An intelligent control method and related device for the working state of an aromatherapy machine

Abstract

The present invention relates to the field of data processing technology, and discloses an intelligent control method and related device for the working state of an aromatherapy machine, the method comprising: preprocessing real-time environmental data and real-time user status data; performing scenario analysis based on the real-time environmental data based on linear relationship analysis to obtain real-time scenario data; extracting features from real-time user facial images based on confidence analysis, and extracting features from real-time user voices; identifying the user's current emotion based on facial expression features and voice emotion features combined with emotion fluctuation analysis; performing user preference analysis based on the current emotion recognition result to obtain user preference data corresponding to the current emotion recognition result; and switching control of the working state of the aromatherapy machine based on real-time scenario data, current emotion recognition results, and user preference data. The present invention coordinates and matches the working state that the aromatherapy machine needs to switch from multiple dimensions, so that the control of the working state of the aromatherapy machine can achieve a more ideal effect.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an intelligent control method and related device for the working state of an aromatherapy machine. Background Art

[0002] An aromatherapy machine is an electrical appliance that improves the air environment by evaporating essential oils. Through the volatile properties of spices or essential oils, spices or essential oils are released into the indoor air to provide a comfortable and pleasant smell environment, which can help people relax and regulate their emotions. With the development of information technology, today's aromatherapy machines are able to achieve autonomous control of their working state. However, in the current control of the working state of aromatherapy machines, the switching control of the working state is usually only carried out through unilateral data analysis, and the switching of the working state is not carried out from multiple data analysis. At the same time, the current relevant data analysis does not involve the analysis of application scenarios, sentiment analysis, and preference analysis. As a result, it is impossible to obtain sufficiently comprehensive and accurate data on the control of the working state of the aromatherapy machine, which makes the working state switched and controlled by the aromatherapy machine unsatisfactory and fails to bring users a high-quality user experience. Summary of the invention

[0003] The purpose of the present invention is to overcome the shortcomings of the prior art. The present invention provides an intelligent control method and related devices for the working state of an aromatherapy machine, which coordinately match the working state that the aromatherapy machine needs to switch from multiple dimensions, so that the control of the working state of the aromatherapy machine can achieve a more ideal effect.

[0004] In order to solve the above technical problems, the present invention provides an intelligent control method for the working state of an aromatherapy machine, the method comprising:

[0005] Acquire real-time environment data and real-time user status data, and pre-process the real-time environment data and real-time user status data to obtain pre-processed real-time environment data and real-time user status data;

[0006] Based on linear relationship analysis, pre-processed real-time environmental data is used to perform scenario analysis to obtain real-time scenario data;

[0007] Extracting a real-time user facial image and a real-time user voice from the preprocessed real-time user status data, performing feature extraction on the real-time user facial image based on confidence analysis to obtain facial expression features, and performing feature extraction on the real-time user voice to obtain voice emotion features;

[0008] Based on the facial expression features and voice emotion features combined with emotion fluctuation analysis, the user's current emotion recognition is performed to obtain a current emotion recognition result;

[0009] Performing user preference analysis based on the current emotion recognition result to obtain user preference data corresponding to the current emotion recognition result;

[0010] The aromatherapy machine's working state is switched and controlled based on the real-time scenario data, the current emotion recognition result and the user preference data.

[0011] Optionally, the performing scenario analysis based on the linear relationship analysis using the preprocessed real-time environmental data to obtain the real-time scenario data includes:

[0012] Obtain historical environmental data, perform linear relationship fitting between the historical environmental data and each application scenario, and obtain the target linear relationship;

[0013] Performing aggregation processing based on the historical environmental data to obtain an environmental data aggregation interval;

[0014] Performing distribution interval analysis based on the historical environmental data to obtain distribution interval analysis data, and generating scenario division data based on the environmental data aggregation interval and the distribution interval analysis data;

[0015] Performing cluster analysis based on the historical environmental data and the target linear relationship to obtain cluster analysis results;

[0016] A scenario division model is generated based on the cluster analysis result and the scenario division data, and scenario analysis is performed based on the scenario division model using the preprocessed real-time environmental data to obtain real-time scenario data.

[0017] Optionally, the extracting features of the real-time user facial image based on confidence analysis to obtain facial expression features includes:

[0018] Performing regional object division on the real-time user facial image to obtain the real-time user facial image after regional object division;

[0019] Based on wavelet transform, the real-time user facial image after the regional object division is subjected to amplitude spectrum feature extraction to obtain target amplitude spectrum feature, and the target amplitude spectrum feature is subjected to histogram extraction to obtain a target histogram;

[0020] Extracting a region of interest from the real-time user facial image after the region object division, obtaining a channel pixel grayscale value of the region of interest, and determining facial blood flow distribution characteristics based on the channel pixel grayscale value;

[0021] Performing confidence analysis on the target histogram and facial blood flow distribution features based on an adaptive feature learning mechanism to obtain corresponding self-attention confidence;

[0022] The target histogram and facial blood flow distribution features are fused with the corresponding self-attention confidence to obtain the facial expression features.

[0023] Optionally, extracting features from the real-time user voice to obtain voice emotion features includes:

[0024] Extracting tone features of the real-time user voice to obtain target tone features;

[0025] Performing spectrogram extraction on the real-time user speech to obtain a target spectrogram, and removing silence segments from the target spectrogram to obtain a target spectrogram after the silence segments are removed;

[0026] Based on the target spectrogram after the silence segment is removed, a harmonic intensity trajectory is calculated using linear prediction coding to obtain a target harmonic intensity trajectory, and based on the target harmonic intensity trajectory, a target amplitude spectrum is generated using Fourier transform;

[0027] Based on the wavelet packet basis function, the target spectrogram after the silence segment is removed is decomposed by wavelet packet to obtain the corresponding wavelet packet coefficients;

[0028] A speech emotion feature is generated based on the target tone feature, the target amplitude spectrum and the wavelet packet coefficients.

[0029] Optionally, the current emotion recognition of the user based on the facial expression features and the voice emotion features combined with the emotion fluctuation analysis to obtain the current emotion recognition result includes:

[0030] Performing dimensionality reduction and weighting processing on the facial expression features and the speech emotion features to obtain target facial expression features and target speech emotion features;

[0031] Matching the emotion fluctuation value in the emotion database based on the target facial expression feature and the target voice emotion feature to obtain the target emotion fluctuation value;

[0032] Generate an emotion fluctuation curve using the target emotion fluctuation value based on the emotion correlation function and the emotion recognition network;

[0033] Based on the target facial expression features and the target voice emotion features, Naive Bayes is used to perform emotion classification to obtain a corresponding conditional probability set;

[0034] The user's current emotion recognition is performed based on the emotion fluctuation curve and the conditional probability set to obtain a current emotion recognition result.

[0035] Optionally, performing user preference analysis based on the current emotion recognition result to obtain user preference data corresponding to the current emotion recognition result includes:

[0036] Acquire historical user emotion data and historical user preference data of a preset time series, and perform preference feature evolution analysis on the historical user preference data based on dynamic factors and evolution factors to obtain preference feature evolution data;

[0037] Establishing a mapping relationship between the historical user sentiment data and preference feature evolution data;

[0038] Based on the current emotion recognition result, the mapping relationship is used to perform an aggregate analysis of the user's personalized preferences to obtain user preference data corresponding to the current emotion recognition result.

[0039] Optionally, the switching control of the aromatherapy machine working state based on the real-time scenario data, the current emotion recognition result and the user preference data includes:

[0040] Based on the real-time scenario data, the current emotion recognition result and the user preference data matching the corresponding target working state of the aroma diffuser, the aroma diffuser switches to the target working state to perform related operations.

[0041] Optionally, the target working state of the aroma diffuser corresponding to the real-time scenario data, the current emotion recognition result and the user preference data is matched, and the aroma diffuser is switched to the target working state to perform related operations, specifically including:

[0042] Combine real-time scenario data and user preference data to determine the working mode of the aromatherapy machine;

[0043] The working gear of the aromatherapy machine working mode is further determined based on the emotion recognition result. If the aromatherapy machine cannot obtain the emotion recognition result, the emotion recognition result is assigned to a default value, and the minimum gear in the current working mode is determined as the working gear of the aromatherapy machine in the working mode;

[0044] Control the aroma diffuser to switch to a determined working mode and a determined working gear to work, and estimate a user experience index of the aroma diffuser after the aroma diffuser works for a preset time;

[0045] The estimated user experience index is compared with the preset target value. If the user experience index is not less than the target value, the aromatherapy machine is kept in the current working state. Otherwise, the working gear of the aromatherapy machine is adjusted so that the user experience index of the aromatherapy machine when working in the latest working gear is not less than the preset target value.

[0046] In addition, the present invention also provides an intelligent control device for the working state of an aromatherapy machine, the device comprising:

[0047] A data preprocessing module, used to obtain real-time environment data and real-time user status data, and preprocess the real-time environment data and real-time user status data to obtain preprocessed real-time environment data and real-time user status data;

[0048] A scenario analysis module, used to perform scenario analysis using pre-processed real-time environmental data based on linear relationship analysis to obtain real-time scenario data;

[0049] A feature extraction module is used to extract a real-time user facial image and a real-time user voice from the preprocessed real-time user status data, perform feature extraction on the real-time user facial image based on confidence analysis to obtain facial expression features, and perform feature extraction on the real-time user voice to obtain voice emotion features;

[0050] An emotion recognition module is used to recognize the user's current emotion based on the facial expression features and voice emotion features combined with emotion fluctuation analysis to obtain a current emotion recognition result;

[0051] A preference analysis module, used to perform user preference analysis based on the current emotion recognition result to obtain user preference data corresponding to the current emotion recognition result;

[0052] The aroma diffuser working state control module is used to switch the working state of the aroma diffuser based on the real-time scenario data, the current emotion recognition result and the user preference data.

[0053] In addition, the present invention also provides an electronic device, which includes a processor and a memory, wherein the memory is used to store instructions, and the processor is used to call the instructions in the memory, so that the electronic device executes the above-mentioned intelligent control method for the working state of the aromatherapy machine.

[0054] In addition, the present invention also provides a computer-readable storage medium, which stores computer instructions. When the computer instructions are executed on an electronic device, the electronic device executes the above-mentioned intelligent control method for the working state of the aromatherapy machine.

[0055] In an embodiment of the present invention, scenario analysis is performed based on linear relationship analysis using pre-processed real-time environmental data, so that the scenario in which the user is currently located can be analyzed more accurately. Feature extraction is performed on the real-time user facial image based on confidence analysis, and speech emotion features are generated based on the target tone features, target amplitude spectrum and wavelet packet coefficients generated by the real-time user voice, so that more comprehensive and accurate facial expression features and speech emotion features can be obtained. The current emotion of the user is identified based on facial expression features and speech emotion features combined with emotion fluctuation analysis to avoid excessive deviation between the identified user's current emotion and the actual situation. User preference analysis is performed based on the current emotion recognition result, and the preference data under the current emotion can be determined for the user's current emotion. The switching control of the working state of the aromatherapy machine is performed based on the real-time scenario data, the current emotion recognition result and the user preference data, so as to obtain sufficiently comprehensive and accurate data on the control of the working state of the aromatherapy machine, and coordinately match the working state required to be switched by the aromatherapy machine from multiple dimensions, so that the control of the working state of the aromatherapy machine can achieve a more ideal effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0057] Figure 1 is a flow chart of an intelligent control method for the working state of an aromatherapy machine in an embodiment of the present invention;

[0058] Figure 2 It is a schematic diagram of the structure of the intelligent control device for the working state of the aromatherapy machine in the embodiment of the present invention;

[0059] Figure 3 It is a schematic diagram of the structure of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0061] Embodiment 1:

[0062] See also Figure 1 , Figure 1: is a flow chart of an intelligent control method for the working state of an aromatherapy machine in an embodiment of the present invention, the method comprising:

[0063] S11: Acquire real-time environment data and real-time user status data, and pre-process the real-time environment data and real-time user status data to obtain pre-processed real-time environment data and real-time user status data.

[0064] During the specific implementation of the present invention, real-time environmental data and real-time user status data are obtained through the built-in smart sensor combination of the aromatherapy machine. The built-in smart sensor combination of the aromatherapy machine continuously collects real-time data according to the set duration. The smart sensor combination includes temperature sensors, humidity sensors, cameras and multi-wavelength cameras, etc. The real-time environmental data includes odor data, temperature data and humidity data, etc. The real-time user status data includes real-time video, real-time facial image, real-time voice and real-time location of the user, etc.

[0065] The real-time environment data and the real-time user status data are preprocessed, data cleaning processing and data integration processing are performed on the real-time environment data, noise reduction processing and data enhancement processing are performed on the real-time user status data, and the preprocessed real-time environment data and real-time user status data are obtained.

[0066] S12: Perform scenario analysis using the preprocessed real-time environmental data based on linear relationship analysis to obtain real-time scenario data.

[0067] In the specific implementation process of the present invention, the linear relationship analysis is used to perform scenario analysis using preprocessed real-time environmental data to obtain real-time scenario data, including: obtaining historical environmental data, fitting the linear relationship between the historical environmental data and each application scenario, and obtaining the target linear relationship; performing aggregation processing based on the historical environmental data to obtain the environmental data aggregation interval; performing distribution interval analysis based on the historical environmental data to obtain distribution interval analysis data, and generating scenario division data based on the environmental data aggregation interval and the distribution interval analysis data; performing cluster analysis based on the historical environmental data and the target linear relationship to obtain cluster analysis results; generating a scenario division model based on the cluster analysis results and the scenario division data, and performing scenario analysis using the preprocessed real-time environmental data based on the scenario division model to obtain real-time scenario data.

[0068] Specifically, historical environmental data is obtained, which is the environmental data collected by the aromatherapy machine in the past. A linear relationship is fitted between the historical environmental data and each application scenario. The application scenarios include cooking scenarios and exercise scenarios. The linear relationship between the historical environmental data and each environmental data interval corresponding to each application scenario is fitted through the least squares method to obtain the target linear relationship.

[0069] Aggregation processing is performed based on the historical environmental data, with one type of data in the historical environmental data as an independent variable, such as temperature data in the historical environmental data as an independent variable, and another type of data in the historical environmental data as a dependent variable, and the corresponding data aggregation interval is calculated, such as the humidity aggregation interval of ambient humidity at different ambient temperatures, so as to obtain the environmental data aggregation interval.

[0070] Based on the historical environmental data, distribution interval analysis is performed, distribution span and distribution uniformity analysis is performed on the historical environmental data to obtain target distribution span and target distribution uniformity, distribution interval analysis data is generated according to the target distribution span and target distribution uniformity, and scenario division data is generated based on the environmental data aggregation interval and distribution interval analysis data, and scenario division data is generated using the target linear relationship based on the environmental data aggregation interval and distribution interval analysis data.

[0071] A cluster analysis is performed based on the historical environmental data and the target linear relationship, the number of clusters x is determined according to the target linear relationship, x initial cluster centers are selected, the distance from each data point in the historical environmental data to the initial cluster center is calculated, and the data points are divided into classes with the shortest distance to the initial cluster center, new cluster centers are generated, and it is determined whether the new cluster centers meet preset clustering constraints, the preset clustering constraints include that the new cluster centers are the same as the cluster centers of the previous generation or the number of iterations reaches a preset threshold, if the preset clustering constraints are met, the clustering analysis results are output, if the preset clustering constraints are not met, the calculation of data point distances and the generation of new cluster centers are continued until the preset clustering constraints are met, and the clustering analysis results are obtained.

[0072] A scenario division model is generated based on the cluster analysis results and the scenario division data, and the scenario division model is trained based on the cluster analysis results and the scenario division data. The scenario division model adopts a deep neural network model, and based on the scenario division model, scenario analysis is performed using preprocessed real-time environmental data to obtain the user's current scenario, such as cooking or exercising, that is, to obtain real-time scenario data. The working state of the aromatherapy machine is matched by the real-time scenario data, so that the matched working state of the aromatherapy machine can be more in line with the user's actual needs.

[0073] S13: extracting a real-time user facial image and a real-time user voice from the preprocessed real-time user status data, performing feature extraction on the real-time user facial image based on confidence analysis to obtain facial expression features, and performing feature extraction on the real-time user voice to obtain voice emotion features.

[0074] In the specific implementation process of the present invention, the feature extraction of the real-time user facial image based on confidence analysis to obtain facial expression features includes:

[0075] Performing regional object division on the real-time user facial image to obtain the real-time user facial image after regional object division;

[0076] Based on wavelet transform, the real-time user facial image after the regional object division is subjected to amplitude spectrum feature extraction to obtain target amplitude spectrum feature, and the target amplitude spectrum feature is subjected to histogram extraction to obtain a target histogram;

[0077] Extracting a region of interest from the real-time user facial image after the region object division, obtaining a channel pixel grayscale value of the region of interest, and determining facial blood flow distribution characteristics based on the channel pixel grayscale value;

[0078] Based on the adaptive feature learning mechanism, the target histogram and facial blood flow distribution features are analyzed for confidence to obtain the corresponding self-attention confidence; the target histogram and facial blood flow distribution features are fused with the corresponding self-attention confidence to obtain the facial expression features.

[0079] Furthermore, the extracting features of the real-time user voice to obtain voice emotion features includes:

[0080] Extracting tone features of the real-time user voice to obtain target tone features;

[0081] Performing spectrogram extraction on the real-time user speech to obtain a target spectrogram, and removing silence segments from the target spectrogram to obtain a target spectrogram after the silence segments are removed;

[0082] Based on the target spectrogram after the silence segment is removed, a harmonic intensity trajectory is calculated using linear prediction coding to obtain a target harmonic intensity trajectory, and based on the target harmonic intensity trajectory, a target amplitude spectrum is generated using Fourier transform;

[0083] Based on the wavelet packet basis function, the target spectrogram after the silence segment is removed is decomposed by wavelet packet to obtain the corresponding wavelet packet coefficients;

[0084] A speech emotion feature is generated based on the target tone feature, the target amplitude spectrum and the wavelet packet coefficients.

[0085] Specifically, the real-time user facial image and the real-time user voice are extracted from the preprocessed real-time user status data. The real-time user facial image can be obtained by extracting the user facial image from the preprocessed real-time user status data, and the real-time user voice can be obtained by extracting the real-time user voice from the preprocessed real-time user status data.

[0086] The real-time user facial image is divided into regional objects, the four-corner coordinates of each region of the eyes, cheeks, nose and mouth in the real-time user facial image are obtained, the four-corner coordinates of each region are converted to obtain rectangular division information of the corresponding region, and the real-time user facial image is divided into regional objects according to the rectangular division information, that is, the three regional objects of the user's nose, cheeks, eyes and mouth are segmented to obtain the real-time user facial image after the regional object division.

[0087] Based on wavelet transform, amplitude spectrum feature extraction is performed on the real-time user facial image after regional object division, and two-dimensional wavelet transform based on window Fourier transform is used to perform wavelet transform on the real-time user facial image after regional object division to obtain the amplitude spectrum based on window Fourier transform in each direction and scale in the real-time user facial image after regional object division, that is, to obtain the target amplitude spectrum feature, and histogram extraction is performed on the target amplitude spectrum feature, and a local pattern binary operator calculation of a uniform mode is performed on the target amplitude spectrum feature, and a histogram is generated according to the calculated local pattern binary operator, and the histograms are cascaded to obtain the target histogram.

[0088] A region of interest is extracted from a real-time user facial image after regional object division, and regional images of both sides of the nose and cheeks are selected as the region of interest in the real-time user facial image after regional object division, and the channel pixel grayscale value of the region of interest is obtained, and the average pixel grayscale value of the R channel, the G channel and the B channel of the region of interest is calculated by remote photoplethysmography, that is, the channel pixel grayscale value, and the facial blood flow distribution characteristics are determined based on the channel pixel grayscale values, and the average pixel grayscale values ​​of the R channel, the G channel and the B channel are detrended and independently component analyzed to obtain R channel independent components, G channel independent components and B channel independent components, and the independent component with the largest correlation with the average pixel grayscale value of the G channel is used as the pulse wave feature, and the frequency spectrum of the pulse wave feature is extracted, and the facial blood flow distribution characteristics are obtained by using the frequency spectrum of the pulse wave feature through a blood flow distribution detector.

[0089] Based on the adaptive feature learning mechanism, the target histogram and facial blood flow distribution features are analyzed for confidence, and the self-attention confidence of the target histogram and facial blood flow distribution features is calculated by the self-attention algorithm in the adaptive feature learning mechanism to obtain the self-attention confidence corresponding to the target histogram and the self-attention confidence corresponding to the facial blood flow distribution features, that is, to obtain the corresponding self-attention confidence.

[0090] The target histogram and facial blood flow distribution features are fused with the corresponding self-attention confidences, and the target histogram and facial blood flow distribution features are fused with their respective corresponding self-attention confidences through the softmax function to obtain the facial expression features.

[0091] Tone features are extracted from the real-time user speech, and the tone features are extracted from the real-time user speech by a weighted finite state transcriber to obtain target tone features.

[0092] The spectrogram is extracted from the real-time user speech, and the real-time user speech is short-term segmented, Hamming windowed and Fourier transformed to obtain a target spectrogram.

[0093] The target spectrogram is subjected to silent segment removal, the spectral energy in the target spectrogram is calculated, the spectral energy is compared with a preset threshold, and the silent segment is removed according to the comparison result to obtain the target spectrogram after the silent segment is removed.

[0094] The harmonic intensity trajectory is calculated based on the target spectrogram after the silence segment is removed by using linear prediction coding, the target spectrogram after the silence segment is removed is thinned to obtain a sparse target spectrogram, the power spectrum response curve is determined by using the sparse target spectrogram through linear prediction coding, the peak points of the spectrogram are determined by using the power spectrum response curve, a peak trajectory is generated along the time domain direction according to the peak points of the spectrogram, a differential operation is performed on the peak trajectory to obtain frequency difference values, the frequency difference values ​​are grouped according to similar values, the frequency difference value with the largest number in the group is used as the frequency difference between harmonics by a voting algorithm, the target harmonic intensity trajectory is generated according to the frequency difference between harmonics by using the neighborhood mean, and the target harmonic intensity trajectory is generated according to the target harmonic intensity trajectory by using Fourier transform, the calculation points of Fourier transform are determined, the phase of the target harmonic intensity trajectory is removed according to the calculation points of Fourier transform, and the target amplitude spectrum is obtained.

[0095] Based on the wavelet packet basis function, the target spectrogram after the silence segment is removed is subjected to wavelet packet decomposition to obtain speech segments of known emotion categories; the speech segments are used as the target data set, the target data set is preprocessed to obtain valid speech frames, and a full binary tree structure is constructed as the wavelet packet decomposition tree; the valid speech frames are subjected to wavelet packet decomposition according to the wavelet packet decomposition tree to obtain valid speech frames after wavelet packet decomposition, the normalized energy of the valid speech frames after wavelet packet decomposition in the wavelet packet decomposition tree nodes is calculated, and the Hellinger distance of the wavelet packet decomposition tree nodes is calculated according to the normalized energy; an optimized wavelet packet decomposition tree is constructed according to the Hellinger distance, and a wavelet packet basis function is generated according to the optimized wavelet packet decomposition tree. The target spectrogram is subjected to wavelet packet decomposition through the wavelet packet basis function, which can effectively improve the robustness of speech emotion recognition and obtain the corresponding wavelet packet coefficients.

[0096] The speech emotion feature is generated based on the target tone feature, the target amplitude spectrum and the wavelet packet coefficient, so that the obtained speech emotion feature is more comprehensive and accurate.

[0097] S14: Based on the facial expression features and voice emotion features combined with emotion fluctuation analysis, the user's current emotion recognition is performed to obtain a current emotion recognition result.

[0098] In the specific implementation process of the present invention, the current emotion recognition of the user based on the facial expression features and the voice emotion features combined with the emotion fluctuation analysis is performed to obtain the current emotion recognition result, including:

[0099] Performing dimensionality reduction and weighting processing on the facial expression features and the speech emotion features to obtain target facial expression features and target speech emotion features;

[0100] Matching the emotion fluctuation value in the emotion database based on the target facial expression feature and the target voice emotion feature to obtain the target emotion fluctuation value;

[0101] Generate an emotion fluctuation curve using the target emotion fluctuation value based on the emotion correlation function and the emotion recognition network;

[0102] Based on the target facial expression features and the target voice emotion features, Naive Bayes is used to perform emotion classification to obtain a corresponding conditional probability set;

[0103] The user's current emotion recognition is performed based on the emotion fluctuation curve and the conditional probability set to obtain a current emotion recognition result.

[0104] Specifically, the facial expression features and speech emotion features are subjected to dimensionality reduction and weighting processing, the facial expression features and speech emotion features are subjected to dimensionality reduction by principal component analysis, the facial expression features and speech emotion features are weighted by speech weights and expression weights, and target facial expression features and target speech emotion features are obtained.

[0105] Based on the target facial expression features and the target voice emotion features, the emotion fluctuation values ​​are matched in the emotion database, that is, similarity matching is performed in the facial expression database according to the target facial expression features to obtain facial micro-expression data, and similarity matching is performed in the voice emotion database according to the target voice emotion features and the facial micro-expression data to obtain the emotion fluctuation values ​​corresponding to the facial micro-expression data, that is, the target emotion fluctuation values ​​are obtained.

[0106] Based on the emotion correlation function and the emotion recognition network, the emotion fluctuation curve is generated using the target emotion fluctuation value. According to the emotion correlation function, the emotion recognition network is generated using the target emotion fluctuation value in combination with a preset emotion recognition circuit. According to the emotion recognition network, the final emotion fluctuation analysis is performed using the target emotion fluctuation value to obtain the emotion fluctuation curve.

[0107] Based on the target facial expression features and the target voice emotion features, Naive Bayes is used to perform emotion classification, the target facial expression features are subjected to Naive Bayes emotion speech classification to obtain a first conditional probability set, the target voice emotion features are subjected to Naive Bayes facial emotion classification to obtain a second conditional probability combination, that is, a corresponding conditional probability set is obtained.

[0108] The emotion fluctuation curve and conditional probability set are used to identify the user's current emotion, the first conditional probability set and the second conditional probability set are fused to obtain a target probability set, the target probability set and the emotion fluctuation curve are input into a multi-layer perceptron for emotion recognition, and the current emotion recognition result is obtained. The current emotion recognition result may include happiness, anger, sadness, fear, etc., thereby improving the accuracy of emotion recognition and improving the comprehensiveness of emotion recognition categories.

[0109] S15: Performing user preference analysis based on the current emotion recognition result to obtain user preference data corresponding to the current emotion recognition result.

[0110] In a specific implementation of the present invention, performing user preference analysis based on the current emotion recognition result to obtain user preference data corresponding to the current emotion recognition result includes:

[0111] Acquire historical user emotion data and historical user preference data of a preset time series, and perform preference feature evolution analysis on the historical user preference data based on dynamic factors and evolution factors to obtain preference feature evolution data;

[0112] Establishing a mapping relationship between the historical user sentiment data and preference feature evolution data;

[0113] Based on the current emotion recognition result, the mapping relationship is used to perform an aggregate analysis of the user's personalized preferences to obtain user preference data corresponding to the current emotion recognition result.

[0114] Specifically, historical user emotion data and historical user preference data of a preset time series are obtained, the historical user preference data including the aromatherapy machine working modes preferred by users in different emotion types in different periods, and preference feature evolution analysis is performed on the historical user preference data based on dynamic factors and evolution factors, the dynamic factors including the aromatherapy machine working modes that users are interested in in different periods, and the evolution factors including the relationship between the user's preference for the aromatherapy machine working mode and changes over time, a user preference matrix is ​​generated according to the dynamic factors and evolution factors, and preference feature evolution trend analysis is performed using the historical user preference data according to the user preference matrix to obtain the preference feature evolution trend, that is, to obtain the preference feature evolution data, so as to know the user's preference evolution and improve the flexibility of user preference analysis.

[0115] A mapping relationship between the historical user emotion data and the preference feature evolution data is established, and the mapping relationship between the historical user emotion data and the preference feature evolution data is mined according to the meta-stretching network parameters in the meta-mapping network to obtain the corresponding mapping relationship, and the meta-stretching network parameters include a stretching function and an offset function.

[0116] Based on the current emotion recognition result, the mapping relationship is used to perform an aggregate analysis of the user's personalized preferences to obtain user preference data corresponding to the current emotion recognition result. The user's preferred types are different under different emotions, so the preference data corresponding to the user's current emotion type can be known, so that the obtained user preference data is more in line with the user's current emotional state.

[0117] S16: Switching control of the aromatherapy machine's working state is performed based on the real-time scenario data, the current emotion recognition result and the user preference data.

[0118] In the specific implementation process of the present invention, the switching control of the aromatherapy machine's working state based on the real-time scenario data, the current emotion recognition result and the user preference data includes: matching the corresponding target working state of the aromatherapy machine based on the real-time scenario data, the current emotion recognition result and the user preference data, and the aromatherapy machine switches to the target working state to perform related operations.

[0119] In a preferred embodiment, step S16 specifically includes steps S161 to S163, or includes steps S161 to S165.

[0120] Step S161: Determine the working mode of the aromatherapy machine in combination with the real-time scenario data and the user preference data.

[0121] It is understandable that the working state of the aromatherapy machine includes a plurality of preset different working modes, each working mode has a plurality of different working gears, and the aromatherapy smell, aromatherapy concentration and atomization power outputted by the aromatherapy machine at each working gear are not completely consistent. Therefore, when the aromatherapy machine works at different working gears in different working modes, the fragrance it produces for the user's environment is not completely the same, so that the user can obtain a different fragrance experience.

[0122] Step S161 is to first normalize the real-time scenario data Q and the user preference data P, and then multiply them by the preset first weight coefficient a1 and the second weight coefficient a2 to obtain a calculation result M (i.e., M=Q* a1+P* a2, "*" represents multiplication), and determine the working mode of the aromatherapy machine based on the calculation result M by looking up a table, wherein the table presets the digital range of M corresponding to each working mode.

[0123] Among them, a1 and a2 are both between 0 and 1, determined by empirical values ​​or through deep learning of a neural network model.

[0124] It can be understood that the first weight coefficient a1 and the second weight coefficient a2 corresponding to the real-time scenario data Q in different numerical ranges and / or the user preference data P in different numerical ranges can be predetermined respectively, so as to ensure that the aromatherapy machine can work in the most matching working module in the corresponding usage environment and enable the working mode to take into account the user preferences to the greatest extent.

[0125] Step S162: On the basis of determining the working mode of the aromatherapy machine in step S161, further determine the working gear of the working mode of the aromatherapy machine based on the current emotion recognition result T.

[0126] In this step, by pre-establishing the association between each working gear in different working modes and the emotion recognition results, the latest emotion recognition results can be used to reversely determine the working gear of the aromatherapy machine.

[0127] If the aromatherapy machine cannot obtain the emotion recognition result due to reasons such as the user temporarily leaving the usage environment of the aromatherapy machine or the user is in sleep mode, the emotion recognition result is assigned to the default value T0. At this time, the minimum gear in the current working mode is determined as the working gear of the aromatherapy machine in this working mode.

[0128] Step S163: Control the fumigator to switch to the working mode determined in step S161 and the working gear determined in step S162.

[0129] Some preferred embodiments further include step S164 and step S165, which further improve the working state of the aromatherapy machine by coordinated matching in multiple dimensions to perform switching control so as to achieve the best fragrance effect matching the user.

[0130] Step S164: estimating the user experience index of the aromatherapy machine after the aromatherapy machine has been operating for a preset time period.

[0131] The user experience index may be estimated by any existing estimation method, for example, directly using the difference between the latest emotion recognition result T and the default value T0 of the emotion recognition result as the user experience index.

[0132] Step S165: Compare the estimated user experience index with the preset target value. If the user experience index is not less than the target value, keep the aromatherapy machine in the current working state. Otherwise, when the user experience index is less than the target value, adjust the working gear of the aromatherapy machine to improve the user experience index, so that the user experience index of the aromatherapy machine when working in the latest working gear is not less than the preset target value.

[0133] The present invention matches the target working state of the corresponding aromatherapy machine based on the real-time scenario data, the current emotion recognition result and the user preference data: the corresponding working state is matched in the database according to the real-time scenario data, the current emotion recognition result and the user preference data, and the working state of the aromatherapy machine can be collaboratively matched from multiple dimensions, so that the switched working state is more in line with the user's needs, avoiding the one-sidedness and limitations brought about by matching the working state in a single dimension. At the same time, the working state of the aromatherapy machine matched in a single dimension cannot well fit the user's required state, thereby improving the intelligence level of the aromatherapy machine's working state control, and the aromatherapy machine switches to the target working state to perform related operations, that is, the aromatherapy machine switches to the target working state to perform aromatherapy spray operations.

[0134] In an embodiment of the present invention, scenario analysis is performed based on linear relationship analysis using pre-processed real-time environmental data, so that the scenario in which the user is currently located can be analyzed more accurately. Feature extraction is performed on the real-time user facial image based on confidence analysis, and speech emotion features are generated based on the target tone features, target amplitude spectrum and wavelet packet coefficients generated by the real-time user voice, so that more comprehensive and accurate facial expression features and speech emotion features can be obtained. The current emotion of the user is identified based on facial expression features and speech emotion features combined with emotion fluctuation analysis to avoid excessive deviation between the identified user's current emotion and the actual situation. User preference analysis is performed based on the current emotion recognition result, and the preference data under the current emotion can be determined for the user's current emotion. The switching control of the working state of the aromatherapy machine is performed based on the real-time scenario data, the current emotion recognition result and the user preference data, so as to obtain sufficiently comprehensive and accurate data on the control of the working state of the aromatherapy machine, and coordinately match the working state required to be switched by the aromatherapy machine from multiple dimensions, so that the control of the working state of the aromatherapy machine can achieve a more ideal effect.

[0135] Embodiment 2:

[0136] See also Figure 2 , Figure 2 : is a schematic diagram of the structure of an intelligent control device for the working state of an aromatherapy machine in an embodiment of the present invention, the device comprises:

[0137] Data preprocessing module 21: used to obtain real-time environment data and real-time user status data, and preprocess the real-time environment data and real-time user status data to obtain preprocessed real-time environment data and real-time user status data;

[0138] Scenario analysis module 22: used to perform scenario analysis based on linear relationship analysis using pre-processed real-time environmental data to obtain real-time scenario data;

[0139] Feature extraction module 23: used to extract real-time user facial images and real-time user voices from the pre-processed real-time user status data, perform feature extraction on the real-time user facial images based on confidence analysis to obtain facial expression features, and perform feature extraction on the real-time user voices to obtain voice emotion features;

[0140] Emotion recognition module 24: used to recognize the current emotion of the user based on the facial expression features and voice emotion features combined with emotion fluctuation analysis to obtain a current emotion recognition result;

[0141] Preference analysis module 25: used to perform user preference analysis based on the current emotion recognition result to obtain user preference data corresponding to the current emotion recognition result;

[0142] Aroma diffuser working state control module 26: used for switching the working state of the aroma diffuser based on the real-time scenario data, the current emotion recognition result and the user preference data.

[0143] In the specific implementation process of the present invention, the specific implementation method of the device item can refer to the implementation method of the above-mentioned method item, which will not be repeated here.

[0144] In an embodiment of the present invention, scenario analysis is performed based on linear relationship analysis using pre-processed real-time environmental data, so that the scenario in which the user is currently located can be analyzed more accurately. Feature extraction is performed on the real-time user facial image based on confidence analysis, and speech emotion features are generated based on the target tone features, target amplitude spectrum and wavelet packet coefficients generated by the real-time user voice, so that more comprehensive and accurate facial expression features and speech emotion features can be obtained. The current emotion of the user is identified based on facial expression features and speech emotion features combined with emotion fluctuation analysis to avoid excessive deviation between the identified user's current emotion and the actual situation. User preference analysis is performed based on the current emotion recognition result, and the preference data under the current emotion can be determined for the user's current emotion. The switching control of the working state of the aromatherapy machine is performed based on the real-time scenario data, the current emotion recognition result and the user preference data, so as to obtain sufficiently comprehensive and accurate data on the control of the working state of the aromatherapy machine, and match the working state that the aromatherapy machine needs to switch from multiple dimensions, so that the control of the working state of the aromatherapy machine can achieve a more ideal effect.

[0145] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the intelligent control method of the working state of the aromatherapy machine of any one of the above embodiments is implemented. The computer-readable storage medium includes, but is not limited to, any type of disk (including floppy disk, hard disk, optical disk, CD-ROM, and magneto-optical disk), ROM (Read-Only Memory), RAM (Random Access Memory), EPROM (Erasable Programmable Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory, magnetic card or optical card. That is, the storage device includes any medium that can store or transmit information in a readable form by a device (for example, a computer, a mobile phone), which can be a read-only memory, a disk or an optical disk, etc.

[0146] Embodiment three:

[0147] See also Figure 3 , Figure 3 It is a schematic diagram of the structure of an electronic device in an embodiment of the present invention.

[0148] The embodiment of the present invention further provides an electronic device, such as Figure 3 As shown, the electronic device includes a memory 31, a processor 33, and a computer program 32 stored in the memory 31 and executable on the processor 33. Those skilled in the art will appreciate that Figure 3The electronic device shown does not constitute a limitation on all devices, and may include more or fewer components than shown, or combine certain components. The memory 31 can be used to store the computer program 32 and various functional modules, and the processor 33 runs the computer program 32 stored in the memory 31, thereby executing various functional applications and data processing of the device. The memory can be an internal memory or an external memory, or include both internal and external memories. The internal memory can include a read-only memory (ROM), a programmable ROM (PROM), an electrically programmable ROM (EPROM), an electrically erasable programmable ROM (EEPROM), a flash memory, or a random access memory. The external memory can include a hard disk, a floppy disk, a ZIP disk, a U disk, a magnetic tape, etc. The processor 33 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general processor may be a microprocessor, a single chip microcomputer or the processor 33 may also be any conventional processor, etc. The processor and memory disclosed in the present invention include but are not limited to these types of processors and memories. The processor and memory disclosed in the present invention are only examples and not limitations.

[0149] As an embodiment, the electronic device includes: one or more processors 33, a memory 31, and one or more computer programs 32, wherein the one or more computer programs 32 are stored in the memory 31 and are configured to be executed by the one or more processors 33, and the one or more computer programs 32 are configured to execute the intelligent control method of the working state of the aromatherapy machine in any of the above embodiments. Please refer to the above embodiments for the specific implementation process, which will not be repeated here.

[0150] In an embodiment of the present invention, scenario analysis is performed based on linear relationship analysis using pre-processed real-time environmental data, so that the scenario in which the user is currently located can be analyzed more accurately. Feature extraction is performed on the real-time user facial image based on confidence analysis, and speech emotion features are generated based on the target tone features, target amplitude spectrum and wavelet packet coefficients generated by the real-time user voice, so that more comprehensive and accurate facial expression features and speech emotion features can be obtained. The current emotion of the user is identified based on facial expression features and speech emotion features combined with emotion fluctuation analysis to avoid excessive deviation between the identified user's current emotion and the actual situation. User preference analysis is performed based on the current emotion recognition result, and the preference data under the current emotion can be determined for the user's current emotion. The switching control of the working state of the aromatherapy machine is performed based on the real-time scenario data, the current emotion recognition result and the user preference data, so as to obtain sufficiently comprehensive and accurate data on the control of the working state of the aromatherapy machine, and match the working state that the aromatherapy machine needs to switch from multiple dimensions, so that the control of the working state of the aromatherapy machine can achieve a more ideal effect.

[0151] In addition, the above is a detailed introduction to the intelligent control method and related devices for the working state of an aromatherapy machine provided by an embodiment of the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as a limitation to the present invention.

Claims

1. An intelligent control method for the working state of an aromatherapy machine, characterized in that: The method comprises: Acquire real-time environment data and real-time user status data, and pre-process the real-time environment data and real-time user status data to obtain pre-processed real-time environment data and real-time user status data; Based on linear relationship analysis, pre-processed real-time environmental data is used to perform scenario analysis to obtain real-time scenario data; Extracting a real-time user facial image and a real-time user voice from the preprocessed real-time user status data, performing feature extraction on the real-time user facial image based on confidence analysis to obtain facial expression features, and performing feature extraction on the real-time user voice to obtain voice emotion features; Based on the facial expression features and voice emotion features combined with emotion fluctuation analysis, the user's current emotion recognition is performed to obtain a current emotion recognition result; Performing user preference analysis based on the current emotion recognition result to obtain user preference data corresponding to the current emotion recognition result; Switching control of the working state of the aromatherapy machine is performed based on the real-time scenario data, the current emotion recognition result and the user preference data; Among them, the feature extraction of the real-time user speech to obtain speech emotion features includes: extracting tone features of the real-time user speech to obtain target tone features; extracting spectrograms of the real-time user speech to obtain target spectrograms, removing silence segments from the target spectrograms to obtain target spectrograms after the silence segments are removed; calculating harmonic intensity trajectories using linear prediction coding based on the target spectrograms after the silence segments are removed to obtain target harmonic intensity trajectories, and generating target amplitude spectrum based on the target harmonic intensity trajectories using Fourier transform; performing wavelet packet decomposition on the target spectrograms after the silence segments are removed based on wavelet packet basis functions to obtain corresponding wavelet packet coefficients; and generating speech emotion features based on the target tone features, target amplitude spectrum and wavelet packet coefficients.

2. The intelligent control method of the working state of the aromatherapy machine according to claim 1 is characterized in that: The method of performing scenario analysis based on linear relationship analysis using pre-processed real-time environmental data to obtain real-time scenario data includes: Obtain historical environmental data, perform linear relationship fitting between the historical environmental data and each application scenario, and obtain the target linear relationship; Performing aggregation processing based on the historical environmental data to obtain an environmental data aggregation interval; Performing distribution interval analysis based on the historical environmental data to obtain distribution interval analysis data, and generating scenario division data based on the environmental data aggregation interval and the distribution interval analysis data; Performing cluster analysis based on the historical environmental data and the target linear relationship to obtain cluster analysis results; A scenario division model is generated based on the cluster analysis result and the scenario division data, and scenario analysis is performed based on the scenario division model using the preprocessed real-time environmental data to obtain real-time scenario data.

3. The intelligent control method of the working state of the aromatherapy machine according to claim 1 is characterized in that: The step of extracting features from the real-time user facial image based on confidence analysis to obtain facial expression features includes: Performing regional object division on the real-time user facial image to obtain the real-time user facial image after regional object division; Based on wavelet transform, the real-time user facial image after the regional object division is subjected to amplitude spectrum feature extraction to obtain target amplitude spectrum feature, and the target amplitude spectrum feature is subjected to histogram extraction to obtain a target histogram; Extracting a region of interest from the real-time user facial image after the region object division, obtaining a channel pixel grayscale value of the region of interest, and determining facial blood flow distribution characteristics based on the channel pixel grayscale value; Performing confidence analysis on the target histogram and facial blood flow distribution features based on an adaptive feature learning mechanism to obtain corresponding self-attention confidence; The target histogram and facial blood flow distribution features are fused with the corresponding self-attention confidence to obtain the facial expression features.

4. The intelligent control method of the working state of the aromatherapy machine according to claim 1 is characterized in that: The method of performing current emotion recognition of the user based on the facial expression features and the voice emotion features combined with emotion fluctuation analysis to obtain the current emotion recognition result includes: Performing dimensionality reduction and weighting processing on the facial expression features and the voice emotion features to obtain target facial expression features and target voice emotion features; Matching the emotion fluctuation value in the emotion database based on the target facial expression feature and the target voice emotion feature to obtain the target emotion fluctuation value; Generate an emotion fluctuation curve using the target emotion fluctuation value based on the emotion correlation function and the emotion recognition network; Based on the target facial expression features and the target voice emotion features, Naive Bayes is used to perform emotion classification to obtain a corresponding conditional probability set; The user's current emotion recognition is performed based on the emotion fluctuation curve and the conditional probability set to obtain a current emotion recognition result.

5. The intelligent control method of the working state of the aromatherapy machine according to claim 1 is characterized in that: The performing user preference analysis based on the current emotion recognition result to obtain user preference data corresponding to the current emotion recognition result includes: Acquire historical user emotion data and historical user preference data of a preset time series, and perform preference feature evolution analysis on the historical user preference data based on dynamic factors and evolution factors to obtain preference feature evolution data; Establishing a mapping relationship between the historical user sentiment data and preference feature evolution data; Based on the current emotion recognition result, the mapping relationship is used to perform an aggregate analysis of the user's personalized preferences to obtain user preference data corresponding to the current emotion recognition result.

6. The intelligent control method of the working state of the aromatherapy machine according to claim 1, characterized in that: The switching control of the aromatherapy machine working state based on the real-time scenario data, the current emotion recognition result and the user preference data includes: Based on the real-time scenario data, the current emotion recognition result and the user preference data matching the corresponding target working state of the aroma diffuser, the aroma diffuser switches to the target working state to perform related operations.

7. An intelligent control device for the working state of an aromatherapy machine, characterized in that: The device comprises: A data preprocessing module, used to obtain real-time environment data and real-time user status data, and preprocess the real-time environment data and real-time user status data to obtain preprocessed real-time environment data and real-time user status data; A scenario analysis module, used to perform scenario analysis based on linear relationship analysis using pre-processed real-time environmental data to obtain real-time scenario data; A feature extraction module is used to extract a real-time user facial image and a real-time user voice from the preprocessed real-time user status data, perform feature extraction on the real-time user facial image based on confidence analysis to obtain facial expression features, and perform feature extraction on the real-time user voice to obtain voice emotion features; An emotion recognition module is used to recognize the user's current emotion based on the facial expression features and voice emotion features combined with emotion fluctuation analysis to obtain a current emotion recognition result; A preference analysis module, used to perform user preference analysis based on the current emotion recognition result to obtain user preference data corresponding to the current emotion recognition result; An aromatherapy machine working state control module, used to switch the working state of the aromatherapy machine based on the real-time scenario data, the current emotion recognition result and the user preference data; Among them, the feature extraction module extracts features from the real-time user speech to obtain speech emotion features, specifically including: extracting tone features from the real-time user speech to obtain target tone features; extracting spectrograms from the real-time user speech to obtain target spectrograms, removing silence segments from the target spectrograms to obtain target spectrograms after the silence segments are removed; calculating harmonic intensity trajectories using linear prediction coding based on the target spectrograms after the silence segments are removed to obtain target harmonic intensity trajectories, and generating target amplitude spectra based on the target harmonic intensity trajectories using Fourier transform; performing wavelet packet decomposition on the target spectrograms after the silence segments are removed based on wavelet packet basis functions to obtain corresponding wavelet packet coefficients; generating speech emotion features based on the target tone features, target amplitude spectrum and wavelet packet coefficients.

8. An electronic device, comprising a processor and a memory, characterized in that: The memory is used to store instructions, and the processor is used to call the instructions in the memory, so that the electronic device executes the intelligent control method for the working state of the aromatherapy machine according to any one of claims 1 to claim 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed on the electronic device, the electronic device executes the intelligent control method for the working state of the aromatherapy machine according to any one of claims 1 to 6.

Images