Atomization equipment control method and device, controller, equipment and readable storage medium

Abstract

The invention discloses an atomization equipment control method and device, a controller, equipment and a readable storage medium. The method comprises the steps that detection data collected by a sensor in atomization equipment is obtained; preprocessing the detection data according to historical suction data and current temperature to obtain denoised detection data; the detection data comprises at least two types of detection data; inputting the at least two types of detection data into a pre-trained behavior recognition model to obtain a behavior recognition result recognized by the behavior recognition model; the behavior recognition result is used for representing the use state of the atomization equipment; under the condition that the behavior recognition result represents that the atomization equipment is in the suction state, the atomization equipment is controlled to conduct atomization operation. The method and the device can be used for solving the problem of misjudgment of atomization equipment caused by external interference factors in related technologies.

Description

Technical Field

[0001] The present application relates to the field of electronic atomization technology, and in particular to an atomization device control method, device, controller, device and readable storage medium. Background Art

[0002] Atomizer devices usually rely on a single air pressure sensor to determine whether the device is in the suction state. However, in related technologies, this judgment method is easily interfered with by external factors and leads to misjudgment. For example, transportation vibration, falling, changes in ambient air flow, etc. will affect the air pressure sensor, which in turn causes the atomizer device to make wrong judgments, resulting in incorrect atomization and oil supply of the atomizer device, seriously affecting the user experience and the service life of the device. Summary of the Invention

[0003] The present application provides an atomization device control method, device, controller, device and readable storage medium, which are used to solve the problem of erroneous judgment of the atomization device due to external interference factors in the related art.

[0004] The present application provides a method for controlling an atomizing device, the method comprising:

[0005] Obtain detection data collected by sensors in atomization equipment;

[0006] Preprocessing the detection data according to historical suction data and current temperature to obtain denoised detection data; the detection data includes at least two types of detection data;

[0007] Inputting at least two types of detection data into a pre-trained behavior recognition model to obtain a behavior recognition result identified by the behavior recognition model; the behavior recognition result is used to characterize the usage status of the atomization device;

[0008] When the behavior recognition result indicates that the atomization device is in a suction state, the atomization device is controlled to perform an atomization operation.

[0009] In some embodiments, preprocessing the detection data according to historical suction data and current temperature to obtain the denoised detection data includes:

[0010] Adjusting a preset suction threshold according to the historical suction data and the current temperature data to obtain an adjusted target suction threshold;

[0011] Inputting at least two types of detection data into a pre-trained behavior recognition model to obtain a behavior recognition result identified by the behavior recognition model includes:

[0012] When the current suction data is greater than or equal to the target suction threshold and the first duration is within a first preset duration range, inputting at least two types of detection data into the pre-trained behavior recognition model to obtain a behavior recognition result identified by the behavior recognition model;

[0013] The first duration is the duration during which the suction data is greater than or equal to the target suction threshold.

[0014] In some embodiments, inputting at least two types of detection data into a pre-trained behavior recognition model to obtain a behavior recognition result identified by the behavior recognition model includes:

[0015] Inputting at least two types of detection data into the behavior recognition model to obtain index parameters output by the behavior recognition model; the index parameters respectively correspond to each type of detection data;

[0016] According to the indicator parameters and the indicator conditions corresponding to the indicator parameters, the indicator parameters that meet the indicator conditions are used as target indicator parameters;

[0017] When the ratio of the number of the target index parameters to the number of the index parameters is greater than or equal to a preset ratio, a behavior recognition result of the atomization device being in a puffing state is output.

[0018] In some embodiments, the behavior recognition model includes at least two recognition sub-models, and the at least two recognition sub-models correspond to at least two types of detection data respectively;

[0019] Inputting at least two types of detection data into the behavior recognition model to obtain indicator parameters output by the behavior recognition model includes:

[0020] selecting, according to a type of target detection data, a target recognition sub-model corresponding to the target detection data from at least two recognition sub-models, wherein the target detection data is one of the at least two types of detection data;

[0021] The target detection data is input into the corresponding target recognition sub-model to obtain the index parameters corresponding to the target detection data output by the target recognition sub-model.

[0022] In some embodiments, the detection data includes at least one of suction data, acceleration data, posture angle data, and temperature data within a preset time period.

[0023] In some embodiments, controlling the atomization device to perform an atomization operation includes:

[0024] Sending a start instruction to the oil supply solenoid valve, wherein the start instruction is used to control the oil supply solenoid valve to be turned on, so as to conduct the atomized matrix to the atomizing core; or,

[0025] Sending a heating instruction to the heating component, wherein the heating instruction is used to control the heating component to heat the atomizer core after a first preset time period; or,

[0026] A wake-up instruction is sent to the power management component, where the wake-up instruction is used to control the power supply component to supply power to the atomizer core after a second preset time period.

[0027] In some embodiments, the detection data includes suction data collected by the air pressure sensor within a preset time period;

[0028] The method further comprises:

[0029] When the current suction data is greater than or equal to the target suction threshold, and a second duration during which the current suction data is greater than or equal to the target suction threshold is within a second preset duration range, controlling the atomizing device to be in the first state;

[0030] The first state includes the power supply component supplying power to the heating component, so that the heating component heats the atomizer core at a first preset power.

[0031] In some embodiments, the detection data includes acceleration data collected by a three-axis accelerometer within a preset time period and attitude angle data collected by a gyroscope within the preset time period;

[0032] The method further comprises:

[0033] Determining current posture information of the atomizing device according to the acceleration data and the posture angle data;

[0034] When it is determined that the atomization device is in a specified posture according to the posture information, the atomization device is controlled to be in a second state, and the second state includes the power supply component supplying power to the heating component so that the heating component heats the atomization core at a second preset power.

[0035] In one embodiment, an atomization device is provided, comprising:

[0036] An acquisition module is used to obtain detection data collected by sensors in the atomization device;

[0037] a preprocessing module, configured to preprocess the detection data according to historical suction data and current temperature to obtain the detection data after denoising; the detection data includes at least two types of detection data;

[0038] an identification module, configured to input at least two types of detection data into a pre-trained behavior recognition model to obtain a behavior recognition result identified by the behavior recognition model, wherein the behavior recognition result is used to characterize the usage status of the atomization device;

[0039] The first control module is configured to control the atomization device to perform an atomization operation when the behavior recognition result indicates that the atomization device is in a puffing state.

[0040] In some embodiments, the pre-processing module is configured to adjust a preset suction threshold according to the historical suction data and the current temperature data to obtain an adjusted target suction threshold;

[0041] The recognition module is configured to input the at least two types of detection data into the pre-trained behavior recognition model to obtain a behavior recognition result identified by the behavior recognition model when the current suction data is greater than or equal to the target suction threshold and the first duration is within a first preset duration range;

[0042] The first duration is the duration during which the suction data is greater than or equal to the target suction threshold.

[0043] In some embodiments, the identification module includes:

[0044] A first identification submodule is configured to input at least two types of detection data into the behavior recognition model to obtain indicator parameters output by the behavior recognition model; the indicator parameters respectively correspond to each type of detection data;

[0045] A first determining submodule is configured to, based on the indicator parameter and the indicator condition corresponding to the indicator parameter, select the indicator parameter that meets the indicator condition as the target indicator parameter;

[0046] The second determining submodule is configured to output a behavior recognition result of the atomizing device being in a puffing state when the ratio of the number of the target index parameters to the number of the index parameters is greater than or equal to a preset ratio.

[0047] In some embodiments, the behavior recognition model includes at least two recognition sub-models, and the at least two recognition sub-models correspond to at least two types of detection data respectively; the first recognition sub-module is used to select a target recognition sub-model corresponding to the target detection data from the at least two recognition sub-models according to the type of the target detection data, and the target detection data is one of the at least two types of detection data;

[0048] The target detection data is input into the corresponding target recognition sub-model to obtain the index parameters corresponding to the target detection data output by the target recognition sub-model.

[0049] In some embodiments, the detection data includes at least one of suction data, acceleration data, posture angle data, and temperature data within a preset time period.

[0050] In some embodiments, the first control module includes:

[0051] A first sending submodule is configured to send a start instruction to the oil supply solenoid valve, wherein the start instruction is configured to control the oil supply solenoid valve to be turned on, so as to conduct the atomized matrix to the atomizing core; or

[0052] The second sending submodule is used to send a heating instruction to the heating component, wherein the heating instruction is used to control the heating component to heat the atomizer core after a first preset time period; or

[0053] The third sending submodule is used to send a wake-up instruction to the power management component, where the wake-up instruction is used to control the power supply component to supply power to the atomizer core after a second preset time period.

[0054] In some embodiments, the detection data includes suction data collected by the air pressure sensor within a preset time period; the device further includes:

[0055] a second control module, configured to control the atomizing device to be in a first state when the current suction data is greater than or equal to the target suction threshold and a second duration during which the current suction data is greater than or equal to the target suction threshold is within a second preset duration range;

[0056] The first state includes the power supply component supplying power to the heating component, so that the heating component heats the atomizer core at a first preset power.

[0057] In some embodiments, the detection data includes acceleration data collected by a three-axis accelerometer within a preset time period and attitude angle data collected by a gyroscope within the preset time period; the device further includes:

[0058] a determination module, configured to determine current posture information of the atomization device based on the acceleration data and the posture angle data;

[0059] The third control module is used to control the atomization device to be in a second state when it is determined that the atomization device is in a specified posture according to the posture information. The second state includes the power supply component supplying power to the heating component so that the heating component heats the atomization core at a second preset power.

[0060] In one embodiment, a controller is provided, which is configured to implement the method described in any of the above embodiments when executed.

[0061] In one embodiment, an atomizing device is provided, wherein the atomizing device includes a sensor, a heating component, and a controller as described in the above embodiment.

[0062] The sensor is used to collect detection data;

[0063] The heating component is used to generate heat so that the atomizing device can perform an atomizing operation.

[0064] In one embodiment, a readable storage medium is provided. When instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the method described in any of the above embodiments.

[0065] In this embodiment, detection data collected by sensors in the atomizing device is obtained; the detection data is preprocessed according to historical suction data and current temperature to obtain the denoised detection data; the detection data includes at least two types of detection data; the at least two types of detection data are input into a pre-trained behavior recognition model to obtain a behavior recognition result identified by the behavior recognition model; the behavior recognition result is used to characterize the usage status of the atomizing device; when the behavior recognition result characterizes that the atomizing device is in a suction state, the atomizing device is controlled to perform an atomizing operation. In this way, multi-dimensional data collected by multiple sensors can be deployed, and the behavior recognition model can be used to cross-validate the data. The pre-trained model can be used to learn the real feature combination of the atomizing device in the suction state, and the interference data of the suction behavior can be filtered more accurately, which can improve the recognition accuracy and reduce the probability of the atomizing device being falsely triggered. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 This is a flow chart of a method for controlling an atomization device provided in an embodiment of the present application;

[0067] Figure 2 This is a flow chart of another atomization device control method provided in an embodiment of the present application;

[0068] Figure 3 This is a block diagram of an atomization device control device provided in an embodiment of the present application;

[0069] Figure 4 FIG. 1 is a structural block diagram of an electronic device in an embodiment. DETAILED DESCRIPTION

[0070] The present invention will be further described in detail below by means of specific embodiments in conjunction with the accompanying drawings. Similar elements in different embodiments are numbered with associated similar elements. In the following embodiments, many detailed descriptions are provided to enable the present application to be better understood. However, those skilled in the art will readily appreciate that some of the features may be omitted in different circumstances, or may be replaced by other elements, materials, or methods. In some cases, some operations related to the present application are not shown or described in the specification. This is to avoid the core portion of the present application being overwhelmed by excessive descriptions, and for those skilled in the art, it is not necessary to describe these related operations in detail. They will fully understand the related operations based on the description in the specification and the general technical knowledge in the art.

[0071] In addition, the features, operations, or characteristics described in the specification may be combined in any appropriate manner to form various embodiments. Furthermore, the steps or actions in the method description may be reordered or adjusted in a manner readily apparent to those skilled in the art. Therefore, the various sequences in the specification and drawings are provided solely for the purpose of clearly describing a particular embodiment and are not intended to be mandatory, unless otherwise specified.

[0072] The serial numbers assigned to components herein, such as "first," "second," etc., are used solely to distinguish the objects being described and do not convey any sequential or technical meaning. References to "connection" and "coupling" herein, unless otherwise specified, include both direct and indirect connections (couplings).

[0073] Before introducing the system security startup method, device, electronic device and storage medium provided by this application, the application scenarios involved in each embodiment of this application are first introduced. This application can be applied to the scenario of controlling the atomization device, and the atomization device control method provided in the embodiment of this application can be applied to the controller of the atomization device. Among them, the atomization device can be used to heat or vibrate the atomization matrix (liquid substance) into an aerosol (mist particles).

[0074] At present, traditional atomizing equipment usually relies on a single air pressure sensor to judge the suction status of the equipment. When the user inhales, the air pressure inside the atomizing equipment changes. The sensor detects the drop in air pressure and transmits a signal to the equipment control device. The control device starts atomization and oil supply accordingly. However, the air pressure sensor is very sensitive and external factors can easily interfere with its detection results. For example, during transportation, vibrations such as vehicle bumps and handling collisions will cause the air inside the equipment to shake and the air pressure to fluctuate, which will be mistakenly judged as the user inhaling; when the equipment falls accidentally, the instantaneous impact will cause a sudden change in air pressure, which will also trigger an erroneous judgment; in environments such as air-conditioning outlets and near fans, the flowing airflow will change the air pressure around the equipment, misleading the air pressure sensor to send an erroneous signal.

[0075] If the air pressure sensor misjudges, the device will initiate atomization and oil supply without actual puff demand. On the one hand, the atomizer core burns dry without the atomizer matrix soaking in, which not only produces an unpleasant burnt smell and degrades the user experience, but also accelerates the wear and tear of the atomizer core, shortening its service life. On the other hand, continued incorrect oil supply can cause atomizer matrix to accumulate inside the device, increasing the risk of leakage. Leaked e-liquid can corrode the device's internal circuitry, causing more serious malfunctions and even posing a safety hazard.

[0076] In order to solve the above problems, the present application provides a method, device, controller, equipment and readable storage medium for controlling an atomizing device. By deploying multi-dimensional data collected by multiple sensors, the data can be cross-validated using a behavior recognition model. The pre-trained model can learn the actual feature combination of the atomizing device in the puffing state, and more accurately filter the interference data of the puffing behavior, thereby improving the recognition accuracy and reducing the probability of the atomizing device being falsely triggered.

[0077] The method provided in the embodiment of the present application is described in detail below through specific embodiments and their application scenarios in conjunction with the accompanying drawings.

[0078] Figure 1 This is a flow chart of a method for controlling an atomizing device provided in an embodiment of the present application. Figure 1 As shown, the method can be applied to a controller of an atomizing device. The atomizing device can be used to heat or vibrate an atomizing matrix (liquid substance) to break it into aerosol (mist-like particles).

[0079] The method may include the following steps.

[0080] Step 101: Acquire detection data collected by sensors in an atomization device.

[0081] Among them, the atomization device can be provided with multiple sensors for collecting detection data.

[0082] Considering that traditional atomization equipment only relies on a single air pressure sensor to judge the suction status, it is easily interfered by external factors, such as transportation vibration, changes in ambient airflow, etc., which may lead to wrong judgment. Therefore, in this application, multiple sensors are set on the atomization equipment to collect information from multiple angles, complement and verify each other, and reduce the risk of misjudgment.

[0083] The sensor may include at least two types of sensors, and the at least two types of sensors are used to detect detection data of the atomization device in different aspects.

[0084] Specifically, the multiple sensors may include at least two types of sensors, such as an air pressure sensor, an acceleration sensor, an airflow sensor, a temperature sensor, a pressure sensor, and a gyroscope, to detect at least two types of detection data.

[0085] In this step, while the device is in standby mode, the system controller can continuously query the sensor data at a preset query frequency, for example, at a 50Hz frequency (i.e., 50 times per second). This allows the capture of short-term transient signals (such as momentary airflow fluctuations and slight vibrations) through high-frequency sampling, avoiding the omission of critical data due to long sampling intervals.

[0086] To ensure the temporal consistency and relevance of the collected data, each sensor synchronizes its clocks through hardware or is time-shared by the controller, ensuring that all data is precisely aligned in timestamps, accurate to the millisecond level. For example, when a user takes a puff, the air pressure sensor detects a drop in air pressure, while the accelerometer registers a slight hand shake, and the airflow sensor captures a steady inward airflow. These data are generated almost simultaneously, providing a reliable foundation for subsequent analysis.

[0087] Because the raw data collected by the sensors may contain noise or outliers, preprocessing is required. First, digital filtering techniques, such as low-pass filters, are used to remove high-frequency noise, such as high-frequency acceleration signals generated by transportation vibrations. Second, valid data thresholds are set for each sensor based on the device's actual usage scenario and normal operating range. For example, the airflow velocity must be greater than the natural wind speed threshold to avoid misjudgments caused by ambient wind. Finally, statistical methods, such as the Z-score (standard score), are used to detect and mark sudden abnormal data, such as the impact acceleration data generated when the device falls, and treat it as interference signals for processing.

[0088] After preprocessing, the data from each sensor is integrated into a feature vector containing multi-dimensional information, such as [air pressure change rate, mean acceleration, airflow velocity, and grip pressure]. This feature vector contains information about the various states of the device at a given moment, providing a more comprehensive picture of the device's current state and providing rich and valuable input data for accurate judgment using subsequent behavior recognition models.

[0089] Step 102 : pre-processing the detection data according to historical suction data and current temperature to obtain denoised detection data.

[0090] The detection data includes at least two types of detection data.

[0091] Considering that temperature changes can cause changes in the volume of gas inside the device, the detection value of the air pressure sensor (suction data) deviates from the actual user's suction force. For example, in a high temperature environment (such as 40°C), gas expansion may reduce the drop in air pressure under the same suction force, causing the original data to underestimate the actual force; moreover, temperature affects the viscosity of the atomization matrix, indirectly changing the inhalation flow rate, making the suction data mismatched with the actual atomization needs. In addition, the suction range of different users may vary significantly, and the same user's puffing habits may change in different time periods (such as a decrease in suction when fatigued), so dynamic thresholds need to be learned through historical data.

[0092] Therefore, in this step, environmental interference can be corrected by temperature compensation, and further correction can be made based on habit adaptation of historical data. In this way, the original detection data can be converted into detection data with "accurate physical meaning and clear user intention" after the interference data is removed.

[0093] Step 103: Input at least two types of detection data into a pre-trained behavior recognition model to obtain a behavior recognition result identified by the behavior recognition model.

[0094] The behavior recognition result is used to characterize the usage state of the atomizing device, and the usage state may include the atomizing device being in a suction state and the atomizing device not being in a suction state.

[0095] This behavior recognition model can convert the raw data collected by multiple sensors (such as air pressure, acceleration, airflow velocity, etc.) into an accurate judgment of the usage status of the atomizer device (such as whether it is in the puffing state). This method is different from the traditional single sensor threshold judgment. It can comprehensively analyze multi-dimensional information, effectively distinguish between real user operations and external interference, and solve the problem of single air pressure sensor prone to misjudgment.

[0096] In this step, the detection data of at least two types of sensors can be input into the behavior recognition model. The behavior recognition model can automatically extract key indicators from the at least two types of input detection data, and output recognition results based on the extracted key indicators. The behavior recognition results can be used to characterize the usage status of the atomization device, and the usage status can include the atomization device being in a suction state and the atomization device not being in a suction state.

[0097] For example, when the sensor includes an air pressure sensor, the air pressure data inside the atomizing device can be collected through the air pressure sensor. The behavior recognition model can analyze the amplitude index, speed index, and duration index of the air pressure drop based on the multiple collected air pressure data to determine whether it conforms to the air pressure change pattern of normal inhalation.

[0098] For example, if the sensor includes an accelerometer, it can detect the acceleration change index of the atomization device in three-dimensional space (such as the X-axis, Y-axis, and Z-axis). When the user holds the device normally and performs a puff operation, the slight movement of the hand will cause the device to produce a specific acceleration change; however, during transportation, the acceleration change pattern caused by the vibration of the device is different from that during puffing. The accelerometer can record these differences. The behavior recognition model can identify the vibration frequency index, intensity index, and direction change index in the acceleration data based on the multiple acceleration data collected, and distinguish between the regular vibration when holding the hand and puffing and the irregular vibration during transportation.

[0099] And when the sensor includes an airflow sensor, the airflow speed and direction at the entrance of the atomizing device can be detected. When the user performs a real inhalation operation, a stable airflow toward the inside of the device will be generated, and the airflow sensor can accurately sense the characteristics of this airflow; while the speed and direction of interfering airflows such as natural wind in the environment are often unstable. The data from the airflow sensor can distinguish between the real inhalation airflow and the environmental interfering airflow. The behavior recognition model can identify the airflow speed index, direction index, and airflow duration index based on the multiple airflow data collected, and then judge whether it is a stable inward airflow generated by the user's active inhalation based on the stability and continuity of the airflow speed and direction.

[0100] Step 104 : When the behavior recognition result indicates that the atomizing device is in a suction state, control the atomizing device to perform an atomizing operation.

[0101] The above technical solution is adopted to obtain the detection data collected by the sensor in the atomizing device; pre-process the detection data according to the historical suction data and the current temperature to obtain the denoised detection data; the detection data includes at least two types of detection data; the at least two types of detection data are input into the pre-trained behavior recognition model to obtain the behavior recognition result identified by the behavior recognition model; the behavior recognition result is used to characterize the usage status of the atomizing device; when the behavior recognition result characterizes that the atomizing device is in the suction state, the atomizing device is controlled to perform the atomizing operation. In this way, the multi-dimensional data collected by multiple sensors can be deployed, and the behavior recognition model can be used to cross-validate the data. The pre-trained model can learn the real feature combination of the atomizing device in the suction state, and more accurately filter the interference data of the suction behavior, which can improve the recognition accuracy and reduce the probability of the atomizing device being falsely triggered.

[0102] Considering that multiple sensors (such as air pressure, acceleration, airflow, etc.) generate a large amount of raw data when collecting data at high frequencies (such as using 50Hz polling in step 101), and the computational complexity of the behavior recognition model is positively correlated with the amount of input data, if a large amount of raw data is input, the recognition efficiency of the behavior recognition model will be low. In addition, redundant data may also introduce "noise interference", causing the model to deviate from the judgment of key features (such as misjudging environmental vibrations as user operations). Therefore, before using the behavior recognition model to identify the detection data, invalid data can be filtered out through rules or algorithms to retain key information that is strongly related to the target behavior.

[0103] Furthermore, temperature fluctuations affect the physical properties of the gas inside the device (such as thermal expansion and contraction), leading to deviations in the suction data detected by the pressure sensor. For example, at low temperatures, the gas density increases, resulting in a greater drop in pressure at the same suction force, which could cause the fixed threshold to misjudge as "over-suction." Meanwhile, at high temperatures, the gas expands, and even slight fluctuations in ambient airflow can easily trigger the fixed threshold, leading to misjudgments.

[0104] Furthermore, by analyzing users' past suction data (e.g., average suction force and suction force fluctuation range), personalized thresholds can be established. For example, if user A habitually puffs slowly at a speed of 0.2-0.3 kPa / s, historical data can reflect their stable suction force variation trend; while user B occasionally experiences short bursts of intense suction force of 0.5 kPa / s, historical data can record the frequency and intensity of these sudden behaviors.

[0105] Therefore, in some embodiments, preprocessing the detection data based on historical suction data and current temperature to obtain the denoised detection data can be achieved in the following manner.

[0106] Optionally, historical suction data and current suction data inside the atomizing device, as well as current temperature data, may be obtained, and then the preset suction threshold may be adjusted according to the historical suction data and the current temperature data to obtain an adjusted target suction threshold.

[0107] The preset suction threshold refers to a pre-set suction intensity threshold within the atomizer's control logic that determines whether the user triggers atomization. This threshold can be the system's default benchmark, used to distinguish between normal inhalation behavior and non-target behaviors such as accidental inhalation and interference.

[0108] In this embodiment, the threshold range may be adjusted based on the current temperature.

[0109] For example, in a high temperature environment (>30°C), the threshold can be adjusted upward by 0.2kPa to avoid interference from ambient airflow; in a low temperature environment (<10°C), the threshold can be adjusted downward by 0.1kPa to compensate for detection deviations caused by changes in gas density.

[0110] Finally, the target suction threshold can be calculated by combining the statistical characteristics of historical suction data with the current temperature compensation value. For example, if the historical average suction is 0.3kPa and the current temperature compensation value is +0.05kPa, the target threshold is set to 0.3×1.2+0.05=0.41kPa.

[0111] When the target suction threshold is obtained, the behavior recognition result can be obtained based on at least two types of detection data through the behavior recognition model when the current suction data is greater than or equal to the target suction threshold and the first duration is within a first preset duration range.

[0112] The first duration is a duration during which the suction data is greater than or equal to the target suction threshold.

[0113] In a possible implementation, historical temperature data may be first acquired, and then a mapping relationship between historical temperature and historical suction data may be established. For example, for every 10° C. increase in temperature, the suction detection value decreases by 0.1 kPa.

[0114] Then, the current suction data (detected by the air pressure sensor) and temperature data (monitored by the temperature sensor) are continuously obtained. Then, the suction data of the user's recent N times (e.g., N = 50 times) can be extracted from the local storage of the device to calculate statistical features (e.g., mean, standard deviation, quantile).

[0115] After obtaining the above data, the current suction data can be calibrated based on the current temperature and the preset temperature-suction correction formula. For example, if the preset formula is "corrected suction = original suction + 0.01 × (current temperature - 25°C)", when the temperature is 35°C, the original suction value will be increased by 0.1kPa, and then adaptive adjustments can be made based on historical data.

[0116] For example, if the historical suction data fluctuates slightly (standard deviation < 0.05 kPa), it means that the user habits are stable, and the threshold is set to 1.2 times the historical average to reduce misjudgment; if there are high suction peaks in the historical data (such as the frequency of exceeding the preset threshold is > 10%), the threshold is lowered by 10% to ensure that strong suction behavior is responded to in a timely manner.

[0117] Then, when the target suction threshold is obtained, when the adjusted current suction data is greater than or equal to the target suction threshold and the first duration is within the first preset duration range, at least two types of the detection data can be input into the pre-trained behavior recognition model to obtain the behavior recognition result identified by the behavior recognition model.

[0118] The first duration is a duration during which the suction data is greater than or equal to the target suction threshold.

[0119] By adopting the above technical solutions, by adapting to individual usage habits and environmental changes, it is ensured that the device can respond accurately in different scenarios, reducing the phenomenon of "no response" or "false triggering"; avoiding frequent misoperations due to unreasonable thresholds (such as dry burning, overload), and reducing hardware loss; personalized dynamic adjustment strategies can meet users' needs for device intelligence and adaptability, and enhance product differentiation advantages.

[0120] In some embodiments, inputting at least two types of detection data into the pre-trained behavior recognition model to obtain a behavior recognition result identified by the behavior recognition model can be achieved by the following steps:

[0121] In step 1021, at least two types of the detection data are input into the behavior recognition model to obtain an indicator parameter corresponding to each of the detection data output by the behavior recognition model.

[0122] The indicator parameters correspond to each type of detection data and can be key information extracted from the raw detection data by the behavior recognition model to characterize the inherent laws or behavioral patterns of the detection data. The type and meaning of the indicator parameters can depend on the specific dimensions of the detection data (such as time series, numerical fluctuations, state changes, etc.) and the recognition target of the model (such as identifying specific patterns of "puffing state").

[0123] For example, when the detection data is the suction data collected by the air pressure sensor within a preset time period, the corresponding indicator parameter can be the rate of the suction data within the preset time period (such as the suction increases by 0.1 kPa per second), which is used to determine whether there is a "suction start / end" trend.

[0124] Alternatively, when the detection data is acceleration data collected by a three-axis accelerometer within the preset time period, the corresponding indicator parameter can be the maximum and minimum values ​​of the acceleration of each axis, which can reflect the range of acceleration change within the time period and be used to distinguish between "holding still" (small peak value) and "hand movement" (large peak value); the corresponding indicator parameter can also be the calculated time domain characteristics of the acceleration data within the preset time period, for example, the average value of the X / Y / Z axis acceleration within the preset time period, which can reflect the static posture of the device (for example, when the device is placed flat, the Z axis average value is close to the gravity acceleration g = 9.8m / s 2); the corresponding index parameter can also be the calculated frequency domain characteristics of the acceleration data within the preset time period, which can be used to analyze the frequency components of the signal. For example, the time domain signal is converted into the frequency domain through Fourier transform, the energy proportion of each frequency component is calculated, and the periodic motion pattern is identified. Different frequency intervals can also be divided (such as low frequency 0-3Hz, medium frequency 3-10Hz, high frequency>10Hz), the power value of each interval is calculated, and different types of movements are distinguished (such as low frequency corresponds to stable grip, high frequency corresponds to sudden jitter). If the energy in the 2-5Hz frequency band is significant, it may correspond to a slight tremor when the hand is gripped; high-frequency energy above 10Hz may come from the rapid movement during suction.

[0125] Alternatively, when the detection data is the posture angle data collected by the gyroscope within the preset time period, the corresponding index parameter can be the calculated time domain feature of the posture angle data within the preset time period, which can be used to capture the time series law of posture changes. For example, the increase and deceleration rate of the angle per unit time (such as ° / s) can be used to reflect the speed of posture changes; and the corresponding index parameter can also be the calculated spatial feature of the posture angle data within the preset time period, which can be used to describe the combination pattern of multi-dimensional postures, such as combining the pitch angle (Pitch), roll angle (Roll), and yaw angle (Yaw) into a three-dimensional posture vector, calculating its modulus (reflecting the comprehensive amplitude of the posture change) and direction (reflecting the main change dimension). If the modulus suddenly increases and the direction is concentrated in the pitch angle dimension, the user can perform a "head-up and inhale" action.

[0126] In step 1022, based on the indicator parameter and the indicator condition corresponding to the indicator parameter, the indicator parameter that meets the indicator condition is used as the target indicator parameter.

[0127] Among them, on the basis of extracting the multi-dimensional indicator parameters in step 1021, the target indicator parameters that are strongly correlated with the target behavior (such as the puffing state) can be selected through preset rule screening (i.e., the indicator conditions corresponding to the indicator parameters). The purpose is to filter the noise data through the indicator conditions, focus on the effective information, reduce the complexity of the model calculation and improve the judgment accuracy.

[0128] In a possible implementation, the indicator condition corresponding to the indicator parameter may be determined according to the statistical rules of historical data.

[0129] Optionally, corresponding indicator conditions may be set by analyzing a large amount of historical data labeled as "puffing status" and calculating the distribution range (such as mean ± standard deviation) of each indicator parameter.

[0130] For example, if statistics show that the mean acceleration vector amplitude during puffing is 1.2 g and the standard deviation is 0.3 g, the indicator condition can be set to ≥ 0.9 g (mean - standard deviation) to cover more than 90% of real actions.

[0131] In another possible implementation, the indicator condition corresponding to the indicator parameter may be determined in combination with physical logic laws.

[0132] Optionally, you can combine the device hardware characteristics and usage scenarios to set indicator conditions that conform to physical laws.

[0133] For example, if the inward airflow speed detected by the airflow sensor is ≥1.5m / s, and the natural ambient wind speed is usually <1m / s, otherwise it can be determined as invalid airflow.

[0134] It should be noted that, through the indicator parameters and the corresponding indicator conditions, the indicator parameters that meet the indicator conditions are used as target indicator parameters, and the extracted multi-dimensional indicator parameters can be layered filtered and combined for verification.

[0135] Specifically, first, you can apply conditions to each indicator parameter independently, and only keep the indicator parameters that meet the conditions.

[0136] For example, if the extracted air pressure change rate (the drop in air pressure per unit time) is less than or equal to the set threshold of 0.2kPa / s, it can be regarded as no effective suction action; if the acceleration vector amplitude (the combined intensity of the three-axis acceleration) is less than or equal to the set 0.5g, it can be regarded as the atomization device being stationary or slightly shaking; if the airflow direction consistency (the proportion of the inward airflow duration) is less than or equal to the set condition 80%, it can be regarded as chaotic airflow direction, which is environmental interference.

[0137] However, since some indicator parameters may trigger indicator conditions independently due to environmental interference (such as transportation vibration causing acceleration amplitude to exceed the standard), further screening can be carried out through the combination of multiple types of indicator parameters.

[0138] Alternatively, AND logic can be used. For example, the combination is valid only when "the pressure change rate is ≥ 0.2 kPa / s, the acceleration vector amplitude is ≥ 0.5 g, and the airflow direction consistency is ≥ 80%."

[0139] Priority logic can also be used. For example, if the temperature sensor detects that the atomization chamber temperature is less than room temperature (the device is not preheated), the acceleration index condition will automatically increase by 20% (to avoid false positives during cold start).

[0140] Dynamic adjustments can also be used to adapt to changing scenarios. For example, in high-temperature environments, the airflow sensor index condition can be increased to 2.0m / s (to compensate for detection deviation caused by gas expansion), and for users whose historical data frequently shows "low acceleration + high pressure changes", the acceleration index condition can be reduced to 0.3g.

[0141] In this way, through the layer-by-layer screening of indicator conditions and logical combinations, the transformation from the "full set of original indicator parameters" to the "subset of key indicator parameters" can be achieved, which can not only retain the core information used for behavior recognition, but also eliminate the influence of environmental interference and hardware noise. It can ensure that the data input into the behavior recognition model is both "accurate" and "efficient", and improve the accuracy and real-time performance of the final judgment results.

[0142] In step 1023, when the ratio of the number of the target indicator parameters to the number of the indicator parameters is greater than or equal to a preset ratio, a recognized behavior recognition result is obtained.

[0143] In this step, after selecting the target indicator parameters in step 1022, the overall matching degree of the multi-dimensional indicator parameters is quantitatively evaluated to determine whether the current device state corresponds to the target behavior (e.g., puffing state). The purpose is to avoid misjudgment of a single or a few features by using the feature matching ratio to ensure the reliability of the behavior recognition results.

[0144] The target behavior (e.g., puffing) requires that multiple parameters (e.g., air pressure drop, inward airflow, and acceleration consistent with a gripping action) simultaneously meet conditions, rather than being triggered by a single parameter. For example, if only the air pressure characteristic meets the standard (number of target parameters = 1), but the acceleration, airflow, and other characteristics do not meet the standard (total number of parameters = 5), the ratio is 20%, indicating that most parameters do not support the "puffing state" judgment.

[0145] The ratio of the number of target indicator parameters to the number of indicator parameters reflects the degree of consistency in the multi-sensor data's support for the target behavior. A higher ratio indicates that more features from different dimensions match the target behavior pattern, and the confidence level of the judgment is higher. For example, a ratio of 100% indicates that all features meet the criteria and fully match the target behavior; a ratio of 60% indicates that most features meet the criteria. Whether to accept this confidence level depends on the business scenario.

[0146] The preset ratio threshold can be used to balance accuracy and fault tolerance.

[0147] In one possible implementation, a reasonable threshold (such as 60% or 70%) can be set by statistically analyzing the distribution of the proportion of standard index parameters in actual puffing behavior (such as ≥4 / 5 features meeting the standard in 80% of samples).

[0148] In another possible implementation, the preset ratio threshold may also use pre-set data.

[0149] Therefore, by adopting the above technical solution, it is possible to extract indicator parameters through the behavior recognition model, complete the dimensionality reduction compression of the original detection data, and then filter the target indicator parameters through the indicator conditions, further eliminate irrelevant parameters, and finally only need to perform simple numerical calculations on the ratio of the filtered indicator parameters, avoiding complex model reasoning on the full amount of data, greatly reducing computing time and energy consumption. In this way, by comparing the indicator conditions of multiple detection data, it is possible to avoid misjudgments caused by single sensor noise or anomalies (such as accidental vibration triggering the accelerometer), and by calculating the ratio of the number of target indicator parameters to the total number of indicator parameters (such as ≥60%), abstract behavior recognition can be converted into quantifiable mathematical judgments, avoiding the randomness caused by fuzzy decision-making.

[0150] In some embodiments, the detection data includes at least one of suction data, acceleration data, posture angle data, and temperature data within a preset time period.

[0151] Optionally, the detection data may include at least one of the suction data collected by the pressure sensor within a preset time period, the acceleration data collected by the three-axis accelerometer within the preset time period, the attitude angle data collected by the gyroscope within the preset time period, and the temperature data collected by the temperature sensor within the preset time period.

[0152] Considering that at least two types of detection data (such as the time domain signal of the accelerometer, the continuous value of the pressure sensor, and the three-dimensional vector of the gyroscope) have completely different data forms, noise distributions, and physical meanings, if all the detection data are directly spliced ​​into high-dimensional vectors and input into a single model (such as inputting three-axis acceleration + pressure + temperature data, a total of 6 dimensions), the model parameters will increase exponentially with the input dimension, requiring more computing power and training data, as well as the dimensional differences of different physical quantities (such as acceleration m / s 2 vs air pressure Pa) will lead to imbalanced gradient optimization and make it difficult for the model to learn cross-dimensional associations.

[0153] Therefore, in some embodiments, the behavior recognition model may include at least two recognition sub-models.

[0154] At least two of the identification sub-models correspond to at least two types of detection data.

[0155] In this way, each sub-model only needs to correspond to the indicator parameter rules of a single data dimension (such as the peak frequency of acceleration), which greatly reduces the complexity of the model; and can avoid "irrelevant feature interference": for example, the fluctuation of temperature data (ambient temperature changes) is irrelevant to the holding action. Through independent sub-models, it can prevent it from interfering with the determination of the conditions of the acceleration indicator parameters. And it can improve the heterogeneous capability of the behavior recognition model. When the source of the detection data changes (such as changing from a MEMS accelerometer to a fiber optic gyroscope), it is only necessary to retrain the corresponding sub-model for the new sensor, and the overall recognition logic does not need to be reconstructed, thereby improving the system's compatibility with hardware iterations.

[0156] Optionally, based on the type of target detection data, a target recognition sub-model corresponding to the target detection data can be selected from the at least two recognition sub-models, and then each target detection data can be input into the corresponding target recognition sub-model to obtain the indicator parameters corresponding to the target detection data output by the target recognition sub-model.

[0157] The target detection data is one of at least two types of detection data.

[0158] The recognition sub-model may include a time series analysis sub-model, a motion pattern recognition sub-model, a posture change detection sub-model, and a heat conduction sub-model.

[0159] For example, the corresponding sub-model can be matched first according to the type of detection data (air pressure, acceleration, gyroscope, temperature) to ensure that each type of data is processed by the most suitable model.

[0160] For example, the suction data collected by the air pressure sensor is a continuous value that changes with time, so its rate of change and waveform characteristics need to be analyzed, so it can be matched with the corresponding timing analysis sub-model; the acceleration data collected by the three-axis accelerometer reflects the intensity and stability of the device's movement, so the continuity and amplitude stability need to be judged, so it can be matched with the corresponding motion pattern recognition sub-model; the attitude angle data collected by the gyroscope needs to analyze the angle change rate and exclude violent shaking, so it can be matched with the corresponding attitude change detection sub-model; the temperature data collected by the temperature sensor reflects the working state of the atomizer core, so the cooling rate needs to be matched, so it can be matched with the corresponding heat conduction sub-model.

[0161] For the timing analysis sub-model corresponding to the suction data (air pressure value), the air pressure value sequence within a preset time period (such as 1 second data sampled at 50Hz, a total of 50 points) can be input into the timing analysis sub-model, and then the air pressure difference / time interval between adjacent time points is calculated to obtain the descent rate per second (unit: kPa / s), and the first-order derivative of the air pressure curve is calculated to determine whether it presents a single-peak waveform (the derivative drops sharply at the beginning of inhalation and rises back at the end of inhalation, forming a single peak). If the descent rate is between 0.5 and 2 kPa / s and the first-order derivative is a single-peak waveform, the output is "air pressure characteristics meet the standards."

[0162] For the motion pattern recognition sub-model corresponding to the acceleration data, the time domain sequence of the three-axis acceleration (50 sampling points each on the X / Y / Z axis) can be input into the motion pattern recognition sub-model, and the The vector amplitude reflecting the overall motion intensity is calculated to detect whether the detection vector amplitude is continuously stable within 300ms (fluctuation range ≤±0.1g). If the acceleration vector amplitude is continuously stable for ≥300ms (condition 3), the output is "acceleration feature meets the standard".

[0163] For the attitude change detection sub-model corresponding to the gyroscope data, the pitch angle and roll angle sequence within the preset time period can be input into the attitude change detection sub-model, and the angle difference / time interval between adjacent time points can be calculated to obtain the change rate per second. If the change rate of the pitch angle and roll angle are both <5° / s, it means that the device attitude is stable (no violent shaking or flipping), and the output is "attitude feature meets the standard".

[0164] For the heat conduction sub-model corresponding to the temperature data, the temperature value sequence within the preset time period is input into the heat conduction characteristic sub-model, and the temperature difference / time interval between adjacent time points is calculated. If the cooling rate is between -0.3 and -0.1°C / s (i.e., a decrease rate of 0.1 to 0.3°C / s), it means that the atomizer core is cooled by the inhaled airflow, and the output is "temperature characteristic meets the standard."

[0165] By adopting the above technical solution, through this "classification modeling + multi-condition filtering + proportional decision-making" mechanism, the system can accurately identify real puffing behavior, while effectively suppressing non-target events such as environmental interference and hardware noise, thereby improving the reliability of the atomization equipment and user experience.

[0166] Considering that the main safety risk of atomizer equipment comes from mismatching the timing of oil and power supply (e.g., heating first, resulting in dry burning, or oil supply before heating, resulting in liquid leakage), a safety buffer can be created by triggering commands in stages and setting a preset time delay to ensure that each component starts in the order of "oil supply - pre-filling - heating / power supply".

[0167] In some embodiments, controlling the atomization device to perform an atomization operation may include at least one of the following methods:

[0168] S1. Sending a start instruction to the oil supply solenoid valve, wherein the start instruction is used to control the oil supply solenoid valve to be turned on, so as to conduct the atomized substrate to the atomizer core.

[0169] For example, after the start command is sent, the solenoid valve needs to wait for a first preset time period (such as 200ms) before it is turned on to deliver the atomized matrix to the atomization core.

[0170] This delay prevents premature oil supply when the system is not ready for heating or power, preventing liquid from accumulating outside the atomizer core and causing leakage. Furthermore, the delay ensures that the heating element has enough time to start after oil supply (e.g., if the heating command in S2 is triggered after the first preset time period), avoiding the risk of dry burning due to "oil supply but no heating" (e.g., if the heating command in S2 is triggered after the first preset time period).

[0171] S2. Sending a heating instruction to the heating component, where the heating instruction is used to control the heating component to heat the atomizer core after a first preset time period.

[0172] For example, after the heating instruction is triggered, the heating component needs to wait for a first preset time period (such as 300ms) before starting to heat the atomizer core.

[0173] Delayed heating ensures that the oil supply solenoid valve in S1 is open, allowing the atomizer to fully soak the atomizer core (e.g., a 200ms delay allows sufficient oil to flow), preventing dry burning. Furthermore, if the oil supply fails (e.g., the solenoid valve fails to open), the delayed heating prevents dry burning in the absence of oil. This "oil supply → delay → heating" sequence creates a safety barrier.

[0174] S3. Send a wake-up instruction to the power management component, where the wake-up instruction is used to control the power supply component to supply power to the atomizer core after a second preset time period.

[0175] For example, after the wake-up command is triggered, the power supply component needs to wait for a second preset time period (such as 400ms) before supplying power to the atomizer core.

[0176] In this way, after S1 is oiled and S2 triggers the heating command, the power supply is further delayed to ensure the complete process of "oil in place - heating module ready - power on and heating".

[0177] For example, 200ms (S1 oil supply) - 300ms (S2 heating delay) - 400ms (S3 power supply delay), the total buffer time can be as long as 900ms, which fully avoids timing disorder.

[0178] Through one or more of the modes S1-S3, S1+S2+S3 can be triggered simultaneously.

[0179] The execution order is: first send S1 (oil flow command), the solenoid valve is turned on with a delay of 200ms, then send S2 (heating command), the heating component is started with a delay of 300ms (need to wait for the end of S1 for 200ms, and then delay 100ms), then send S3 (power supply command), the power supply is delayed by 400ms (need to wait for the total delay of S1+S2 for 500ms, and then delay -100ms, which can actually be adjusted to be logically later than the heating start).

[0180] This ensures that each link is executed in sequence, avoids heating / power supply when oil is not in place, and only triggers S1+S3.

[0181] If the system mistakenly triggers S1 (oil flow) and S3 (power supply) but does not trigger S2 (heating), during the power supply delay (e.g. 400ms), the system can use sensors to detect whether there is a heating command (if not detected, power supply is prohibited) to prevent the atomizer core from drying out due to power on when there is no heating.

[0182] By adopting the above technical solution, a multi-layer safety buffer zone can be built through phased command delay and cross-component logical interlocking. The preset time period ensures that oil flow, heating, and power supply are strictly executed in sequence. During the delay of any link, the preconditions (such as oil in place, heating ready) can be detected to block abnormal startup. Moreover, they can be used in combination, and the delay time is superimposed to form a longer safety window, minimizing the risks of dry burning, leakage, etc.

[0183] In some embodiments, when the current suction data is greater than or equal to the target suction threshold, and the second duration during which the current suction data is greater than or equal to the target suction threshold is within a second preset duration range, the atomization device is controlled to be in the first state.

[0184] The first state includes the power supply component supplying power to the heating component, so that the heating component heats the atomizer core at a first preset power.

[0185] In this embodiment, the target suction threshold is a reference value that is dynamically adjusted based on the user's historical puffing habits, device performance, and ambient temperature. The current suction data meets the standard, indicating that the suction intensity applied by the user is sufficient to trigger atomization (such as normal inhalation strength).

[0186] In this way, when the current suction data is greater than or equal to the target suction threshold, slight airflow fluctuations (such as a small drop in air pressure caused by ambient wind) can be excluded, and only the effective suction generated by active suction can be identified. When the suction intensity meets the standard and the duration is reasonable, it is determined to be "effective suction", which can trigger the first state. After receiving the effective suction determination signal, the power supply component immediately supplies power to the heating component to ensure that the heating action is synchronized with the suction.

[0187] When the power supply component supplies power to the heating component, a relatively small power can be used for preheating.

[0188] The power supply assembly provides low voltage / current (such as 30%-50% of the rated power) to the heating assembly, so that the temperature of the atomizer core slowly rises to the preheating temperature range (such as 60-80℃, which is 100-200℃ lower than the normal atomization temperature). This can preheat the atomizer core and put it in a "ready to work state". When you officially draw, you can quickly reach the optimal atomization temperature and the smoke volume is more stable.

[0189] It should be noted that when the user's suction force continues to increase (current suction force ≥ target threshold × 1.5) or the duration exceeds 500ms, the system will determine it as "formal suction" and switch to normal heating. For example, if the rated power is 100%, if there is no further suction action within 2 seconds after preheating, the power supply will be automatically stopped and the machine will enter standby mode to avoid idling energy consumption.

[0190] In other embodiments, the current posture information of the atomizing device may be determined based on the acceleration data and the posture angle data; and when it is determined that the atomizing device is in a specified posture based on the posture information, the atomizing device may be controlled to be in the second state.

[0191] The second state includes the power supply component supplying power to the heating component, so that the heating component heats the atomizer core at a second preset power.

[0192] In this embodiment, the device tilt angle can be determined by calculating the static acceleration component (projection of gravity acceleration), and the dynamic rotation process can be identified by calculating the attitude angle (pitch angle, roll angle) through integration.

[0193] In this way, the static attitude of the acceleration is combined with the dynamic changes of the gyroscope to build a more robust attitude judgment model.

[0194] For example, when the roll angle is ≈0° and the pitch angle is ≈0°, the posture of the atomizer device is held horizontally; when the roll angle is ≈0° and the pitch angle is ≈90°, the posture of the atomizer device is held vertically; when the acceleration amplitude suddenly increases and the attitude angle change rate is >5° / s, the atomizer device changes from a stationary state to a moving state.

[0195] When the device maintains a specified posture (such as horizontal holding) for more than a preset time (such as 300ms), preheating can be triggered. This can avoid false triggering due to short posture changes (such as hand shaking).

[0196] And when a specific posture conversion process (such as from vertical to horizontal) is detected and the acceleration change conforms to the preset pattern (such as acceleration first and then deceleration), preheating is triggered immediately.

[0197] It should also be noted that, similar to the first state, the power supply component outputs 30%-50% of the rated power (such as 3-5W), and the preheating lasts for 1-2 seconds, so that the temperature of the atomizer core rises to 60-80°C.

[0198] By adopting the above technical solution, by integrating acceleration and posture angle data, the second state is transformed from "passive response" to "active prediction". Through the triggering logic of physical posture, it not only improves the smoothness of user experience, but also reduces energy consumption and safety risks through precise control.

[0199] Figure 2 This is a flow chart of another atomization device control method provided in an embodiment of the present application. Figure 2 As shown, the method can be applied to the controller of the atomization device.

[0200] The method may include the following steps.

[0201] Step 201: Acquire detection data collected by sensors in an atomization device.

[0202] Among them, the atomization device can be provided with multiple sensors for collecting detection data.

[0203] The detection data includes suction data collected by the pressure sensor within a preset time period, acceleration data collected by the three-axis accelerometer within the preset time period, attitude angle data collected by the gyroscope within the preset time period, and temperature data collected by the temperature sensor within the preset time period.

[0204] Step 202: Acquire historical suction data and current temperature data inside the atomizing device.

[0205] Step 203 : adjusting the preset suction threshold according to the historical suction data and the current temperature data to obtain an adjusted target suction threshold.

[0206] Step 204 , when the current suction data is greater than or equal to the target suction threshold and the first duration is within a first preset duration range, select a target recognition sub-model corresponding to the target detection data from the at least two recognition sub-models according to the type of the target detection data.

[0207] The target detection data is one of at least two types of detection data. The behavior recognition model includes at least two recognition sub-models, and different recognition sub-models correspond to different detection data.

[0208] Step 205: input the target detection data into the corresponding target recognition sub-model to obtain the index parameters corresponding to the target detection data output by the target recognition sub-model.

[0209] Step 206: Based on the indicator parameter and the indicator condition corresponding to the indicator parameter, the indicator parameter that meets the indicator condition is used as the target indicator parameter.

[0210] Step 207 : When the ratio of the number of the target index parameters to the number of the index parameters is greater than or equal to a preset ratio, outputting a behavior recognition result of the atomizing device being in a puffing state.

[0211] Step 208 : When the behavior recognition result indicates that the atomizing device is in the suction state, the atomizing device is controlled to perform an atomizing operation.

[0212] For example, a start instruction can be sent to the oil supply solenoid valve, which is used to control the oil supply solenoid valve to be turned on so as to conduct the atomized matrix to the atomizer core; or, a heating instruction can be sent to the heating component, which is used to control the heating component to heat the atomizer core after a first preset time period; or, a wake-up instruction can be sent to the power management component, which is used to control the power supply component to supply power to the atomizer core after a second preset time period.

[0213] By adopting the above technical solution, multi-dimensional data collected by multiple sensors can be deployed and cross-validated using a behavior recognition model. The pre-trained model can learn the real feature combination of the atomizing device in the puffing state, and more accurately filter the interference data of the puffing behavior, which can improve the recognition accuracy and reduce the probability of the atomizing device being triggered incorrectly.

[0214] Figure 3 This is a block diagram of an atomization device control device provided in an embodiment of the present application. Figure 3 As shown, the device 300 includes:

[0215] An acquisition module 301 is used to acquire detection data collected by sensors in an atomization device;

[0216] A preprocessing module 302 is configured to preprocess the detection data based on historical suction data and current temperature to obtain denoised detection data; the detection data includes at least two types of detection data;

[0217] The recognition module 303 is used to input at least two types of detection data into a pre-trained behavior recognition model to obtain a behavior recognition result identified by the behavior recognition model, and the behavior recognition result is used to characterize the usage status of the atomization device;

[0218] The first control module 304 is configured to control the atomizing device to perform an atomizing operation when the behavior recognition result indicates that the atomizing device is in a puffing state.

[0219] In some embodiments, the pre-processing module is configured to adjust a preset suction threshold according to the historical suction data and the current temperature data to obtain an adjusted target suction threshold;

[0220] an identification module, configured to input at least two types of detection data into a pre-trained behavior recognition model to obtain a behavior recognition result identified by the behavior recognition model when the current suction data is greater than or equal to the target suction threshold and the first duration is within a first preset duration range;

[0221] The first duration is the duration during which the suction data is greater than or equal to the target suction threshold.

[0222] In some embodiments, the identification module includes:

[0223] A first identification submodule is configured to input at least two types of detection data into a behavior recognition model to obtain indicator parameters output by the behavior recognition model; the indicator parameters respectively correspond to each type of detection data;

[0224] A first determining submodule is configured to, based on the indicator parameter and the indicator condition corresponding to the indicator parameter, take the indicator parameter that satisfies the indicator condition as the target indicator parameter;

[0225] The second determining submodule is configured to output a behavior recognition result of the atomizing device being in a puffing state when the ratio of the number of the target index parameters to the number of the index parameters is greater than or equal to a preset ratio.

[0226] In some embodiments, the behavior recognition model includes at least two recognition sub-models, and the at least two recognition sub-models correspond to at least two types of detection data respectively; the first recognition sub-module is used to select a target recognition sub-model corresponding to the target detection data from the at least two recognition sub-models according to the type of the target detection data, and the target detection data is one of the at least two types of detection data;

[0227] The target detection data is input into the corresponding target recognition sub-model to obtain the index parameters corresponding to the target detection data output by the target recognition sub-model.

[0228] In some embodiments, the detection data includes at least one of suction data, acceleration data, posture angle data, and temperature data within a preset time period.

[0229] In some embodiments, the first control module includes:

[0230] The first sending submodule is used to send a start instruction to the oil supply solenoid valve, and the start instruction is used to control the oil supply solenoid valve to be turned on, so as to conduct the atomized matrix to the atomizer core; or the second sending submodule is used to send a heating instruction to the heating component, and the heating instruction is used to control the heating component to heat the atomizer core after a first preset time period; or the third sending submodule is used to send a wake-up instruction to the power management component, and the wake-up instruction is used to control the power supply component to supply power to the atomizer core after a second preset time period.

[0231] In some embodiments, the detection data includes suction data collected by the air pressure sensor within a preset time period; the device further includes:

[0232] a second control module, configured to control the atomizing device to be in the first state when the current suction data is greater than or equal to the target suction threshold and a second duration during which the current suction data is greater than or equal to the target suction threshold is within a second preset duration range;

[0233] The first state includes the power supply component supplying power to the heating component, so that the heating component heats the atomizer core at a first preset power.

[0234] In some embodiments, the detection data includes acceleration data collected by a three-axis accelerometer within a preset time period and attitude angle data collected by a gyroscope within the preset time period; the device further includes:

[0235] a determination module, configured to determine current posture information of the atomization device based on the acceleration data and the posture angle data;

[0236] The third control module is used to control the atomization device to be in a second state when it is determined that the atomization device is in a specified posture according to the posture information. The second state includes the power supply component supplying power to the heating component so that the heating component heats the atomization core at a second preset power.

[0237] In this embodiment, detection data collected by sensors in the atomizing device is obtained; the detection data is preprocessed based on historical suction data and current temperature to obtain denoised detection data; the detection data includes at least two types of detection data; the at least two types of detection data are input into a pre-trained behavior recognition model to obtain a behavior recognition result identified by the behavior recognition model; the behavior recognition result is used to characterize the usage status of the atomizing device; when the behavior recognition result characterizes that the atomizing device is in a suction state, the atomizing device is controlled to perform an atomizing operation. In this way, by deploying multi-dimensional data collected by multiple sensors and cross-validating the data using a behavior recognition model, the pre-trained model can learn the actual feature combination of the atomizing device in the suction state, more accurately filter out the interference data of the suction behavior, and improve the recognition accuracy to reduce the probability of the atomizing device being falsely triggered.

[0238] The present invention also provides an electronic device, see Figure 4 , including: a processor 601, a memory 602, and a computer program 6021 stored in the memory and capable of running on the processor, and when the processor executes the program, the atomization device control method of the aforementioned embodiment is implemented.

[0239] The present invention also provides a readable storage medium, which, when the instructions in the storage medium are executed by a processor of an electronic device, enables the electronic device to execute the atomization device control method of the aforementioned embodiment.

[0240] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0241] It should be noted that the various information and data obtained in the embodiments of the present invention are all obtained with the authorization of the information / data holder.

[0242] The algorithm and display provided herein are not inherently related to any particular computer, virtual system or other device. Various general-purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious that the structure required for constructing this type of system. In addition, the present invention is not directed to any specific programming language. It should be understood that various programming languages ​​can be utilized to realize the content of the present invention described herein, and the above description of specific languages ​​is for the purpose of disclosing the best mode of the present invention.

[0243] In the description provided herein, numerous specific details are described. However, it is understood that embodiments of the present invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques are not shown in detail so as not to obscure the understanding of this description.

[0244] Similarly, it should be understood that in order to streamline the present invention and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as reflected in the claims below, inventive aspects lie in less than all the features of the individual embodiments disclosed above. Accordingly, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of the present invention.

[0245] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and in addition may be divided into multiple submodules or subunits or subcomponents. All features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed herein may be combined in any combination, except that at least some of such features and / or processes or units are mutually exclusive. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0246] It should be noted that the above embodiments illustrate rather than limit the invention, and that those skilled in the art may devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between brackets should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention may be implemented by means of hardware comprising several different elements and by means of appropriately programmed computers. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third etc. does not indicate any order. These words may be interpreted as names.

[0247] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for controlling an atomization device, characterized in that: The method comprises: Obtain detection data collected by sensors in atomization equipment; Preprocessing the detection data according to historical suction data and current temperature to obtain denoised detection data; the detection data includes at least two types of detection data; Inputting at least two types of detection data into a pre-trained behavior recognition model to obtain a behavior recognition result identified by the behavior recognition model; the behavior recognition result is used to characterize the usage status of the atomization device; When the behavior recognition result indicates that the atomization device is in a suction state, the atomization device is controlled to perform an atomization operation.

2. The method according to claim 1, characterized in that The pre-processing of the detection data according to the historical suction data and the current temperature to obtain the denoised detection data includes: Adjusting a preset suction threshold according to the historical suction data and the current temperature data to obtain an adjusted target suction threshold; Inputting at least two types of detection data into a pre-trained behavior recognition model to obtain a behavior recognition result identified by the behavior recognition model includes: When the current suction data is greater than or equal to the target suction threshold and the first duration is within a first preset duration range, inputting at least two types of detection data into the pre-trained behavior recognition model to obtain a behavior recognition result identified by the behavior recognition model; The first duration is the duration during which the suction data is greater than or equal to the target suction threshold.

3. The method according to claim 1, characterized in that Inputting at least two types of detection data into a pre-trained behavior recognition model to obtain a behavior recognition result identified by the behavior recognition model includes: Inputting at least two types of detection data into the behavior recognition model to obtain index parameters output by the behavior recognition model; the index parameters respectively correspond to each type of detection data; According to the indicator parameters and the indicator conditions corresponding to the indicator parameters, the indicator parameters that meet the indicator conditions are used as target indicator parameters; When the ratio of the number of the target index parameters to the number of the index parameters is greater than or equal to a preset ratio, a behavior recognition result of the atomization device being in a puffing state is output.

4. The method according to claim 3, characterized in that The behavior recognition model includes at least two recognition sub-models, and the at least two recognition sub-models correspond to at least two types of detection data respectively; Inputting at least two types of detection data into the behavior recognition model to obtain indicator parameters output by the behavior recognition model includes: selecting, according to a type of target detection data, a target recognition sub-model corresponding to the target detection data from at least two recognition sub-models, wherein the target detection data is one of the at least two types of detection data; The target detection data is input into the corresponding target recognition sub-model to obtain the index parameters corresponding to the target detection data output by the target recognition sub-model.

5. The method according to claim 4, characterized in that The detection data includes at least one of suction data, acceleration data, posture angle data, and temperature data within a preset time period.

6. The method according to claim 1, characterized in that The controlling the atomizing device to perform an atomizing operation includes: Sending a start instruction to the oil supply solenoid valve, wherein the start instruction is used to control the oil supply solenoid valve to be turned on, so as to conduct the atomized matrix to the atomizing core; or, Sending a heating instruction to the heating component, wherein the heating instruction is used to control the heating component to heat the atomizer core after a first preset time period; or, A wake-up instruction is sent to the power management component, where the wake-up instruction is used to control the power supply component to supply power to the atomizer core after a second preset time period.

7. The method according to claim 2, characterized in that The detection data includes suction data collected by the air pressure sensor within a preset time period; The method further comprises: When the current suction data is greater than or equal to the target suction threshold, and a second duration during which the current suction data is greater than or equal to the target suction threshold is within a second preset duration range, controlling the atomizing device to be in the first state; The first state includes the power supply component supplying power to the heating component, so that the heating component heats the atomizer core at a first preset power.

8. The method according to claim 1, characterized in that The detection data includes acceleration data collected by the three-axis accelerometer within a preset time period and attitude angle data collected by the gyroscope within the preset time period; The method further comprises: Determining current posture information of the atomizing device according to the acceleration data and the posture angle data; When it is determined that the atomization device is in a specified posture according to the posture information, the atomization device is controlled to be in a second state, and the second state includes the power supply component supplying power to the heating component so that the heating component heats the atomization core at a second preset power.

9. An atomizing device, characterized in that: The device comprises: An acquisition module is used to obtain detection data collected by sensors in the atomization device; a preprocessing module, configured to preprocess the detection data according to historical suction data and current temperature to obtain the detection data after denoising; the detection data includes at least two types of detection data; an identification module, configured to input at least two types of detection data into a pre-trained behavior recognition model to obtain a behavior recognition result identified by the behavior recognition model, wherein the behavior recognition result is used to characterize the usage status of the atomization device; The first control module is configured to control the atomization device to perform an atomization operation when the behavior recognition result indicates that the atomization device is in a puffing state.

10. A controller, characterized in that: The controller is configured to implement the method according to any one of claims 1 to 8 when executed.

11. An atomizing device, characterized in that: The atomizing device includes a sensor, a heating component, and the controller according to claim 10, The sensor is used to collect detection data; The heating component is used to generate heat so that the atomizing device can perform an atomizing operation.

12. A readable storage medium, characterized in that: When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to implement the method according to any one of claims 1 to 8.

Images