A method, system and readable medium for identifying electronic cigarette types based on deep learning

Abstract

The present invention relates to the technical field of electronic cigarette type identification, and more specifically, to a method, system, and readable medium for electronic cigarette type identification based on deep learning. The method comprises the following steps: Step 1: training to generate a type training model; Step 2: obtaining the formaldehyde content value in the electronic cigarette aerosol data, bringing it into the type discrimination model, and obtaining a discrimination coefficient and a discrimination result; Step 3: if there is new electronic cigarette aerosol data, bringing the discrimination coefficient into the type discrimination model to generate a new type discrimination model, and proceeding to Step 4; if not, proceeding to Step 4; Step 4: stopping training. By establishing a neural network model and introducing a process for obtaining the discrimination coefficient, the model maintains its recognition ability for old task data while processing new data, enabling the model to maintain its prediction results for old task data, preventing the model from biasedly predicting new task data, and reducing the impact of counterfeit data on samples from the same batch.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic cigarette type identification, and in particular to a method, system, and readable medium for electronic cigarette type identification based on deep learning. Background Art

[0002] E-cigarettes have the same appearance and similar taste as cigarettes. They are mainly used to simulate the feeling of smoking without affecting health, so as to help people quit smoking or provide smoking. Nowadays, there are many types of e-cigarettes, and many three-no e-cigarettes are also circulating on the market. The identification of e-cigarettes mainly lies in the different additives in the e-liquid components. Directly testing the components in the e-liquid requires a lot of time, financial resources and energy. Nowadays, most of the components of the aerosol formed after the e-cigarette oil is burned are tested for component concentration. However, due to different ambient temperatures, there are differences in the concentration of aerosol components, and e-cigarettes are often sold mixed with counterfeits in order to avoid sampling inspections. Therefore, it is necessary to establish a neural network model to test the data and improve the recognition performance of the component concentration of e-cigarettes in the same batch. Summary of the Invention

[0003] The object of the present invention is to provide a method, system and readable medium for identifying electronic cigarette types based on deep learning, which are used to solve the above-mentioned technical problems.

[0004] The embodiments of the present invention are achieved through the following technical solutions:

[0005] A method for identifying electronic cigarette types based on deep learning, the method comprising the following steps:

[0006] Step 1: Train and generate category discrimination model;

[0007] Step 2: Obtain the formaldehyde content value in the electronic cigarette aerosol data, bring it into the type discrimination model, and obtain the discrimination coefficient and discrimination result;

[0008] Step 3: If there is new e-cigarette aerosol data, the discrimination coefficient is brought into the category discrimination model to generate a new category discrimination model, and the process returns to step 2. If there is no new e-cigarette aerosol data, the process proceeds to step 4.

[0009] Step 4: Stop training.

[0010] Optionally, the process of the type discrimination model discriminating the aerosol formaldehyde content value specifically includes the following steps:

[0011] Step 21: Obtain the formaldehyde content value in the aerosol as formaldehyde concentration data and obtain environmental data;

[0012] Step 22: Perform environmental compensation on the formaldehyde concentration data according to the environmental data, and output the concentration data after environmental compensation as the final value data;

[0013] Step 23: And obtain the judgment result based on the final value data.

[0014] Optionally, the formaldehyde concentration data includes first concentration data and second concentration data, and the environmental data includes environmental humidity data.

[0015] Optionally, the environmental compensation process includes the following steps:

[0016] Step 221: When the absolute value of the difference between the first concentration data and the second concentration data is greater than 0, gradually increasing the ambient humidity by equal amounts to obtain multiple sets of concentration signals consisting of the first concentration data and the second concentration data;

[0017] Step 222: When the absolute value of the difference between the first concentration data and the second concentration data is 0, output the current ambient humidity value and the first concentration data or the second concentration data as final value data.

[0018] Optionally, in step 2, the process of obtaining the discriminant coefficient is specifically as follows:

[0019] Step 24: Establish a task set, wherein the task set includes a number of discrimination results and final value data corresponding to the discrimination results, and the final value data serves as task data;

[0020] Step 25: Train the task set based on the current category discrimination model to obtain a new task dataset and a new category discrimination model;

[0021] Step 26: Use the feature extractor to obtain the features of all task data in the new task dataset, calculate the feature mean of all sample data in the task set, and calculate the Euclidean distance between the features of multiple sample data and the feature mean;

[0022] Step 27: Retain several task data corresponding to the sample data feature with the smallest Euclidean distance, and use all the retained task data as the discrimination coefficient.

[0023] Optionally, there is a certain distance between the environmental data collection position and the electronic cigarette aerosol data collection position.

[0024] Optionally, the first concentration data is collected at a constant temperature of 320°C, and the second concentration data is collected at a constant temperature of 35°C.

[0025] A deep learning-based e-cigarette type recognition system includes the following:

[0026] The first sensor is used to collect formaldehyde content values in electronic cigarette aerosol data as first concentration data at a constant temperature of 320° C.;

[0027] The second sensor is used to collect the formaldehyde content value in the electronic cigarette aerosol data at a constant temperature of 35° C. as the second concentration data;

[0028] Humidity sensor, used to detect environmental humidity data and output the environmental humidity value;

[0029] The central controller is used to receive the detection data of the first sensor, the second sensor and the humidity sensor, and substitute the data into the type discrimination model to obtain the discrimination result.

[0030] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for identifying electronic cigarette types based on deep learning.

[0031] An electronic device includes a memory and one or more programs, wherein the one or more programs are stored in the memory and is configured so that one or more processors execute operating instructions contained in the one or more programs for performing a method for identifying electronic cigarette types based on deep learning.

[0032] The technical solutions of the embodiments of the present invention have at least the following advantages and beneficial effects:

[0033] The present invention has a reasonable design and a simple structure. By eliminating the influence of ambient humidity on formaldehyde concentration detection, the detection accuracy of formaldehyde content in electronic cigarette aerosol data is improved. Furthermore, by establishing a neural network model and introducing a process for obtaining a discriminant coefficient, the model maintains the ability to recognize old task data while processing new data. The model can maintain the prediction results of old task data, thereby alleviating the forgetting of old task data, preventing the model from biasedly predicting new task data, and reducing the influence of counterfeit data on samples from the same batch. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A flowchart of a method for identifying electronic cigarette types based on deep learning provided by the present invention;

[0035] Figure 2 A schematic diagram of the process of using the type discrimination model provided by the present invention to discriminate the aerosol formaldehyde content value;

[0036] Figure 3 A schematic diagram of the process of obtaining the discrimination coefficient provided by the present invention. DETAILED DESCRIPTION

[0037] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of 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. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0038] like Figure 1 As shown, the present invention provides one embodiment: a method, system and readable medium for identifying electronic cigarette types based on deep learning, comprising the following steps:

[0039] Step 1: Train and generate category discrimination model;

[0040] Step 2: Obtain the formaldehyde content value in the electronic cigarette aerosol data, bring it into the type discrimination model, and obtain the discrimination coefficient and discrimination result;

[0041] Step 3: If there is new e-cigarette aerosol data, the discrimination coefficient is brought into the category discrimination model to generate a new category discrimination model, and the process returns to step 2. If there is no new e-cigarette aerosol data, the process proceeds to step 4.

[0042] Step 4: Stop training.

[0043] like Figure 2 As shown, in this embodiment, the process of the type discrimination model discriminating the aerosol formaldehyde content value specifically includes the following steps:

[0044] Step 21: Obtain the formaldehyde content value in the aerosol as formaldehyde concentration data and obtain environmental data;

[0045] Step 22: Perform environmental compensation on the formaldehyde concentration data according to the environmental data, and output the concentration data after environmental compensation as the final value data;

[0046] Step 23: And obtain the judgment result based on the final value data.

[0047] In this embodiment, the formaldehyde concentration data includes first concentration data and second concentration data, and the environmental data includes environmental humidity data.

[0048] Common methods for detecting formaldehyde concentration include spectrophotometry, chromatography, fluorescence, and polarography. However, formaldehyde detection devices manufactured using these methods are often costly and complex to operate, hindering their widespread application in applications such as air purifiers and HVAC systems. Currently, formaldehyde concentration is typically detected using electrochemical principles. This principle involves a redox reaction between oxygen and formaldehyde at an electrode, generating a signal proportional to the formaldehyde concentration. The formaldehyde concentration is then calculated based on the signal. While this detection method offers a fast response, it has limitations in terms of stability and lifespan. Because formaldehyde gas diffuses at varying humidities, the measured formaldehyde concentration may deviate from the actual formaldehyde concentration. Therefore, a humidity sensor and compensation algorithm are employed to compensate the formaldehyde sensor for temperature and / or humidity.

[0049] In this embodiment, the environmental compensation process includes the following steps:

[0050] Step 221: When the absolute value of the difference between the first concentration data and the second concentration data is greater than 0, gradually increasing the ambient humidity by equal amounts to obtain multiple sets of concentration signals consisting of the first concentration data and the second concentration data;

[0051] Step 222: When the absolute value of the difference between the first concentration data and the second concentration data is 0, output the current ambient humidity value and the first concentration data or the second concentration data as final value data.

[0052] At a certain humidity value, when the absolute value of the difference between the first concentration data and the second concentration data is 0, it indicates that the diffusion rate of formaldehyde is the same at this humidity value.

[0053] like Figure 3 As shown, in this embodiment, in step 2, the process of obtaining the discrimination coefficient is specifically as follows:

[0054] Step 24: Establish a task set, wherein the task set includes a number of discrimination results and final value data corresponding to the discrimination results, and the final value data serves as task data;

[0055] Step 25: Train the task set based on the current category discrimination model to obtain a new task dataset and a new category discrimination model;

[0056] Step 26: Use the feature extractor to obtain the features of all task data in the new task dataset, calculate the feature mean of all sample data in the task set, and calculate the Euclidean distance between the features of multiple sample data and the feature mean;

[0057] Step 27: Retain several task data corresponding to the sample data feature with the smallest Euclidean distance, and use all the retained task data as the discrimination coefficient.

[0058] By introducing the process of obtaining the discriminant coefficient, the model can maintain the ability to recognize old task data while processing new data. It can also enable the model to maintain the prediction results of old task data, thereby alleviating the forgetting of old task data, preventing the model from being biased in predicting new task data, and reducing the impact of fake data on samples in the same batch.

[0059] In this embodiment, there is a certain distance between the environmental data collection position and the electronic cigarette aerosol data collection position.

[0060] In this embodiment, the first concentration data is collected at a constant temperature of 320° C., and the second concentration data is collected at a constant temperature of 35° C.

[0061] like Figure 2 As shown, a deep learning-based e-cigarette type recognition system includes the following:

[0062] The first sensor is used to collect formaldehyde content values in electronic cigarette aerosol data as first concentration data at a constant temperature of 320° C.;

[0063] The second sensor is used to collect the formaldehyde content value in the electronic cigarette aerosol data at a constant temperature of 35° C. as the second concentration data;

[0064] Humidity sensor, used to detect environmental humidity data and output the environmental humidity value;

[0065] The central controller is used to receive the detection data of the first sensor, the second sensor and the humidity sensor, and substitute the data into the type discrimination model to obtain the discrimination result.

[0066] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for identifying electronic cigarette types based on deep learning.

[0067] An electronic device includes a memory and one or more programs, wherein the one or more programs are stored in the memory and configured to execute, by one or more processors, operating instructions contained in the one or more programs for performing an electronic cigarette type identification method based on deep learning

[0068] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A method for identifying electronic cigarette types based on deep learning, characterized in that: The steps of the method include: Step 1: Train and generate category discrimination model; Step 2: Obtain the formaldehyde content value in the electronic cigarette aerosol data, bring it into the type discrimination model, and obtain the discrimination coefficient and discrimination result; Step 3: If there is new e-cigarette aerosol data, the discrimination coefficient is brought into the category discrimination model to generate a new category discrimination model, and then proceed to step 4; if there is no new e-cigarette aerosol data, then proceed to step 4; Step 4: Stop training; The process of the type discrimination model discriminating the aerosol formaldehyde content value specifically includes the following steps: Step 21: Obtain the formaldehyde content value in the aerosol as formaldehyde concentration data and obtain environmental data; Step 22: Perform environmental compensation on the formaldehyde concentration data according to the environmental data, and output the concentration data after environmental compensation as the final value data; The formaldehyde concentration data includes first concentration data and second concentration data, and the environmental data includes environmental humidity data; The environmental compensation process includes the following steps: Step 221: When the absolute value of the difference between the first concentration data and the second concentration data is greater than 0, gradually increasing the ambient humidity by equal amounts to obtain multiple sets of concentration signals consisting of the first concentration data and the second concentration data; Step 222: When the absolute value of the difference between the first concentration data and the second concentration data is 0, output the current ambient humidity value and the first concentration data or the second concentration data as final value data; In step 2, the process of obtaining the discrimination coefficient and the discrimination result is specifically as follows: Step 24: Establish a task set, wherein the task set includes a number of final value data, and the final value data serves as task data; Step 25: Train the task set based on the current category discrimination model to obtain a new task dataset and a new category discrimination model; Step 26: Use the feature extractor to obtain the features of all task data in the new task dataset, calculate the feature mean of all sample data in the task set, and calculate the Euclidean distance between the features of multiple sample data and the feature mean; Step 27: If there is a sample data feature whose Euclidean distance is less than the threshold, then the task data corresponding to the sample data feature and all the retained task data are used as the discrimination coefficient, and the discrimination result is output as authentic; if not, the discrimination coefficient is zero, and the discrimination result is output as fake.

2. The method for identifying electronic cigarette types based on deep learning according to claim 1, characterized in that: There is a certain distance between the environmental data collection position and the electronic cigarette aerosol data collection position.

3. The method for identifying electronic cigarette types based on deep learning according to claim 1, characterized in that: The first concentration data is collected at a constant temperature of 320° C., and the second concentration data is collected at a constant temperature of 35° C.

4. An electronic cigarette type identification system based on deep learning, characterized in that: A method for identifying electronic cigarette types based on deep learning according to any one of claims 1 to 3, comprising the following contents: The first sensor is used to collect formaldehyde content values in electronic cigarette aerosol data as first concentration data at a constant temperature of 320° C.; The second sensor is used to collect the formaldehyde content value in the electronic cigarette aerosol data at a constant temperature of 35° C. as the second concentration data; Humidity sensor, used to detect environmental humidity data and output the environmental humidity value; The central controller is used to receive the detection data of the first sensor, the second sensor and the humidity sensor, and substitute the data into the type discrimination model to obtain the discrimination result.

5. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for identifying electronic cigarette types based on deep learning according to any one of claims 1 to 3.

6. An electronic device comprising a memory and one or more programs, wherein the one or more programs are stored in the memory and configured to have one or more processors execute the operating instructions contained in the one or more programs for performing the deep learning-based electronic cigarette type identification method according to any one of claims 1 to 3.

Images