wavelet transform, the db4 wavelet basis function is selected, the decomposition layer number is set to 3 layers according to the signal frequency characteristics, and the soft threshold method is used to remove high-frequency noise. The threshold is set to 3 times the standard deviation of the noise. The electrostatic characteristic parameters are extracted from the preprocessed electrostatic signal, and the time domain features such as mean, variance, kurtosis, and skewness are calculated. The power spectral density is calculated by fast Fourier transform, and the frequency domain features such as the main frequency are extracted, which is the frequency corresponding to the maximum value of the power spectral density, and the frequency band energy ratio, which is the ratio of the energy of a specific frequency band to the total energy. Calculate the time-frequency domain features such as wavelet energy entropy, which is the normalized entropy value of the energy of wavelet coefficients at each scale. Finally, a 15-dimensional feature vector is formed and standardized so that all eigenvalues ​​are scaled to the interval [-1,1]. The standardized 15-dimensional electrostatic feature vector is input into the pre-trained support vector machine model, and the model is used to predict the concentrations of PM2.5, PM10, ozone and sulfur dioxide in the current indoor air. PM2.5 refers to particles with an aerodynamic diameter of less than or equal to 2.5 microns. PM10 refers to particles with an aerodynamic diameter of less than or equal to 10 microns. According to the "Ambient Air Quality Standard" (GB3095-2012), the individual air quality index (IAQI) is calculated based on the predicted pollutant concentration, and the maximum value is taken as the comprehensive air quality index (AQI). According to the AQI value, the air quality level is divided into 0-50 for excellent, 51-100 for good, 101-150 for light pollution, 151-200 for moderate pollution, 201-300 for heavy pollution, and greater than 300 for severe pollution. The evaluation results are updated every minute to achieve real-time monitoring and evaluation of indoor air quality. In practical applications, the high-precision electrostatic sensor array collects electrostatic signals on the sofa surface at a frequency of 100Hz, forming a signal sequence containing 1000 data points every 10 seconds. A 1-second sliding window is used to update 100 new data points per second to maintain the real-time nature of the data. The original signal is first median filtered with a window size of 5, which effectively removes 95% of the impulse noise. Subsequently, the db4 wavelet 3-layer decomposition is applied, and the soft threshold is set to 3 times the noise standard deviation (about 0.05mV), which successfully filters out high-frequency noise and improves the signal-to-noise ratio by 8dB. A 15-dimensional feature vector is extracted from the processed signal, including time domain features including mean, which ranges from -0.5mV to 0.5mV, and variance, whose typical value is 0.01mV 2, kurtosis, its normal distribution is about 3, skewness, its range is -0.5 to 0.5; frequency domain features are calculated by 1024-point FFT, the main frequency is usually between 20-50Hz, and the energy ratio of 0-100Hz band is about 80%; wavelet energy entropy is usually between 2.5-3.5. After the feature vector is standardized, it is input into the pre-trained support vector machine model. The root mean square error of the model's prediction of PM2.5 on the test set is 5μg / m 3 , for PM10: 8μg / m 3 , 10ppb for ozone, and 5ppb for sulfur dioxide. Calculate AQI based on the prediction results, such as PM2.5 prediction value of 75μg / m 3 When the air quality level is 104, the corresponding IAQI is 104. The maximum value of the IAQI of all pollutants is taken as the final AQI to determine the air quality level. The evaluation results are updated every minute. In a typical 8-hour test, the air quality level changes three times, from "good" to "slightly polluted" and then back to "good", are accurately captured, and the results of professional air quality monitoring equipment are consistent with 90%.

[0039] Step S106, if the evaluation result shows that the air quality is less than the air quality threshold, the air purification device of the smart sofa is triggered, the air purifier is turned on, and the ventilation volume of the fresh air system is increased to improve the indoor air environment.

[0040] The air quality index measured in real time by the air quality sensor is obtained, and it is determined whether it exceeds a preset air quality threshold according to the air quality index; if the air quality index exceeds the preset air quality threshold, a start instruction is sent to the air purification device through a microcontroller; the air quality index and indoor temperature and humidity data are processed by a fuzzy control algorithm to obtain control instructions for the air purifier working mode and the ventilation volume of the fresh air system; the wind speed of the air purifier and the ventilation volume of the fresh air system are adjusted according to the control instructions; the deviation between the air quality index and the target value is calculated by a PID control algorithm to obtain the adjustment amount of the purification measure intensity; the wind speed of the air purifier and the ventilation volume of the fresh air system are modified according to the adjustment amount; the adjusted air purifier wind speed is calculated by the following formula:

[0041] V fan =V base +△V

[0042] Where V fan Indicates the wind speed of the air purifier, V base represents the basic wind speed, ΔV represents the adjustment amount of the wind speed; the modified wind speed and ventilation volume are sent to the air purification device, and the air purification device performs corresponding purification measures according to the received wind speed and ventilation volume. This formula is used to calculate the adjusted air purifier wind speed.

[0043] Exemplarily, according to the real-time evaluated air quality index AQI, a multi-level air quality threshold is set with reference to the ambient air quality standard GB3095-2012, the excellent level threshold is 100, the light pollution level threshold is 150, the moderate pollution level threshold is 200, and the heavy pollution level threshold is 300. When the AQI exceeds the corresponding threshold, different levels of air purification measures are triggered. Through the ArduinoMega2560 microcontroller built into the smart sofa, a linkage mechanism between the air quality assessment result and the air purification device is established. The microcontroller communicates with the air quality sensor through the I2C interface to obtain real-time AQI data, and controls the air purifier and the fresh air system through the RS485 interface. When the AQI exceeds the set threshold, the microcontroller sends a start command to automatically adjust the purifier wind speed and the fresh air system ventilation volume according to the degree of pollution. A fuzzy control algorithm is used to dynamically adjust the working mode of the air purifier and the ventilation volume of the fresh air system according to the AQI value and indoor temperature and humidity data. The fuzzy control process includes three steps: fuzzification, reasoning, and defuzzification. Convert AQI and temperature and humidity data into fuzzy sets; perform reasoning based on preset fuzzy rules; convert fuzzy output into specific control instructions. For example, when AQI is between 100-150, the air purifier runs at medium speed and the fresh air system works at 30% ventilation volume; when AQI is greater than 200, the air purifier runs at full speed and the fresh air system works at 80% ventilation volume. Set up a feedback control mechanism and use the PID control algorithm to dynamically adjust the intensity of purification measures according to the deviation between AQI and the target value. Adjust the duration and intensity of air purification measures by real-time monitoring of AQI change trends. When AQI is lower than the set threshold for 30 consecutive minutes, the PID controller calculates the corresponding control amount, gradually reduces the air purifier wind speed and the fresh air system ventilation volume until it returns to the standby state, realizing intelligent and energy-saving operation. At the same time, the microcontroller exchanges real-time data with the air purification device through the MQTT protocol to ensure the timely execution of control instructions and real-time update of status feedback. In actual application, the smart sofa integrates a control system based on ArduinoMega2560 and monitors air quality in real time through the SharpGP2Y1010AU0F dust sensor. Data is collected every 5 seconds and the average value is taken for 20 consecutive times to calculate the AQI. When the AQI reaches 101, a light pollution level response is triggered: the air purifier starts and runs at a medium speed of 1200rpm, and the fresh air system is ventilated at 30% of the ventilation volume (about 150m 3 If the AQI rises to 201, a moderate pollution response is triggered: the purifier speed is increased to 2000rpm, and the fresh air system ventilation volume is increased to 80% (about 400m 3The control algorithm uses fuzzy control of triangular membership function. The input variables are AQI (range 0-500) and indoor relative humidity (range 20%-80%), and the output variables are purifier speed (range 0-3000rpm) and fresh air ventilation volume (range 0-500m 3 / h). The fuzzy rule base contains 25 rules, such as "IFAQIis High AND Humidityis Medium THENF an_Speedis High ANDVentilationis High". Defuzzification uses the center of gravity method to obtain the specific control quantity. At the same time, the PID controller takes the target AQI value of 100 as the set point, the proportional coefficient Kp=0.5, the integral time Ti=300s, the differential time Td=60s, and adjusts the control output every 60 seconds. It communicates with the air purification device through the MQTT protocol, and the publish / subscribe topics include "air_quality / aqi", "air_purifier / speed" and "ventilation / flow_rate", and the message refresh rate is 1Hz. In a typical operation, when the AQI is detected to rise from 80 to 130, the purification measures are started within 5 seconds, and the AQI drops to 95 after 15 minutes, and then remains stable for 30 minutes. The purifier and fresh air system are gradually adjusted to a low-speed operation state, achieving precise air quality control and energy efficiency optimization.

[0044] The description of the above embodiments is only used to help understand the technical solutions and core ideas of the present application. Ordinary technical personnel in this field should understand that they can still modify the technical solutions recorded in the aforementioned embodiments, or replace some of the technical features therein with equivalents. However, these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A data collection and analysis method for a smart sofa, characterized in that: The method comprises: By setting up an electrostatic sensor array on the smart sofa, the electrostatic signal intensity and polarity distribution, human body charge, and environmental humidity data of different areas on the sofa surface are collected in real time to obtain electrostatic distribution data; Based on the static electricity distribution data, according to different sofa materials and shapes, a static electricity accumulation model is established, and static electricity measurement equipment is used to conduct accumulation experiments under different conditions to obtain static electricity distribution measurement data, and the static electricity measurement data is normalized; Based on the normalized electrostatic measurement data, wavelet transform and frequency domain analysis are used for signal processing to extract electrostatic characteristic parameters used to reflect the data change trend, obtain air quality indicators, and use correlation analysis methods to obtain the correlation strength between different electrostatic characteristic parameters and air quality indicators, and screen out electrostatic characteristic parameters associated with air quality; Under different air quality conditions, including pollutant concentration and particulate matter content, the mapping relationship model between electrostatic characteristic parameters and air quality is trained through the support vector machine algorithm based on the electrostatic characteristic parameters; The real-time collected sofa electrostatic signal strength is input into the trained mapping relationship model between electrostatic features and air quality to obtain the changes in electrostatic features and evaluate the indoor air quality in real time. If the evaluation result shows that the air quality is lower than the air quality threshold, the air purification device of the smart sofa will be triggered to turn on the air purifier and increase the ventilation volume of the fresh air system to improve the indoor air environment.

2. The method according to claim 1, characterized in that The electrostatic sensor array is arranged on the smart sofa to collect the electrostatic signal intensity and polarity distribution, human body charge, and environmental humidity data in different areas of the sofa surface in real time to obtain the electrostatic distribution data, including: An electrostatic sensor array is set on the surface of the sofa, wherein the electrostatic sensor array is arranged in a grid shape, and an electrostatic sensor is placed at each grid point to obtain electrostatic signal distribution data; Using wavelet transform to filter and reduce noise on the electrostatic signal distribution data to obtain filtered electrostatic signal data; Calculate the electric field intensity distribution on the sofa surface according to the filtered electrostatic signal data, identify the electrostatic accumulation in the human body contact area by analyzing the charge density, and obtain the electrostatic accumulation area data; The electrostatic accumulation area data is classified and processed using a K-means clustering algorithm to identify different types of electrostatic interference sources and obtain a classification result of the electrostatic interference source; The electrostatic safety status is evaluated according to the electrostatic interference source classification result.

3. The method according to claim 1, characterized in that Based on the static electricity distribution data, according to different sofa materials and shapes, a static electricity accumulation model is established, static electricity measurement equipment is used to perform accumulation experiments under different conditions, static electricity distribution measurement data is obtained, and static electricity measurement data is normalized, including: Clustering algorithm is used to classify sofa materials, and leather, fabric and suede material categories are obtained; Determine the corresponding resistivity range according to the material category and obtain material characteristic parameters; Combining the material characteristic parameters with the three-dimensional shape data of the sofa, constructing a static electricity accumulation model; Designing an experimental scheme according to the electrostatic accumulation model, and using an electrostatic measurement device to conduct an electrostatic accumulation experiment under different conditions specified in the experimental scheme; Performing data cleaning on the electrostatic measurement data obtained from the electrostatic accumulation experiment, and using a cubic spline interpolation method to complete missing data in the electrostatic measurement data; The static electricity measurement data is normalized to a range of 0-1 by a maximum and minimum value normalization method to obtain normalized static electricity distribution data; Establishing a static electricity accumulation prediction model based on the normalized static electricity distribution data; It also includes: extracting the detailed features of the pleats and seams on the sofa surface based on the geometric shape data of the sofa, judging the pleat and seam area, determining the electrostatic accumulation risk coefficient of the area, and establishing a static electricity accumulation model by combining the materials and electrostatic distribution data of different sofas.

4. The method according to claim 3, characterized in that The method extracts the detailed features of the fold seams on the sofa surface according to the geometric shape data of the sofa, determines the fold seam area, determines the electrostatic accumulation risk coefficient of the area, and establishes an electrostatic accumulation model by combining the materials and electrostatic distribution data of different sofas, including: According to the 3D scanning data of the sofa, the curvature distribution of the sofa surface is calculated by using the discrete Gaussian curvature calculation method to obtain the surface wrinkle area; For the surface wrinkle area, extract the seam line features to determine the precise position and range of the wrinkle seam area; Using a structured light three-dimensional scanning method to measure the seam depth and width of the fold seam area, and establishing a geometric and electrical characteristic model of the fold seam area; Performing electrostatic field simulation to obtain the electric field intensity distribution in the fold seam area; If the electric field intensity distribution exceeds a preset threshold, it is determined that there is a risk of static electricity accumulation in the wrinkle seam area; Extract the electrostatic characteristic parameters of the sofa material from the pre-established material database, including dielectric constant, surface resistivity and charge decay time constant; Combine the charge density and electric field strength information in the static electricity distribution data to build a basic static electricity accumulation model that combines material properties; A static electricity accumulation prediction model is constructed using a multi-layer perceptron neural network, wherein the input layer of the neural network includes material parameters, geometric features and environmental factors, and the output layer represents the static electricity accumulation amount in different areas; The model parameters of the neural network are optimized by a back propagation algorithm to obtain a final static electricity accumulation model.

5. The method according to claim 1, characterized in that The method is based on the normalized electrostatic measurement data, uses wavelet transform and frequency domain analysis to perform signal processing, extracts electrostatic characteristic parameters used to reflect the data change trend, obtains air quality indicators, uses correlation analysis methods to obtain the correlation strength between different electrostatic characteristic parameters and air quality indicators, and screens out electrostatic characteristic parameters associated with air quality, including: The normalized electrostatic measurement data is subjected to wavelet transformation and denoised using the soft threshold method; Perform fast Fourier transform on the denoised electrostatic data to obtain frequency domain representation, calculate the power spectrum density, and extract frequency domain characteristic parameters from it, including main frequency, frequency band energy ratio and harmonic distortion; For the frequency domain characteristic parameters, combined with the time domain characteristics, an electrostatic characteristic parameter set is formed to establish an association between the electrostatic characteristics and the air quality; Acquire the synchronously collected air quality data, perform data cleaning and preprocessing on the air quality data, remove outliers using the moving median method, fill in missing values ​​using the linear interpolation method, and obtain a normalized air quality index data set through z-score standardization; The Pearson correlation coefficient is used to calculate the correlation between the electrostatic characteristic parameters and the standardized air quality indicators. If the correlation coefficient is greater than the preset threshold, the subset of electrostatic characteristic parameters that are strongly correlated with the air quality indicators is screened out, and the principal component analysis method is used for dimensionality reduction. The first preset principal components whose cumulative contribution rate reaches the preset value are selected, and the electrostatic characteristic parameters for air quality assessment are obtained through linear combination.

6. The method according to claim 1, characterized in that The mapping relationship model between electrostatic characteristic parameters and air quality is trained by a support vector machine algorithm according to electrostatic characteristic parameters under different air quality conditions, including pollutant concentration and particulate matter content, including: Obtain electrostatic characteristic parameters and corresponding pollutant concentration and particulate matter content data under different air quality conditions to construct a training data set; Performing Z-score standardization on the training data set, screening the electrostatic feature parameters after standardization using a recursive feature elimination method, and determining the optimal feature combination; According to the optimal feature combination, a grid search method is used to optimize the kernel function type, regularization parameter C and kernel function parameter γ of the support vector machine to obtain an optimal parameter combination; Using the optimal feature combination and the optimal parameter combination, a support vector machine algorithm is used to train a mapping relationship model between electrostatic feature parameters and air quality; For different air quality conditions, corresponding sub-models are trained respectively, and the sub-models are trained based on the mapping relationship model to improve the adaptability of the sub-models under various pollution levels.

7. The method according to claim 1, characterized in that The real-time collected sofa electrostatic signal strength is input into the trained mapping relationship model between electrostatic features and air quality to obtain the change of electrostatic features and evaluate the indoor air quality in real time, including: Acquire an electrostatic signal sequence of the sofa surface collected by the electrostatic sensor array, wherein the electrostatic signal sequence includes preset data points; Performing data preprocessing on the electrostatic signal sequence, removing burst noise using a median filter algorithm, and performing signal denoising by wavelet transform; Extract electrostatic characteristic parameters from the electrostatic signal after signal denoising, calculate time domain characteristics and frequency domain characteristics, and obtain a multi-dimensional feature vector; If the multidimensional feature vector has not been standardized, the multidimensional feature vector is standardized so that all eigenvalues ​​are scaled to the interval [-1, 1]; Inputting the standardized multidimensional electrostatic feature vector into a pre-trained support vector machine model to predict the concentrations of PM2.5, PM10, ozone and sulfur dioxide in the current indoor air; The single air quality index is calculated according to the predicted pollutant concentration, the maximum value is taken as the comprehensive air quality index, the air quality grade interval to which the comprehensive air quality index belongs is determined, and the current indoor air quality grade is determined.

8. The method according to claim 1, characterized in that If the evaluation result shows that the air quality is less than the air quality threshold, the air purification device of the smart sofa is triggered to turn on the air purifier and increase the ventilation volume of the fresh air system to improve the indoor air environment, including: Obtaining an air quality index measured in real time by an air quality sensor, and determining whether the air quality exceeds a preset air quality threshold according to the air quality index; If the air quality index exceeds a preset air quality threshold, a start instruction is sent to the air purification device through the microcontroller; The air quality index and indoor temperature and humidity data are processed by a fuzzy control algorithm to obtain control instructions for the air purifier working mode and the ventilation volume of the fresh air system; Adjust the wind speed of the air purifier and the ventilation volume of the fresh air system according to the control instruction; The PID control algorithm is used to calculate the deviation between the air quality index and the target value, and the adjustment amount of the purification measure intensity is obtained; Modify the wind speed of the air purifier and the ventilation volume of the fresh air system according to the adjustment amount; The following formula is used to calculate the adjusted air purifier wind speed:

9. Where V fan Indicates the wind speed of the air purifier, V base represents the basic wind speed, and ΔV represents the adjustment of the wind speed; The modified wind speed and ventilation volume are sent to the air purification device, and the air purification device performs corresponding purification measures according to the received wind speed and ventilation volume; This formula is used to calculate the adjusted air purifier wind speed.

Images