Household electrical appliance, and smell detection method and system for household electrical appliance

Abstract

The invention discloses a household appliance and a smell detection method and system.The smell detection method for the household appliance comprises the steps that environment data in a cavity of the household appliance are obtained, and the environment data comprise gas data; the environment data is input into a preset machine learning model, a classification result is obtained, the preset machine learning model is used for predicting the current odor attribute in the cavity according to the environment data, and the preset machine learning model is iteratively updated on the basis of feedback data, the feedback data is obtained from a user associated with the household appliance and is at least used for representing the predicted satisfaction degree of the classification result. According to the scheme, the smell detection logic of the household electrical appliance can be customized according to user preferences, and the intelligent degree of the household electrical appliance can be improved.

Description

Technical Field

[0001] Embodiments of the present invention relate to the technical field of household appliances, and particularly to a household appliance, an odor detection method and system for a household appliance. Background Art

[0002] As one of the five senses, smell affects all aspects of people's daily life and work. For example, the odor emitted by a household appliance during use will directly affect the user experience, which is also one of the evaluation indicators for users to choose a household appliance.

[0003] Taking a refrigerator as an example, the refrigerator extends the shelf life of food by forming a low-temperature enclosed space (which can be called a chamber). Therefore, people usually use the refrigerator to store perishable foods such as meat and vegetables. The food placed in the refrigerator will inevitably produce odors. Since the chamber for storing food is enclosed, the odors produced by the food cannot dissipate on their own, and the odors will become stronger over time. If the odors in the chamber cannot be detected and removed in time, it will have an adverse impact on the user experience.

[0004] There are mainly two existing odor detection solutions for refrigerators: The first is to detect the gas concentration in the chamber. The disadvantage is that it cannot distinguish the odor type, so it is impossible to specifically remove the odors that cause discomfort to users. The second is to detect specific odor types predefined in the laboratory. However, it is far from enough to detect odors only according to the laboratory definition. On the one hand, different users have very different and unpredictable preferences for odors. It is possible that the laboratory defines an odor as unpleasant, User A agrees that this odor is unpleasant, but User B actually thinks this odor is pleasant. At this time, the same set of odor detection logic cannot meet the preference feelings of all users. On the other hand, sometimes a user's preference for the same odor will change over time, while the odor detection logic predefined in the laboratory is fixed and unchanged.

[0005] Therefore, the existing odor detection logic of household appliances is not intelligent enough to meet the personalized needs of users. Summary of the Invention

[0006] An object of embodiments of the present invention is to provide an improved odor detection method for a household appliance.

[0007] Therefore, an embodiment of the present invention provides a method for detecting the smell of a household appliance, including: obtaining environmental data in the cavity of the household appliance, where the environmental data includes gas data; inputting the environmental data into a preset machine learning model and obtaining a classification result, where the preset machine learning model is used to predict the current smell attribute in the cavity according to the environmental data, and the preset machine learning model is iteratively updated based on feedback data, and the feedback data is obtained from a user associated with the household appliance and is at least used to characterize the prediction satisfaction for the classification result.

[0008] When performing smell detection in the prior art, the same set of detection logics is applied to household appliances of all users, which cannot reflect the personal preferences of users and is not personalized and intelligent enough. In contrast, the preset machine learning model provided in this implementation can be iteratively updated according to user feedback, so that the smell detection logic of the household appliance can better match the personal preferences of the users who use the household appliance. Thus, users are allowed to customize the smell detection logic in the household appliance in a personalized manner. The household appliance applying this implementation has a higher degree of intelligence, can perform smell detection based on the same preference feelings as the users, and makes the smell detection result conform to the user's preferences.

[0009] Optionally, the process of iteratively updating the preset machine learning model based on the feedback data includes: receiving the feedback data; determining the corresponding environmental data according to the generation time of the feedback data; constructing a training set and a validation set based on at least the feedback data and the corresponding environmental data received within a period of time; training the preset machine learning model based on the training set to obtain an updated preset machine learning model; and validating the updated preset machine learning model based on the validation set. Thus, the preset machine learning model is iteratively updated according to user feedback and the environmental data corresponding to this user feedback, so that the prediction result of the preset machine learning model better conforms to the user's preferences, and more personalized and intelligent smell detection of household appliances is realized.

[0010] Optionally, the validation set further includes a basic data set, where the basic data set includes standard environmental data and corresponding standard classification results. Thus, it is ensured that the prediction result of the preset machine learning model always conforms to the basic evaluation standard, and the situations of overfitting and prediction distortion are avoided. For example, the smell of spoiled food is usually considered unpleasant, so it is used as the content of the basic data set to ensure that the iteratively updated preset machine learning model always makes predictions on a certain baseline.

[0011] Optionally, the environmental data is collected periodically. Determining the corresponding environmental data according to the generation time of the feedback data includes: determining the environmental data with the collection time closest to the generation time of the feedback data as the corresponding environmental data. Thereby, the feedback data is accurately matched with the environmental data in the household appliance chamber that triggered the current user feedback, ensuring that the training data provided for training the preset machine learning model itself has high accuracy and ensuring that the model training result better conforms to the user's preference.

[0012] Optionally, the preset machine learning model is iteratively updated periodically as the feedback data accumulates. Thereby, the model is continuously trained with the accumulation of user usage data, making the prediction of the preset machine learning model more targeted and capable of reflecting user personalization. Further, the preset machine learning model is updated periodically. After accumulating a sufficient amount of feedback data in each period, a unified iterative update is performed once to ensure a better update effect of the preset machine learning model while saving costs.

[0013] Optionally, the preset machine learning model is iteratively updated based on the feedback data on the basis of the initial machine learning model, wherein the initial machine learning model is trained based on a basic data set, and the basic data set includes standard environmental data and corresponding standard classification results. When the household appliance leaves the factory, the initial machine learning model can be preset, and the machine learning algorithm is automatically updated according to the user's feedback on the odor in the chamber to obtain an iteratively updated preset machine learning model to reflect user personalization and intelligence. Further, the same initial machine learning model is preset for multiple household appliances. As multiple users use them, the preset machine learning models of each household appliance are iteratively updated according to the feedback of the users associated with them, increasingly reflecting the personal preferences of the users associated with them. Thus, the differences in the prediction results of the same environmental data by each preset machine learning model may become larger and larger. Thereby, user-personalized odor detection is realized.

[0014] Optionally, the initial machine learning model includes multiple sub-models, and different sub-models are associated with different odor preferences. The preset machine learning model is iteratively updated based on the feedback data on the basis of the initial machine learning model, including: obtaining the odor preferences of the user for at least one odor; determining the corresponding sub-model according to the obtained odor preferences; and iteratively updating the preset machine learning model based on the feedback data on the basis of the determined sub-model. Thus, when the user uses the household appliance for the first time, they can obtain an odor detection solution with a certain personalized customization function, and the personalized customization effect will be gradually enhanced as the user uses it. Further, the user uploads personal preference options, and the background records and analyzes them and establishes a connection with the initial machine learning model. Refrigerators of users with the same or similar personal preferences initially use the same initial machine learning model, and then gradually update into a more personalized preset machine learning model as the usage time increases.

[0015] Optionally, the method further includes: periodically sending a prompt message, where the prompt message includes a feedback form; receiving the feedback form, and generating the feedback data based on the feedback form. Thus, the user is periodically and actively prompted to submit feedback, and as much feedback data as possible is collected for training the preset machine learning model to optimize the model training effect.

[0016] Optionally, the prompt message and / or the feedback form are transmitted through the display and / or input unit of the household appliance, and / or the prompt message and / or the feedback form are transmitted through a terminal device associated with the household appliance. Thus, the user is reminded to give feedback through the human-computer interaction interface provided on the household appliance, and the feedback data is received to be used as the basis for model training. Based on the terminal device, it is convenient for the user to receive the feedback form and upload the feedback data anytime and anywhere.

[0017] Another object of the embodiments of the present invention is to provide an improved household appliance.

[0018] Therefore, an embodiment of the present invention provides a household appliance, including: a body defining a chamber; an odor detector disposed in the chamber to collect environmental data in the chamber; a control module disposed in the body, the control module communicating with the odor detector to receive the environmental data, the control module being configured to execute the above method and generate a control instruction according to the current odor attribute in the chamber; and a deodorizing module disposed in the chamber, in response to receiving the control instruction, the deodorizing module adjusting an operating parameter according to the control instruction.

[0019] Therefore, the odor detection method provided based on this implementation solution can perform odor detection according to the user's personal preferences, remove odors in a timely manner when detecting odors that the user does not like, and intelligently keep the odor in the chamber in line with the user's preferences. Further, the actions of model iteration and update are all completed within the household appliance, so that the household appliance itself can complete model optimization without relying on external server support or the network, and has a fast response speed.

[0020] Optionally, the household appliance is selected from: a refrigeration appliance, a washing machine, a dishwasher, and an oven.

[0021] Optionally, the household appliance further includes: a communication module disposed on the body, and the control module receives feedback data through the communication module; and / or, a display and / or input unit disposed on the body, and the control module receives feedback data through the display and / or input unit. Thus, the user can transmit feedback data through network communication or directly input feedback data through the display and / or input unit, improving the convenience of feedback.

[0022] Another object of the embodiments of the present invention is to provide an improved odor detection system for a household appliance.

[0023] Therefore, the embodiments of the present invention provide an odor detection system for a household appliance, including: a household appliance including a body and a control module, where the control module is used to execute the above method; a server communicatively connected to the control module, and the server is used to synchronize the preset machine learning model to the control module.

[0024] Thus, the server can be used to store historical data, including historical environmental data and corresponding feedback data, and retrain the preset machine learning model based on the historical data, so that the current odor attributes predicted by the household appliance based on the updated preset machine learning model are more in line with the user's preferences. Further, when the server is externally disposed to the household appliance, it is beneficial to reduce the number of components in the household appliance and reduce costs.

[0025] Optionally, the server is used to iteratively update the preset machine learning model based on the feedback data and synchronize the updated preset machine learning model to the control module. Thus, with the support of the high computing power of the server, the iteration response speed of the preset machine learning model can be further improved.

[0026] Optionally, the system further includes: a communication module disposed on the body, and the control module communicates with the server through the communication module. Thus, the household appliance establishes a communication connection with the outside world through the communication module to achieve remote update of the preset machine learning model. The outside can be, for example, a server.

[0027] Optionally, the system further includes a display and / or input unit disposed on the body and communicating with the control module, and the display and / or input unit is configured to receive the feedback data. Thus, the user can submit feedback data locally on the household appliance, and the household appliance together with the corresponding environmental data is summarized to the server side, so that the server iteratively updates the preset machine learning model based on these data. Description of the Drawings

[0028] Figure 1 is a flowchart of a method for detecting the smell of a household appliance according to an embodiment of the present invention;

[0029] Figure 2 is a flowchart of the iterative update process of the preset machine learning model based on the feedback data in the embodiment of the present invention;

[0030] Figure 3 is a schematic diagram of a typical application scenario of a household appliance according to an embodiment of the present invention;

[0031] Figure 4 is a schematic diagram of the principle of a smell detection system for a household appliance according to an embodiment of the present invention;

[0032] In the drawings:

[0033] 1 - Household appliance; 10 - Body; 101 - Chamber; 102 - Refrigerating chamber; 103 - Freezing chamber; 104 - Door; 11 - Smell detector; 12 - Control module; 13 - Deodorizing module; 14 - Communication module; 15 - Display and / or input unit; 2 - Server. Detailed Embodiments

[0034] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following detailed description of the specific embodiments of the present invention will be given with reference to the accompanying drawings.

[0035] Figure 1 is a flowchart of a method for detecting the smell of a household appliance according to an embodiment of the present invention.

[0036] This implementation scheme can be applied to the scenario of monitoring the smell inside a household appliance, such as monitoring whether there is an abnormal smell inside the household appliance, or monitoring the real-time smell attributes inside the household appliance. The smell attributes can be used to describe the smell pleasantness inside the household appliance. By adopting this implementation scheme, the smell emitted inside the household appliance can be classified, for example, at least including two categories: pleasant smell and unpleasant smell. In actual applications, the classification criteria and the number of classification categories of the smell attributes can be adjusted according to needs. For example, the classification results obtained by adopting this implementation scheme can be further subdivided into five categories: very pleasant smell, pleasant smell, neutral smell, unpleasant smell, and extremely unpleasant smell (also known as very unpleasant smell).

[0037] The items stored in the household appliance can belong to a single category or multiple categories. This implementation can specifically detect the odor attributes of the odors generated by specific types of items in the household appliance, or macroscopically detect the odor attributes of the mixed odors generated by multiple items in the household appliance as a whole.

[0038] The household appliance described in this implementation can be selected from: refrigeration appliances, washing machines, dishwashers, and ovens.

[0039] In the odor detection application scenario of a refrigeration appliance, the odor of the food stored in the chamber (e.g., the refrigerator compartment) of the refrigeration appliance can be detected.

[0040] In the odor detection application scenario of a washing machine, the odor in the washing chamber (e.g., the drum) of the washing machine can be detected. For example, if the user does not take out the wet clothes in time after using the washing machine, unpleasant odors may be generated after the wet clothes are left in the drum for a long time. At this time, by implementing this implementation, the odor can be detected in a timely and accurate manner.

[0041] In the odor detection application scenario of a dishwasher / oven, the odor in the chamber after using the dishwasher / oven can be detected. For example, through this implementation, it can be detected whether there is a smell of cooking fumes, the smell of food residues, etc. remaining in the dishwasher / oven.

[0042] This implementation can be executed by a control module. The control module can be, for example, the single-chip microcomputer of the household appliance, or can also be, for example, the control unit dedicated to implementing this implementation in the household appliance. In this embodiment, the control module can be arranged in the main body of the household appliance.

[0043] Specifically, referring to Figure 1 , the odor detection method of the household appliance described in this embodiment can include the following steps:

[0044] Step S101, obtain the environmental data in the chamber of the household appliance, where the environmental data includes gas data;

[0045] Step S102, input the environmental data into a preset machine learning model and obtain a classification result, where the preset machine learning model is used to predict the current odor attribute in the chamber according to the environmental data, the preset machine learning model is iteratively updated based on feedback data, and the feedback data is obtained from a user associated with the household appliance and is at least used to characterize the prediction satisfaction for the classification result.

[0046] Further, the gas data may include concentration information of at least one gas component in the chamber. Different types of items may produce different types of gas components under different odor attributes, and for the user's intuitive feeling, it is to smell different odors and then generate odor pleasantness evaluations of different levels. Taking food ingredients as an example, food ingredients can include meat, nuts, vegetables, etc. according to types. The odor attributes can include very pleasant, pleasant, neutral, unpleasant, and extremely unpleasant.

[0047] Regarding the judgment criteria for each odor attribute, they can be determined according to the user's sensitivity to odors, and can also be determined according to relevant industry standards. Further, in this implementation scheme, the judgment criteria when the preset machine learning model makes predictions can be corrected in real time according to user feedback, so that the classification results are more in line with the user's personal preferences.

[0048] In a specific implementation, the gas data can be collected by a gas detector disposed in the chamber. Specifically, step S101 may include the steps of: adjusting the operating temperature of the gas detector during the operation of the gas detector, and receiving candidate gas data respectively collected by the gas detector at multiple operating temperatures, where the candidate gas data associated with different operating temperatures is used to characterize the concentrations of different types of gas components; generating the gas data based on the received multiple candidate gas data. Thus, the operating temperature of the gas detector is adjusted within a certain temperature range to enrich the diversity of the output data of the gas detector, so as to realize the detection of multiple gas components.

[0049] Further, the gas detector may include a volatile organic compounds (VOC) sensor. The VOC sensor reacts strongly to specific types of gases at different temperatures. Using this characteristic, the gas detector used in this implementation scheme can detect multiple gas components within a certain temperature range. For example, at a specific operating temperature, the VOC sensor reacts with a specific type of gas component in the air and outputs a resistance value, and this resistance value can be used to characterize the concentration of this specific type of gas component.

[0050] In some embodiments, the VOC sensor may collect the gas data in the chamber at preset intervals. For example, the VOC sensor collects the gas data in the chamber every n seconds.

[0051] Further, each time the gas data is collected, the preset machine learning model outputs a classification result based on the gas data obtained this time. Thus, a corresponding odor attribute prediction result is obtained each time the gas data is collected.

[0052] Alternatively, the detection period of the VOC sensor and the prediction period of the preset machine learning model may not be exactly the same. For example, the VOC sensor collects gas data in the chamber every 10 seconds. The control module can compare the gas data obtained each time and input the latest obtained gas data into the preset machine learning model when a mutation occurs in the specific values of the gas data obtained before and after, to obtain a classification result. In this way, performing subsequent prediction operations only when the gas data mutates helps reduce the computational overhead.

[0053] In some embodiments, the environmental data may further include the temperature and humidity data in the chamber. Specifically, the environmental temperature and humidity in the chamber can be used as compensation values for the gas data and input into the preset machine learning model together as the environmental data. These three types of data can comprehensively reflect the complex gas composition in the chamber, which is beneficial to improving the accuracy of model prediction.

[0054] Taking the odor detection of a refrigeration appliance as an example, in this example, the environmental data may include the gas data and the operating parameters of the refrigeration appliance when the gas data is obtained, where the operating parameters may include the compartment temperature and the compartment humidity. Further, the operating parameters can be directly obtained from the control module of the refrigeration appliance. This can also obtain the environmental data required for model prediction, and the operating parameters are fixed values and are not affected by the external environment when the door of the refrigeration appliance is opened or closed. Therefore, the effect of using the operating parameters as one of the environmental data in improving the reliability of the input data can be expected.

[0055] In a specific implementation, the preset machine learning model can be constructed based on a machine learning algorithm. The machine learning algorithm can be, for example, decision tree, naive Bayes classification, least squares regression, logistic regression, support vector machine, neural network, deep learning, multilayer perceptron (abbreviated as MLP), random forest algorithm, Extreme Gradient Boosting Decision Tree (abbreviated as XGBoost), and K-Nearest Neighbor (abbreviated as KNN) classification algorithm, etc.

[0056] In some embodiments, different odor attributes can be characterized by different numerical values or scores in different intervals, and the scores represent the level of odor pleasantness in the household appliance. For example, a score of 1 indicates good smell, and a score of 0 indicates bad smell. Another example is that a score in the range of [0, 1) indicates bad smell, and a score greater than or equal to 1 indicates good smell.

[0057] In step S102, the output of the preset machine learning model may include a score obtained by scoring the odor pleasantness in the chamber based on environmental data. Based on this score, the current odor attribute in the chamber can be determined.

[0058] In a specific implementation, in response to obtaining the classification result, the control module may control the corresponding components of the household appliance to perform different actions based on the classification result to keep the odor in the chamber at a better odor pleasantness. For example, if the classification result is unpleasant, the control module may control the deodorization module of the refrigeration appliance to work for deodorization.

[0059] In a specific implementation, the user associated with the household appliance may be, for example, the user of the household appliance. When using the household appliance, the user will have a subjective judgment on the odor emitted by the chamber of the household appliance, and the user feeds back this subjective judgment to the household appliance. Based on this user feedback, it can be determined whether the user is satisfied with the odor in the chamber that the user smells, and further determine whether the user is satisfied with the operation result of the control module performing corresponding operations based on the classification result of the preset machine learning model, which can essentially represent whether the user is satisfied with the classification result predicted by the preset machine learning model.

[0060] For example, assume that the preset machine learning model predicts that the odor in the chamber is pleasant at a certain moment, and the control module controls the deodorization module to remain in a standby state, but the user opens the door and finds that the odor in the chamber is unpleasant. In this example, the user can provide feedback data indicating dissatisfaction with the odor in the chamber, and this feedback data can represent that the user is dissatisfied with the classification result predicted by the preset machine learning model.

[0061] Another example, assume that the preset machine learning model predicts that the odor in the chamber is unpleasant at a certain moment, and the control module controls the deodorization module to work, and the user opens the door and finds that the odor in the chamber is pleasant. In this example, the user can also provide feedback data indicating satisfaction with the odor in the chamber. This feedback data can represent that the user is satisfied with the classification result of the preset machine learning model.

[0062] In a specific implementation, referring to Figure 2 , the process of iteratively updating the preset machine learning model based on the feedback data may include the following steps:

[0063] Step S201, receiving the feedback data;

[0064] Step S202, determining the corresponding environmental data according to the generation time of the feedback data;

[0065] Step S203, constructing a training set and a validation set based at least on the feedback data and the corresponding environmental data received within a period of time;

[0066] Step S204: Train the preset machine learning model based on the training set to obtain an updated preset machine learning model;

[0067] Step S205: Validate the updated preset machine learning model based on the validation set.

[0068] Specifically, the feedback data can be received from user input.

[0069] Furthermore, the generation time of the feedback data can be the time when the user inputs the feedback data, or the time when the control module receives the feedback data.

[0070] Furthermore, the received feedback data and the corresponding environmental data can be accumulated within a period of time, and the preset machine learning model can be updated after accumulating enough data.

[0071] Thus, the preset machine learning model is iteratively updated according to the user feedback and the environmental data corresponding to the current user feedback, so that the prediction result of the updated model is more in line with the user preference, realizing more personalized and intelligent odor detection for household appliances.

[0072] In a specific implementation, the environmental data can be collected periodically. Step S202 can include the step of determining the environmental data with the collection time closest to the generation time of the feedback data as the corresponding environmental data.

[0073] Specifically, the periodically collected environmental data can be stored in the storage module. The storage module can be set in the household appliance or externally located and communicate with the control module.

[0074] Furthermore, when storing, the environmental data can be stored corresponding to the collection time of the environmental data. The collection time can be, for example, a timestamp.

[0075] In response to receiving the feedback data, the environmental data with the collection time closest to the generation time of the feedback data can be searched in the storage module according to the generation time of the feedback data, and the found environmental data is determined to correspond to the currently received feedback data.

[0076] Thus, the feedback data and the environmental data in the household appliance chamber that triggered the current user feedback are accurately matched, ensuring that the training data provided for training the preset machine learning model itself has high accuracy and ensuring that the model training result is more in line with the user preference.

[0077] In some embodiments, the proximity between the acquisition time of environmental data and the generation time of feedback data can be a concept of absolute value. That is to say, environmental data whose acquisition time is earlier or later than the generation time of feedback data can both be regarded as the closest environmental data. The control module can select the environmental data with the smallest absolute value of the difference between the acquisition time and the generation time, or use both of these two environmental data for model training.

[0078] In a specific implementation, the training set can be used to train and generate an updated preset machine learning model, and then the validation set is used to verify whether the parameters of the updated preset machine learning model are appropriate.

[0079] Furthermore, for the feedback data and corresponding environmental data received within a period of time, the training set and the validation set can be obtained by dividing them according to a preset ratio. For example, the preset ratio can be 7:3, with the former being the training set and the latter being the validation set.

[0080] Furthermore, if the verification result in step S205 shows that the accuracy of the prediction result of the updated preset machine learning model is higher than the preset threshold, it can be confirmed that the updated preset machine learning model meets the requirements (for example, there is no overfitting phenomenon). At this time, the updated preset machine learning model can be used for the next odor detection scheme, that is, the preset machine learning model used in step S102 can be the updated preset machine learning model described in this example.

[0081] The preset threshold can be, for example, 95%. In practical applications, the specific value of the preset threshold can be adjusted according to user needs.

[0082] In a variant, if the verification result in step S205 shows that the accuracy of the prediction result of the updated preset machine learning model is lower than the preset threshold, then steps S204 and S205 can be re-executed to retrain the preset machine learning model until the updated preset machine learning model passes the verification of the validation set.

[0083] Furthermore, when re-executing steps S204 and S205, an updated training set and validation set can be re-divided.

[0084] In a specific implementation, the validation set can also include a basic data set, where the basic data set includes standard environmental data and corresponding standard classification results.

[0085] Specifically, the basic data set can include laboratory data, which can specifically be standard results pre-calibrated manually. For the odor emitted by a specific type of item, people's judgments on this odor are usually consistent. Then, the environmental data and corresponding classification results of this odor are defined through pre-calibration in the laboratory as the standard environmental data and corresponding standard classification results.

[0086] For example, the odor attribute of rotten meat is usually unpleasant. Then, the environmental data when rotten meat is stored in the chamber can be collected as the standard environmental data and stored in the basic data set correspondingly with the corresponding standard classification result "unpleasant".

[0087] Furthermore, in step S203, a part of the feedback data received within a period of time and the corresponding environmental data can be divided into a training set, and the basic data set and the remaining part of the feedback data received within a period of time and the corresponding environmental data can be jointly divided into a validation set.

[0088] Thus, it is ensured that the prediction result of the preset machine learning model always conforms to the basic evaluation criteria, avoiding overfitting and prediction distortion. For example, the odor of spoiled food is usually considered unpleasant. Then, as the content of the basic data set, it is ensured that the iteratively updated preset machine learning model always maintains a certain baseline for prediction.

[0089] In a specific implementation, the preset machine learning model can be iteratively updated periodically as the feedback data accumulates.

[0090] Specifically, the update period of the preset machine learning model can be a fixed time interval, such as every day, every week, etc.

[0091] Alternatively, the update period of the preset machine learning model can be dynamically adjusted according to the accumulation of feedback data, and an iterative update is triggered every time enough feedback data is accumulated. For example, every time 10 feedback data are received cumulatively, an iterative update of the preset machine learning model is triggered.

[0092] Thus, the model is continuously trained as the user usage data accumulates, making the prediction of the preset machine learning model more targeted and able to reflect user personalization. Further, the preset machine learning model is updated periodically. After enough feedback data is accumulated within each period, a unified iterative update is performed to ensure a better update effect of the preset machine learning model while saving costs.

[0093] In a specific implementation, the preset machine learning model can be obtained by iteratively updating based on the feedback data on the basis of the initial machine learning model, where the initial machine learning model is trained based on the basic data set.

[0094] Specifically, the initial machine learning model can be pre-installed when the household appliance leaves the factory. As the user uses it, environmental data and corresponding feedback data are continuously accumulated, and the initial machine learning model is continuously iteratively updated according to the user's preferences into the preset machine learning model and the updated preset machine learning model.

[0095] For example, an initial machine learning model is generated based on a basic data set, and is iteratively updated as user feedback data accumulates to obtain a preset machine learning model. After verifying that the model passes using the basic data set and a portion of the user feedback data, it is upgraded to a household appliance for the next odor attribute prediction.

[0096] Thus, the machine learning algorithm is automatically updated according to the user's feedback on the odor in the chamber to obtain a continuously iteratively updated preset machine learning model, so as to reflect user personalization and intelligence. Further, the same initial machine learning model is pre-installed in multiple household appliances. As multiple users use them, the preset machine learning models of each household appliance are continuously iteratively updated according to the feedback of the users associated with them, increasingly reflecting the personal preferences of the users associated with each of them. Therefore, the differences in the prediction results of the same environmental data by each preset machine learning model may become larger and larger. Thus, user personalized odor detection is achieved.

[0097] In a specific implementation, the initial machine learning model may include multiple sub-models, and different sub-models are associated with different odor preferences.

[0098] Specifically, different users may have different preferences for the same odor. For example, some people think the odor of durian is pleasant, while some people think the odor of coriander is unpleasant.

[0099] In this specific implementation, for the situation where the same environmental data (representing a specific odor) may correspond to different odor attributes, corresponding sub-models are constructed separately according to different odor attributes. Different sub-models can be trained using different basic data sets. Among them, the difference between different basic data sets is that the same standard environmental data may correspond to different standard classification results. For example, the odor preferences of liking and disliking the odor of durian correspond to different sub-models. In the respective basic data sets of these two sub-models, the standard classification results for the odor of durian are pleasant and unpleasant respectively.

[0100] Further, the preset machine learning model obtained by iterative update based on the feedback data on the basis of the initial machine learning model may include: obtaining the user's odor preference for at least one odor; determining the corresponding sub-model according to the obtained odor preference; and on the basis of the determined sub-model, iteratively updating to obtain the preset machine learning model based on the feedback data.

[0101] For example, after a user purchases a household appliance, the user's odor preference can be obtained through methods such as questionnaires. In response to obtaining the user's odor preference, the control module selects the corresponding sub-model as the initial machine learning model for actual use, and then continuously iteratively updates the initial machine learning model according to the user's subsequent feedback data as the user uses it.

[0102] Suppose the user feedback indicates a preference for the smell of durian. The initial machine learning model includes sub-model A (constructed with environmental data of durian smell and the smell attribute of being pleasant as input data) and sub-model B (constructed with environmental data of durian smell and the smell attribute of being unpleasant as input data). In response to receiving the feedback data that the user likes the smell of durian, the control module can determine that sub-model A is the sub-model corresponding to the user's smell preference.

[0103] Thus, when the user initially uses the household appliance, they can obtain a smell detection solution with certain personalized customization functions, and the personalized customization effect will be gradually enhanced as the user uses it. Further, when the user uploads personal preference options, the background records and analyzes them and establishes a connection with the initial machine learning model. Refrigerators of users with the same or similar personal preferences initially use the same initial machine learning model, and then gradually update to a more personalized preset machine learning model as the usage time increases.

[0104] In a specific implementation, for the feedback data and corresponding environmental data received within a period of time, before constructing the training set and validation set, the acquired data can be preprocessed first.

[0105] Specifically, data preprocessing can include data cleaning to achieve a denoising effect. For example, abnormal data in the acquired data, such as peaks in a continuous data segment, can be removed. Another example is that error data, such as data with packet loss during transmission, can be removed.

[0106] In a specific implementation, the smell detection method of this embodiment can further include the step of regularly sending a prompt message, where the prompt message includes an opinion feedback form; receiving the opinion feedback form and generating the feedback data based on the opinion feedback form.

[0107] Specifically, the prompt message can be sent at a fixed time every day to collect as much feedback data as possible for training the preset machine learning model and optimizing the model training effect.

[0108] Alternatively, every time it is detected that the user uses the household appliance, a prompt message can be sent. For example, after the control module of a refrigeration appliance detects a door opening or closing action, it actively sends a prompt message.

[0109] Alternatively, the feedback data can also be actively triggered and uploaded by the user.

[0110] In a specific implementation, the prompt message and / or the opinion feedback form can be transmitted through the display and / or input unit of the household appliance.

[0111] Specifically, the display and / or input unit can include a User Interface Module (UIM for short).

[0112] For example, a pop-up window can be periodically displayed on the control panel of the household appliance to ask the user whether they are satisfied with the smell in the current chamber.

[0113] Thus, the user is reminded to give feedback through the man-machine interaction interface provided on the household appliance, and feedback data is received as the basis for model training.

[0114] In a specific implementation, the prompt message and / or the feedback form can be transmitted through a terminal device associated with the household appliance. For example, the control module can periodically send a prompt message to the terminal device, and the prompt message can include a feedback form, and the terminal device receives the feedback form filled in and submitted by the user. Thus, the user can receive the feedback form and upload feedback data at any time and place, and the feedback process is more convenient.

[0115] In some embodiments, the terminal device can communicate with the household appliance, or be under the control of the same user as the household appliance. The terminal device can be, for example, a mobile terminal such as a mobile phone, an IPAD, a laptop computer, or can also be other smart home appliances located in the same local area network as the household appliance, such as a washing machine, a range hood, etc. in the user's home.

[0116] In a specific implementation, the prompt message and the feedback form can be displayed or received through different media. For example, the prompt message can be sent through the display and / or input unit of the household appliance, and the feedback form is received on the user's mobile phone.

[0117] As described above, by using the preset machine learning model provided by this implementation scheme, it can be iteratively updated according to user feedback, so that the smell detection logic of the household appliance can better match the personal preferences of the user using the household appliance. Thus, the user is allowed to customize the smell detection logic inside the household appliance, and the household appliance applying this implementation scheme has a higher degree of intelligence, can perform smell detection based on the same preferences as the user, and makes the smell detection result meet the user's preferences.

[0118] Figure 3 It is a schematic diagram of a typical application scenario of a household appliance 1 according to an embodiment of the present invention. Next, this application scenario will be described in detail taking the household appliance 1 as a refrigeration appliance as an example. The refrigeration appliance can include a refrigerator, a cold storage cabinet, etc.

[0119] Specifically, referring to Figure 3 , the household appliance 1 described in this embodiment may include: a body 10, which defines a chamber 101. For example, the chamber 101 of the refrigeration appliance may include a refrigerating chamber 102 and a freezing chamber 103. The chamber 101 may have an opening, and the household appliance 1 may include a door 104 that can open or close the opening of the corresponding chamber 101.

[0120] Furthermore, the household appliance 1 may further include an odor detector 11, which is disposed in the chamber 101 to collect the environmental data in the chamber 101. For example, the odor detector 11 may include a gas detector disposed on the top of the refrigerating chamber 102 for collecting the gas data in the refrigerating chamber 102. Also for example, a gas detector may be disposed near the air return opening of the refrigerating appliance to collect the overall gas data in the refrigerating appliance. The gas detector may be, for example, a VOC sensor.

[0121] Furthermore, the household appliance 1 may further include a control module 12, which is disposed in the main body 10. The control module 12 communicates with the odor detector 11 to receive the environmental data, and the control module 12 may be used to execute the Figure 1 and Figure 2 odor detection method shown above, and generate a control instruction according to the current odor attribute in the chamber 101.

[0122] The control module 12 may include the main control board / microcontroller of the refrigerating appliance. Or, the control module 12 may be a module dedicated to executing the Figure 1 and Figure 2 method shown above.

[0123] Furthermore, the household appliance 1 may further include a deodorizing module 13, which is disposed in the chamber 101. In response to receiving the control instruction, the deodorizing module 13 adjusts the operating parameters according to the control instruction.

[0124] The deodorizing module 13 and the odor detector 11 may be disposed in the same chamber 101. For example, both are disposed in the refrigerating chamber 102. Or, the deodorizing module 13 may be disposed in the air duct of the refrigerating appliance. In response to the control instruction, the deodorizing module 13 operates according to the working parameters indicated by the control instruction, and realizes the overall odor removal of all chambers 101 through the circulating refrigeration system in the refrigerating appliance.

[0125] The deodorizing module 13 may be, for example, an ionizer.

[0126] The odor detector 11 and the control module 12, and the control module 12 and the deodorizing module 13 may communicate in a wired or wireless manner.

[0127] Thus, based on the odor detection method provided by this embodiment, odor detection can be performed according to the user's personal preference. When an odor that the user does not like is detected, the odor is removed in time, and the odor in the chamber 101 is intelligently maintained to meet the user's preference. Further, the actions of model iteration and update are all completed within the household appliance 1, so that the household appliance 1 itself can complete model optimization without relying on an external server and without relying on the network, and the response speed is fast.

[0128] In some embodiments, a temperature and humidity sensor (not shown in the figure) may also be provided in the chamber 101 to collect the temperature and humidity data in the chamber 101.

[0129] In a specific implementation, the household appliance 1 may further include a storage module (not shown in the figure) for storing a preset machine learning model. When the control module executes Figure 1 the steps S102 shown, it can access the storage module to call the preset machine learning model.

[0130] Further, in response to the execution of Figure 2 the steps S201 to S205 shown, the updated preset machine learning model verified by the validation set can overwrite the preset machine learning model originally stored in the storage module for the control module to call when executing step S102 next time.

[0131] In a specific implementation, the household appliance 1 may further include a communication module 14 disposed on the body 10. The control module 12 receives feedback data through the communication module 14. For example, the communication module 14 can communicate with the user's terminal device to receive the feedback data sent by the user through the terminal device. In some embodiments, the communication module 14 can communicate with the terminal device through wireless communication technologies such as Wireless Fidelity (WIFI) and Near Field Communication (NFC).

[0132] The control module 12 can send a prompt message to the user's terminal device through the communication module 14 and receive the opinion feedback form filled in by the user.

[0133] Further, the control module 12 and the odor detector 11 can also communicate through the communication module 14.

[0134] Further, the control module 12 and the deodorization module 13 can also communicate through the communication module 14.

[0135] It should be noted that Figure 3 only the possible installation positions of the control module 12, the deodorization module 13, the odor detector 11, and the communication module 14 on the household appliance 1 are shown exemplarily. In actual applications, the mutual position relationship of each module and the specific installation position on the household appliance 1 can be adjusted as needed. Each module can be independent of each other or integrated on the same chip or integrated into the same functional module. For example, the control module 12 and the communication module 14 can be integrated together.

[0136] In a specific implementation, the household appliance 1 may further include a display and / or input unit 15 disposed on the body 10. The control module 12 receives feedback data through the display and / or input unit 15.

[0137] Specifically, the household appliance 1 may include an interaction panel to implement human-machine interaction. The interaction panel may include a touchpad, a display screen, and the like.

[0138] The interaction panel may include a display and / or input unit 15. In some embodiments, the display and / or input unit 15 may be a User Interface Module (UIM) for displaying the setting parameters or status of the household appliance 1 and / or receiving control instructions input by the user.

[0139] Furthermore, the interaction panel may further include a cover plate (not shown in the figure) disposed in front of the display and / or input unit 15 to at least provide a protection function. The cover plate may be, for example, the front panel of the door 104 of the household appliance 1, or may also be, for example, an independent glass plate covering in front of the UIM. In some embodiments, the display and / or input unit 15 may be closely attached to the rear side of the cover plate, so as to receive, for example, input signals input by the user via touching the cover plate. The display and / or input unit 15 may include a light-emitting member to enable light to pass through corresponding areas of the cover plate.

[0140] In some embodiments, the control module 12 may directly display a prompt message on the display and / or input unit 15 to remind the user to provide feedback. Furthermore, the control module 12 may receive an opinion feedback form input by the user through the display and / or input unit 15 to obtain feedback data.

[0141] For example, the display and / or input unit 15 may be a touch screen, and the user fills in the opinion feedback form through touch operations.

[0142] Also for example, the interaction panel may include a gesture sensing area, and the user fills in the opinion feedback form through gesture operations.

[0143] Thus, the user can either transmit the feedback data through network communication or directly input the feedback data through the display and / or input unit 15, improving the convenience of feedback.

[0144] In a typical application scenario, referring to Figure 3 , assume that durians are stored in the refrigerating chamber 102, the odor detector 11 collects the environmental data in the refrigerating chamber 102 and transmits it to the control module 12. The control module 12 calls a preset machine learning model (denoted as model A) from the storage module and inputs the obtained environmental data into the preset machine learning model. Assume that the classification result predicted by model A is unpleasant, the control module 12 controls the deodorizing module 13 to operate at a first power to remove the odor in the refrigerating chamber 102. Meanwhile, the control module 12 stores the obtained environmental data and the collection time of this environmental data.

[0145] The user opens the refrigerating chamber 102 and finds that the smell emitted from the refrigerating chamber 102 is still unpleasant. Then, the user feeds back data through the display and / or input unit 15, and the feedback data characterizes that the user is dissatisfied with the classification result predicted by model A this time. In response to receiving the feedback data, the control module 12 stores the feedback data and records the generation time of the feedback data. In addition, the control module 12 can also determine the environmental data corresponding to the feedback data.

[0146] After accumulating 10 sets of feedback data and the corresponding environmental data, the control module 12 retrains model A based on these 10 sets of data and the basic data set. After the updated preset machine learning model passes the verification of the verification set (denoted as model B), the control module 12 stores model B in the storage module to replace the original model A.

[0147] In this application scenario, the classification result of model B compared with that of model A for the same input data has at least the following differences: for the environmental data collected when durians are stored in the refrigerating chamber 102, the classification result predicted by model B is extremely unpleasant. Compared with the unpleasant classification result output by model A, based on the extremely unpleasant classification result output by model B, the control module 12 can control the deodorizing module 13 to operate at a second power, and the second power is greater than the first power. Thus, the preset machine learning model is iteratively updated according to user feedback, so that in the scenario where durians are stored in the refrigerating chamber 102, the deodorizing module 13 can actively operate with greater deodorizing ability to better remove the odor in the refrigerating chamber 102 and improve the user's satisfaction with the smell when opening the refrigerating chamber 102 next time.

[0148] In some embodiments, the machine learning algorithms adopted by model A and model B can be the same, and the difference lies in the adjustment of model parameters.

[0149] Figure 4 It is a schematic diagram of the principle of an odor detection system for a household appliance according to an embodiment of the present invention.

[0150] Specifically, referring to Figure 4 , the odor detection system of the household appliance described in this embodiment may include: a household appliance 1 (as shown in Figure 3 ), including a body 10 and a control module 12, and the control module 12 is used to execute the methods shown in the above Figure 1 and Figure 2 ; a server 2, which communicates with the control module 12, and the server 2 is used to synchronize the preset machine learning model to the control module 12.

[0151] Specifically, the preset machine learning model synchronized by the server 2 to the control module 12 can be the preset machine learning model after the most recent iterative update. For example, if the initial machine learning model is not pre-installed when the household appliance 1 leaves the factory, after the household appliance 1 is powered on for the first time, it connects to the server 2 through the communication module 14 and obtains the initial machine learning model from the server 2. Then, as the user feedback data accumulates, the initial machine learning model is continuously iteratively updated. Each time a new version is updated, the server 2 synchronizes the updated preset machine learning model to the control module 12 of the household appliance 1.

[0152] In some embodiments, the server 2 can be integrated into the household appliance 1 to be specifically used for dynamically updating the preset machine learning model for the household appliance 1.

[0153] In some embodiments, the server 2 can be, for example, a background server, which is set at the manufacturer or designer of the household appliance 2. Alternatively, the server 2 can be, for example, the cloud. A single server 2 can communicate with multiple household appliances 1. For each household appliance 1, the server 2 receives the environmental data and the corresponding feedback data of the household appliance 1 to dynamically update the preset machine learning model for the household appliance 1 specifically.

[0154] Thus, the server 2 can be used to store historical data, including the environmental data and the corresponding feedback data collected historically, and retrain the preset machine learning model based on the historical data, so that the current odor attributes predicted by the household appliance 1 based on the updated preset machine learning model are more in line with the user's preferences. Further, when the server 2 is externally disposed to the household appliance 1, it is beneficial to reduce the number of components in the household appliance 1 and lower the cost.

[0155] In a specific implementation, the server 2 can be used to iteratively update the preset machine learning model based on the feedback data and synchronize the updated preset machine learning model to the control module 12. Thus, with the support of the high computing power of the server 2, the iterative response speed of the preset machine learning model can be further improved.

[0156] For example, the server 2 can obtain the feedback data from the household appliance 1 or the user's terminal device, and obtain the corresponding environmental data from the household appliance 1. Further, based on the feedback data and the corresponding environmental data accumulated over a period of time, the server 2 can retrain the preset machine learning model to achieve iterative update of the model.

[0157] In a specific implementation, the system described in this implementation scheme can further include: a communication module 14, which is set on the body 10, and the control module 12 communicates with the server 2 through the communication module 14.

[0158] For example, the communication module 14 may include a WI-FI module, and the control module 12 remotely uploads the collected data (e.g., environmental data and feedback data) to the server 2 through the WI-FI module.

[0159] Furthermore, the WI-FI module can also be used to upgrade the preset machine learning model via Over-The-Air (OTA) technology. For example, an OTA platform can be built at the server 2, and the updated preset machine learning model is upgraded via OTA through the WI-FI module.

[0160] Thus, the household appliance 1 establishes a communication connection with the outside world through the communication module 14 to achieve remote update of the preset machine learning model. The outside world can be, for example, the server 2.

[0161] In a specific implementation, the system described in this embodiment may further include a display and / or input unit 15, which is disposed on the main body 12 and communicates with the control module 12. The display and / or input unit 15 is used to receive feedback data. Thus, the user can submit feedback data locally on the household appliance 1, and the household appliance 1 aggregates the corresponding environmental data and the feedback data to the server 2 side, so that the server 2 iteratively updates the preset machine learning model based on these data.

[0162] In a typical application scenario, in combination with Figure 3 and Figure 4 , taking the household appliance 1 as a refrigerator as an example, the control module 12 of the refrigerator can execute the methods of the above Figure 1 and Figure 2 shown embodiments. The refrigerator may include a storage module to store the preset machine learning model.

[0163] An odor detector 11 (e.g., a VOC sensor) may be provided in the refrigerating chamber 102 of the refrigerator, and the VOC sensor periodically collects gas data in the refrigerating chamber 102.

[0164] In response to receiving the gas data, the control module 12 uploads the obtained gas data and its collection time to the server 2. And the control module 12 calls the preset machine learning model stored in the storage module to predict the current odor attribute in the refrigerating chamber 102 based on the obtained gas data.

[0165] Assuming that the classification result predicted by the preset machine learning model is neutral, the control module 12 does not trigger the deodorization module 13 to work.

[0166] During the regular execution of the above gas data collection / upload operation and the control of the operation of the deodorization module 13 based on the model prediction result, the control module 12 also periodically sends prompt messages on the display and / or input unit 15 and the user's terminal device.

[0167] If the user opens the refrigerating chamber 102 and finds that the odor inside the refrigerating chamber 102 is very strong, an opinion feedback form is filled out through an application (Application, abbreviated as APP) installed on the terminal device, and the opinion feedback form can be directly transmitted to the server 2 to generate feedback data. In response to receiving the feedback data, the server 2 searches for the gas data with the closest acquisition time according to the generation time of the feedback data, and stores these data correspondingly first.

[0168] When the data volume of the feedback data and the corresponding gas data obtained from the user has accumulated for a period of time, the server 2 retrains the preset machine learning model based on the data accumulated during this period.

[0169] Specifically, these data can be preprocessed first. Then, the preprocessed gas data and the corresponding environmental data are used as the user data set, and together with the basic data set, they are used as input data to train the preset machine learning model.

[0170] If the verification set passes the verification, the server 2 synchronously updates the updated preset machine learning model to the refrigerator via the communication module 14 of the refrigerator through the OTA platform.

[0171] If the verification set fails to pass the verification, the model is retrained, or, after waiting for a period of time to accumulate more data, the model is retrained.

[0172] In a variant, the VOC sensor can include an interaction module, and the user can upload feedback data through the interaction module.

[0173] In a variant, the feedback data can be uploaded to the server 2 through the display and / or input module 15 of the refrigerator via the communication module 14.

[0174] Although the specific implementation schemes have been described above, these implementation schemes are not intended to limit the scope of the present disclosure, even in the case of describing a single implementation scheme only with respect to specific features. The feature examples provided in the present disclosure are intended to be illustrative rather than restrictive, unless otherwise stated. In specific implementations, the technical features of one or more dependent claims can be combined with the technical features of the independent claim, and the technical features from the corresponding independent claims can be combined in any appropriate manner rather than only through the specific combinations listed in the claims.

[0175] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention should be subject to the scope defined by the claims.

Claims

1. An odor detection method for a household appliance, characterized in that, Including: Obtaining environmental data in the chamber of the household appliance, where the environmental data includes gas data; Inputting the environmental data into a preset machine learning model and obtaining a classification result, where the preset machine learning model is used to predict the current odor attribute in the chamber based on the environmental data, and the preset machine learning model is iteratively updated based on feedback data, and the feedback data is obtained from a user associated with the household appliance and is at least used to characterize the prediction satisfaction for the classification result.

2. The method according to claim 1, wherein The process of iteratively updating the preset machine learning model based on the feedback data includes: Receiving the feedback data; Determining the corresponding environmental data according to the generation time of the feedback data; Constructing a training set and a validation set at least based on the feedback data and the corresponding environmental data received within a period of time; Training the preset machine learning model based on the training set to obtain an updated preset machine learning model; Validating the updated preset machine learning model based on the validation set.

3. The method according to claim 2, wherein The validation set further includes a basic data set, where the basic data set includes standard environmental data and corresponding standard classification results.

4. The method according to claim 2, characterized in that, The environmental data is collected periodically, and determining the corresponding environmental data according to the generation time of the feedback data includes: Determining the environmental data with the collection time closest to the generation time of the feedback data as the corresponding environmental data.

5. The method according to claim 1, wherein The preset machine learning model is iteratively updated periodically as the feedback data accumulates.

6. The method according to claim 1, wherein The preset machine learning model is iteratively updated based on the feedback data on the basis of an initial machine learning model, where the initial machine learning model is trained based on a basic data set, and the basic data set includes standard environmental data and corresponding standard classification results.

7. The method according to claim 6, characterized in that The initial machine learning model includes multiple sub-models, and different sub-models are associated with different odor preferences. The preset machine learning model being iteratively updated based on the feedback data on the basis of the initial machine learning model includes: Obtaining the odor preference of the user for at least one odor; Determining the corresponding sub-model according to the obtained odor preference; On the basis of the determined sub-model, iteratively updating based on the feedback data to obtain the preset machine learning model.

8. The method according to claim 1, characterized in that it further Including: Regularly sending a prompt message, where the prompt message includes an opinion feedback form; Receiving the opinion feedback form and generating the feedback data based on the opinion feedback form.

9. The method according to claim 8, wherein The prompt message and / or the opinion feedback form are transmitted through the display and / or input unit of the household appliance, and / or, the prompt message and / or the opinion feedback form are transmitted through a terminal device associated with the household appliance.

10. A household appliance, characterized in that, Including: An ontology (10) defining a chamber (101); An odor detector (11) disposed in the chamber (101) to collect environmental data in the chamber (101); A control module (12) is provided in the body (10). The control module (12) communicates with the odor detector (11) to receive the environmental data. The control module (12) is configured to execute the method according to any one of claims 1 to 9 above, and generate a control instruction based on the current odor attribute in the chamber (101). An odor removal module (13) is provided in the chamber (101). In response to receiving the control instruction, the odor removal module (13) adjusts the operating parameters according to the control instruction.

11. The household appliance according to claim 10, characterized in that, The household appliance is selected from: a refrigeration appliance, a washing machine, a dishwasher, and an oven.

12. The household appliance according to claim 10, wherein, Further comprising: A communication module (14) is provided in the body (10). The control module (12) receives feedback data through the communication module (14); and / or, A display and / or input unit (15) is provided in the body (10). The control module (12) receives feedback data through the display and / or input unit (15).

13. An odor detection system for a household appliance, characterized in that, Comprising: A household appliance, including a body (10) and a control module (12). The control module (12) is configured to execute the method according to any one of claims 1 to 9 above; A server (2) communicates with the control module (12). The server (2) is configured to synchronize the preset machine learning model to the control module (12).

14. The system according to claim 13, wherein The server (2) is configured to iteratively update the preset machine learning model based on the feedback data, and synchronize the updated preset machine learning model to the control module (12).

15. The system according to claim 13, wherein Further comprising: A communication module (14) is provided in the body (10). The control module (12) communicates with the server (2) through the communication module (14).

16. The system according to claim 13, wherein Further comprising: A display and / or input unit (15) is provided in the body (10) and communicates with the control module (12). The display and / or input unit (15) is configured to receive the feedback data.

Images