Smoking behavior identification method based on multi-source perception and particle size fingerprint analysis
By combining multi-source sensing and particle size fingerprint analysis with particulate matter sensors and infrared sensors, the problems of high false alarm rate, poor privacy compliance and high operation and maintenance cost of smoke detection in enclosed public places have been solved, achieving high-precision, low false alarm rate, real-time response and low-cost smoke-free environment management.
Patent Information
- Application Number
- CN202511865229.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-11
- Publication Date
- 2026-04-10
AI Technical Summary
Existing smoke detection technologies suffer from high false alarm rates, poor privacy compliance, slow response times, and high maintenance costs in enclosed public spaces, making it difficult to achieve large-scale deployment with high accuracy, low false alarm rates, privacy compliance, and real-time performance.
Employing a multi-source sensing and particle size fingerprinting approach, this method combines particulate matter sensors and infrared sensors to monitor personnel status and particulate matter concentration changes in real time. It uses the particle size distribution index and PM2.5 growth rate to accurately determine smoking behavior, forming a traceable event-level evidence chain, and integrates low-cost hardware for deployment.
It achieves high detection accuracy and low false alarm rate, avoids privacy leaks, reduces operation and maintenance costs, provides instant response and accurate traceability capabilities, and supports smoke-free environment management in large-scale enclosed venues.
Smart Images

Figure FT_1 
Figure FT_2 
Figure SMS_78
Abstract
Description
Technical Field
[0001] This invention belongs to the field of environmental hygiene technology in public spaces, specifically relating to a smoking behavior identification method based on multi-source sensing and particle size fingerprint analysis. Background Technology
[0002] With the in-depth implementation of the "Healthy China Action (2019-2030)," enclosed public places such as universities, transportation hubs, and commercial complexes have been fully included in the key scope of smoke-free environment construction. Among them, university dormitories and public restrooms have become weak links in tobacco control supervision due to their high privacy, random personnel flow, and limited ventilation.
[0003] Existing automated monitoring methods suffer from irreconcilable contradictions regarding accuracy, real-time performance, location capabilities, and privacy compliance. Manual patrols, limited by time and space, cannot achieve full coverage, resulting in a lack of a widely applicable closed-loop management solution. Existing automated monitoring methods can be broadly categorized into three types:
[0004] (1) Threshold-type smoke alarm: mainly adopts the "single-point concentration threshold" triggering logic. In a confined and narrow space, this type of device is easily interfered with by non-cigarette smoke sources such as water vapor, mosquito coils, and aerosol dust, resulting in a high false alarm rate; in addition, the sensor only outputs switch (Boolean value) signals, which cannot provide the spatial coordinates of the smoke source and the complete evidence chain of the event development, making it difficult to support subsequent accurate intervention.
[0005] (2) Visible light + AI visual analysis: Human movements (such as hand and mouth features) and smoke patterns are collected through surveillance cameras. However, in dimly lit dormitory bathrooms with mirror reflections or multiple people obstructing the view, the false negative rate of target detection increases significantly; and false positives are easily generated for non-smoking hand raising movements. More importantly, continuous video stream acquisition involves sensitive information such as facial features and daily life.
[0006] (3) Infrared thermal imaging grayscale monitoring: The grayscale gradient change signal is generated by the hot spot produced by the burning of cigarettes. However, transient heat sources such as human breathing, hot water cups, and power bank heating may form similar gradient changes, thus triggering false alarms; at the same time, if infrared imaging covers toilet areas and confidential areas, there is also a serious risk of privacy infringement and leakage.
[0007] The existing solutions generally rely on single-modal perception and lack coupled modeling of the "human-smoke-space" ternary relationship. This makes it difficult to establish cross-validation and dynamic confidence assessment mechanisms among multi-source heterogeneous data, resulting in rigid event judgment boundaries that cannot adaptively adjust to changes in environmental temperature, humidity, air exchange rate, and personnel density. Alarm information can only output "presence or absence," lacking descriptions of key dimensions such as "identity, location, duration, and spread trajectory," forcing managers to conduct secondary verification and thus missing the best intervention opportunity. In addition, the fragmented communication protocols of various devices make it difficult to seamlessly integrate with existing smart campus or security IoT platforms, resulting in severe data silos and high operation and maintenance costs.
[0008] In summary, there is an urgent need for a new sensing and recognition method that balances high accuracy, low false alarm rate, and privacy compliance in existing smoke-free scenarios in enclosed public places. Summary of the Invention
[0009] This invention aims to solve the following technical problems:
[0010] 1) Existing smoke detection technologies lack anti-interference capabilities, leading to frequent false alarms: Traditional smoke alarms generally rely on a single "concentration threshold triggering" mechanism, lacking the ability to effectively identify aerosol components. They are highly susceptible to interference from non-smoking sources such as water vapor, mosquito coil smoke, and suspended particulate matter, resulting in false alarms. Frequent false alarms not only reduce the system's reliability but also cause users to experience "alarm fatigue," leading to manual power-off and creating blind spots in safety monitoring. Furthermore, AI vision-based solutions have limitations in distinguishing smoking actions from everyday hand gestures (such as grooming or eating), making it difficult to effectively eliminate interference from similar behavioral features, resulting in a high false alarm rate.
[0011] 2) Imaging-based monitoring methods face severe privacy compliance challenges in private settings: In scenarios with extremely high privacy requirements, such as dormitories and public restrooms, deploying cameras or high-resolution thermal imaging equipment inevitably involves the continuous collection and processing of facial features, body movements, and daily living information. This potentially conflicts with the "minimum necessary" principle for personal information processing in Article 6 of the Personal Information Protection Law, easily triggering users' privacy concerns and posing significant compliance risks to administrators, severely restricting the social acceptance and large-scale application of such systems.
[0012] 3) Traditional threshold triggering mechanisms have inherent response lag and insufficient real-time early warning capabilities: Existing alarm systems mostly adopt an "accumulation-trigger" mode, where particulate matter concentration requires an objective amount of time to accumulate from generation and diffusion mass transfer to reaching the sensor threshold. This lag in the physical process makes it difficult for the system to respond immediately in the early stages of smoking behavior (i.e., the moment of ignition), often missing the best window for dissuasion and intervention, thus limiting management effectiveness.
[0013] 4) Distributed hardware architecture drives up construction and maintenance costs, hindering large-scale deployment: Existing intelligent monitoring systems mostly adopt a distributed architecture of "independent sensing + independent transmission + independent computing," lacking intensive reuse of hardware resources and centralized control design. This architecture results in high deployment costs per point and high maintenance costs throughout the entire lifecycle, making it difficult to achieve low-cost, full-coverage sustainable operation in large venues such as campuses and commercial complexes.
[0014] 5) Traditional monitoring methods have low spatial resolution and lack precise source tracing capabilities: Traditional smoke detectors are usually sparsely deployed in large-area grids, which can only report "regional anomalies" and are difficult to accurately locate micro-spaces (such as specific cubicles, corridors, or beds). This coarse-grained monitoring result cannot automatically pinpoint the source of the violation, and still requires manual investigation or retrieval of surveillance footage for secondary source tracing, resulting in relatively low management efficiency.
[0015] The purpose of this invention is to provide a smoking behavior recognition method based on multi-source sensing and particle size fingerprint analysis. This method, without imaging, achieves real-time determination of "personal presence - smoking behavior - precise location" through multi-source data fusion and intelligent analysis of cigarette smoke fluctuation characteristics. It automatically generates a traceable, quantifiable, and interconnected event-level evidence chain. This not only fills the current technological gap in public health governance but also provides universities, commercial complexes, and health institutions with a sustainable and automated closed-loop management method for smoke-free environments.
[0016] The technical solution of the present invention is as follows:
[0017] A smoking behavior recognition method based on multi-source sensing and particle size fingerprint analysis includes:
[0018] The particulate matter sensor module is configured to continuously acquire multi-channel particle size spectra in the target space in an optical-free manner and generate high-resolution particle size distribution sequences over time.
[0019] An infrared sensor module is installed to monitor the entry and exit of personnel in real time, ensuring that detection is triggered only when personnel are present.
[0020] By calculating the short-term changes in the concentration distribution of particulate matter of different particle sizes using parametric formulas, and combining this with human status and / or the concurrent PM2.5 growth rate, accurate judgment of smoking events can be achieved.
[0021] Furthermore, when the following conditions are met:
[0022] ,
[0023] This is considered a violation of smoking regulations;
[0024] in, It is a Boolean value, representing the time... The infrared sensor detected the presence of a person; if true, then someone is present and has not left.
[0025] in, This represents logical AND. Indicates logical OR;
[0026] in, It is an indicator function, when hour, It is 1 if it is true, otherwise it is 0;
[0027] The eigenvalues are explained below:
[0028] :recent indivual Among the values, at least The values satisfy ;
[0029] : A specific threshold for the exponential growth rate of particle size distribution concentration, recently indivual The value must include several or all values that are greater than this value;
[0030] A specific threshold for PM2.5 concentration growth rate; the increase in PM2.5 within the last k seconds must exceed this value.
[0031] Furthermore, the eigenvalues , It is obtained through the following steps:
[0032] 1) Burn flue-cured cigarettes, blended cigarettes, cigars, heated tobacco products, e-cigarette vaporizing liquid, and common non-human-caused smoke sources in a standard experimental chamber and collect time-series spectra of particle size concentrations of different smoke sources.
[0033] 2) Repeat the experiment N ≥ 30 times for each type of combustible material;
[0034] Using the particle size distribution index: ( The difference between PM2.5 and PM1 at any given time) and the rate of change of the particle size distribution index: ( in the past Rate of change per second The feature value is determined by analyzing and extracting the changes in the concentration distribution of smoke source particles (which is a positive integer parameter), focusing on identifying the differences in the rate of change of particle size distribution between the target cigarette smoke and common non-human smoke sources, and then determining the feature value accordingly.
[0035] Furthermore, immediately after each detection of raw particle concentration data of different particle sizes by the particle sensor module, the following calculations are performed:
[0036] Particle size distribution index: ,Right now The difference between PM2.5 and PM1 at any given time;
[0037] Particle size distribution index change rate: ,Right now in the past Rate of change per second It is a positive integer parameter;
[0038] Set of particle size distribution index change rates: That is, recently indivual value, It is a positive integer parameter.
[0039] Furthermore, the structure including the particulate matter sensor module and the infrared sensor module is integrated into a hardware device that can be embedded in the ceiling or wall of a public place to achieve plug-and-play functionality.
[0040] Furthermore, the particulate matter sensor module includes a particulate matter laser scattering sensor, which continuously monitors the concentration of PM1, PM2.5, and PM10 particulate matter in the air at high frequency.
[0041] Furthermore, the particulate matter sensor module also includes a combustion heat source infrared camera and / or a chemical gas sensor.
[0042] Furthermore, the infrared sensor module includes a reflective active infrared detection sensor.
[0043] Furthermore, the infrared sensor module includes any one of the following:
[0044] Integrate a wireless switch or sensor circuit into the door lock to automatically trigger a wireless signal based on the closing state of the lock tongue;
[0045] Passive infrared receiver sensor in conjunction with millimeter-wave radar sensor;
[0046] Ultrasonic ranging sensors monitor changes in the propagation path of sound waves within a confined space. By observing whether the effective space between the ultrasonic ranging sensor and the wall is occupied by a person, the presence or absence of a person can be determined.
[0047] This invention innovatively proposes a recognition technology based on multi-source data fusion and intelligent sensor analysis (i.e., a smoking behavior recognition method based on multi-source sensing and particle size fingerprint analysis). This technology can deeply analyze infrared sensing data streams and particulate matter concentration data streams, accurately identifying the presence of personnel and the unique fluctuation patterns of cigarette smoke. Through cross-validation of multi-dimensional features, the system can effectively eliminate interference from water vapor, mosquito coils, and non-human smoke sources, achieving a balance between high detection accuracy and low false alarm rate. Simultaneously, this solution completely eliminates the need for imaging equipment, fundamentally avoiding the risk of privacy leaks and meeting the market's urgent need for an innovative solution that provides continuous automated monitoring while ensuring privacy and security.
[0048] Specifically, the beneficial effects of the present invention are as follows:
[0049] 1) Addressing the issue of insufficient anti-interference capabilities and frequent false alarms in existing smoke detection technologies: This invention utilizes a novel algorithm to analyze the changes in particle size concentration distribution of particulate matter from different smoke sources, combined with the concurrent PM2.5 growth rate, to effectively distinguish between smoking and other non-cigarette smoke sources such as water vapor, mosquito repellent, and dust. Furthermore, based on the premise of human presence, it effectively eliminates pollution sources caused by human factors outside the designated space, thereby improving detection accuracy.
[0050] 2) Addressing the severe privacy compliance challenges faced by existing imaging-based monitoring methods in private scenarios: This invention completely avoids imaging-based monitoring technologies at the hardware design level, employing active single-point infrared detection sensors and environmental particulate matter sensors, significantly reducing users' privacy concerns and management's compliance risks. It is particularly suitable for restricted scenarios with extremely high privacy protection requirements, such as secure conference rooms, financial trading halls, and high-end hotel rooms.
[0051] 3) Addressing the inherent response lag and insufficient real-time early warning capability of traditional threshold triggering mechanisms: The novel algorithm logic of this invention makes a judgment by analyzing changes in particulate matter size and PM2.5 growth rate, and by observing fluctuations in concentration. This avoids the physical time delay required for particulate matter concentration to reach the preset threshold in traditional threshold triggering mechanisms, significantly improving the real-time response of the system, ensuring effective intervention in the early stages of smoking behavior, and enhancing the immediacy and effectiveness of management.
[0052] 4) Addressing the issue that the existing distributed hardware architecture drives up construction and maintenance costs and restricts large-scale deployment: The smoke source identification logic upon which this invention relies only needs to integrate a standard multi-channel particulate sensor, a low-cost infrared sensing unit, and a basic computing processing module, thereby eliminating the need for high-cost independent monitors and complex controllers. This streamlined hardware architecture significantly reduces the construction and life-cycle maintenance costs of a single location, providing a hardware foundation for low-cost, sustainable, and scalable large-scale deployment in large venues.
[0053] 5) Addressing the issue of low spatial resolution and lack of precise source tracing capabilities in traditional monitoring methods: This invention deploys a low-cost sensing logic unit in each independent micro-space (such as a toilet cubicle). While achieving rapid judgment, this system can automatically lock onto the specific source of the violation, providing high spatial resolution monitoring results and providing clear and reliable basis for subsequent precise point-to-point intervention and event tracing. Attached Figure Description
[0054] Figure 1 This is a schematic diagram of the average response time of the particle size concentration distribution method of the present invention and two traditional threshold methods;
[0055] Figure 2 This is a schematic diagram showing the time series changes in the average PM2.5 and PM1 concentration difference under different smoke sources. Detailed Implementation
[0056] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0057] The smoking behavior recognition method based on multi-source sensing and particle size fingerprint analysis in this embodiment includes:
[0058] A particulate matter sensor is set up to continuously acquire multi-channel particle size spectrum in the target space in an optical-free manner and generate a high-resolution particle size distribution sequence over time; preferably, a particulate matter laser scattering sensor is used to continuously monitor the concentration of particulate matter such as PM1, PM2.5, and PM10 in the air at high frequency.
[0059] Install active infrared sensors: monitor people entering and leaving in real time to ensure that detection is triggered only when people are present; preferably, use reflective active infrared detection sensors.
[0060] By calculating the short-term changes in the concentration distribution of particulate matter of different sizes using parameterized formulas and combining this with human status and the concurrent PM2.5 growth rate, accurate identification of smoking events can be achieved. This algorithm adapts to different environments through adjustable parameters, effectively distinguishing particulate matter from smoking from other interference sources (such as water vapor, mosquito repellent, dust, and other non-cigarette smoke sources), thus improving detection accuracy and reliability.
[0061] Specifically:
[0062] Set detection time point It is a discrete sequence, and the detection interval is... Second. : Boolean value, representing time The infrared sensor detected the presence of a person (true indicates that someone is present and has not left).
[0063] The following calculations were performed immediately after each detection of raw particulate matter concentration data of different particle sizes:
[0064] Particle size distribution index: ( The difference between PM2.5 and PM1 at any given time.
[0065] Particle size distribution index change rate: ( in the past Rate of change per second (It is a positive integer parameter).
[0066] Set of particle size distribution index change rates: (recent indivual value, It is a positive integer parameter.
[0067] When the detected data satisfies: ,
[0068] This is considered a violation of smoking regulations.
[0069] In the above formula: : is a Boolean value, which is the general premise for the judgment and indicates the presence of personnel.
[0070] If c out of the most recent n particle size distribution index change rates are greater than a certain standard measured in the laboratory ( If so, it can be determined that someone is smoking;
[0071] Or if all values of the most recent n particle size distribution index changes are greater than another standard measured in the laboratory ( This can also be interpreted as someone smoking;
[0072] Alternatively, all values of the most recent n particle size distribution index change rates are >0, and the increase in PM2.5 within the most recent k seconds is greater than the corresponding standard measured in the laboratory. This can also be interpreted as someone smoking;
[0073] in, It is an indicator function, when Greater than or equal to The value is 1 if it is true, and 0 otherwise. This represents logical AND. This indicates a logical OR.
[0074] The eigenvalues are explained below:
[0075] :recent indivual Among the values, at least The values satisfy ;
[0076] : A specific threshold for the exponential growth rate of particle size distribution concentration, recently indivual The value must include several or all values that are greater than this value;
[0077] A specific threshold for PM2.5 concentration growth rate; the increase in PM2.5 within the last k seconds must exceed this value.
[0078] (The formula is for illustrative purposes only. In actual applications, the quantization method of the feature value needs to be determined based on the sensor's output characteristics. Different sensors will output different data and have different detection speeds. Therefore, each sensor will have different feature values determined in the laboratory through the following two steps.)
[0079] The above characteristic values are measured and calculated through the following steps:
[0080] 1) Burn flue-cured cigarettes, blended cigarettes, cigars, heated tobacco products, e-cigarette vaporizing liquid, and common interfering sources (mosquito coils, essential oils, cooking fumes) in a standard experimental chamber and collect time-series spectra of particle size concentrations of different smoke sources.
[0081] 2) Repeat the experiment N≥30 times for each type of combustible material, and use the particle size distribution index: ( The difference between PM2.5 and PM1 at any given time and the rate of change of the particle size distribution index: ( in the past Rate of change per second (A positive integer parameter) is used to analyze and extract changes in the concentration distribution of smoke particles. The focus is on identifying the differences in the rate of change of particle size distribution between the target cigarette smoke and common interference sources (such as dust, water vapor, etc.). The characteristic values for practical applications are determined accordingly to improve the accuracy and robustness of detection.
[0082] pass Figure 2 As can be seen, the characteristics of cigarette smoke are very different from other types of smoke sources. The method described in this embodiment can easily identify smoking behavior.
[0083] Furthermore, smoking behavior identification and detection were performed using the method described in this embodiment (particle size concentration distribution method) and two traditional threshold methods, respectively. The identification results are shown in Table 1 below. Figure 1 As shown.
[0084]
[0085] Note: Both Traditional Threshold Method 1 and Traditional Threshold Method 2 are based on the detection of threshold-type smoke alarms using background technology: Traditional Threshold Method 1 refers to setting the threshold of the traditional threshold-type smoke alarm to twice the background PM2.5 value; Traditional Threshold Method 2 refers to setting the threshold of the traditional threshold-type smoke alarm to three times the background PM2.5 value.
[0086] Furthermore, regarding some alternative solutions to the method in this embodiment described above:
[0087] I. Alternative Solutions for Monitoring Personnel Entry and Exit Status: The core advantage of this invention lies in the accurate capture of personnel presence. Besides the optimal embodiment (active infrared sensor AIR, etc.), the following alternative solutions can achieve personnel status sensing gating:
[0088] 1) Wireless triggering mechanism based on door lock status: Integrate a wireless switch or sensing circuit in the door lock mechanism or door frame. The wireless signal is automatically triggered by the closed state of the lock tongue to realize low-power wake-up of the smoke detection module until the lock tongue is opened and the personnel leave the site.
[0089] 2) Passive infrared receiver sensor combined with millimeter-wave radar sensor: A passive infrared (PIR) sensor is used to capture changes in ambient thermal radiation when personnel enter, and this is combined with a millimeter-wave radar sensor for target presence (occupancy) and micro-motion detection. This fusion scheme can significantly improve the accuracy of personnel presence identification and the ability to resist false positives.
[0090] 3) Space Occupation Detection Based on Ultrasonic Ranging: By using ultrasonic ranging sensors to monitor changes in the sound wave propagation path within a confined space, the system identifies the presence and absence of personnel by recognizing the state of the effective space volume between the device and the wall being occupied by the personnel.
[0091] II. Alternative solutions for monitoring air pollution: Based on particulate matter size analysis (using particulate matter sensors), the following dimensions can be introduced as auxiliary verification of smoke source type:
[0092] By introducing chemical gas sensors (e.g., carbon monoxide (CO) and nitrogen oxides (NOx) sensors) as auxiliary sensing dimensions, and utilizing the concentration change characteristics of specific chemical components in cigarette combustion products, the cross-validation of particulate matter fluctuation characteristics is enhanced, thereby improving the confidence of smoke source type classification.
[0093] Furthermore, this invention realizes an intelligent smoke detection method based on a closed-loop management system of "perception-analysis-intervention-source tracing," the most direct product form of which is a highly integrated intelligent hardware device. This device can be embedded in the ceilings or walls of restrooms in public places such as buildings, campuses, and transportation hubs, enabling plug-and-play commercial deployment. Core functional components (such as multi-channel particulate matter sensors, non-imaging personnel status sensing units, and integrated acoustic-optical-electrical dissuasion output modules) can be integrated into a single compact housing, or they can interface with existing building automation systems (BMS / BAS) via a bus architecture, serving as an environmental health and safety monitoring node in smart buildings.
[0094] The system is based entirely on multi-source non-imaging sensors to identify smoke sources. It is suitable for places where image collection is prohibited, such as confidential units, traffic security checkpoints, judicial case-handling areas, and hotel rooms. It can replace traditional video surveillance-based smoking identification solutions, meeting regulatory requirements while ensuring privacy compliance.
[0095] In terms of industry affiliation, this invention spans three major fields: environmental monitoring instrument manufacturing, intelligent building system integration, and public health technical services. From a user scenario perspective, the core market includes industries such as education, healthcare, transportation, and commercial offices, and can be extended to enclosed spaces sensitive to secondhand smoke and requiring unattended management, such as hotels, shopping malls, and underground utility tunnels. Through a multi-layered product combination of "equipment—module—data," this invention can provide a low-cost, highly accurate, and privacy-friendly integrated solution for smoking monitoring and environmental improvement in the aforementioned scenarios without relying on image acquisition.
[0096] The present invention has been disclosed above with reference to preferred embodiments, but these are not intended to limit the present invention. Anyone skilled in the art can make various changes or modifications without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention should be defined by the scope of the claims in this application.
Claims
1. A smoking behavior recognition method based on multi-source sensing and particle size fingerprint analysis, characterized in that, The method includes: The particulate matter sensor module is configured to continuously acquire multi-channel particle size spectra in the target space in an optical-free manner and generate high-resolution particle size distribution sequences over time. An infrared sensor module is installed to monitor the entry and exit of personnel in real time, ensuring that detection is triggered only when personnel are present. By calculating the short-term changes in the concentration distribution of particulate matter of different particle sizes using parametric formulas, and combining this with human status and / or the concurrent PM2.5 growth rate, accurate judgment of smoking events can be achieved.
2. The smoking behavior recognition method based on multi-source sensing and particle size fingerprint analysis as described in claim 1, characterized in that, When the following conditions are met: , This is considered a violation of smoking regulations; in, It is a Boolean value, representing the time... The infrared sensor detected the presence of a person; if true, then someone is present and has not left. in, This represents logical AND. Indicates logical OR; in, It is an indicator function, when hour, It is 1 if it is true, otherwise it is 0. The eigenvalues are explained below: :recent indivual Among the values, at least The values satisfy ; : A specific threshold for the exponential growth rate of particle size distribution concentration, recently indivual The value must include several or all values that are greater than this value; A specific threshold for PM2.5 concentration growth rate; the increase in PM2.5 within the last k seconds must exceed this value.
3. The smoking behavior recognition method based on multi-source sensing and particle size fingerprint analysis as described in claim 2, characterized in that, The eigenvalue , It is obtained through the following steps: 1) Burn flue-cured cigarettes, blended cigarettes, cigars, heated tobacco products, e-cigarette vaporizing liquid, and common non-human-caused smoke sources in a standard experimental chamber and collect time-series spectra of particle size concentrations of different smoke sources. 2) Repeat the experiment N ≥ 30 times for each type of combustible material; Using the particle size distribution index: ,Right now The difference between PM2.5 and PM1 at any given time; And the rate of change of particle size distribution index: ,Right now in the past Rate of change per second It is a positive integer parameter; The analysis extracts changes in the particle size distribution of smoke sources, focusing on identifying the differences in the rate of change of particle size distribution between target cigarette smoke and common non-human-caused smoke sources, and then determines the characteristic values accordingly.
4. The smoking behavior recognition method based on multi-source sensing and particle size fingerprint analysis as described in claim 2, characterized in that, The following calculations are performed immediately after each detection of raw particle concentration data of different particle sizes by the particle sensor module: Particle size distribution index: ,Right now The difference between PM2.5 and PM1 at any given time; Particle size distribution index change rate: ,Right now in the past Rate of change per second It is a positive integer parameter; Set of particle size distribution index change rates: That is, recently indivual value, It is a positive integer parameter.
5. A smoking behavior recognition method based on multi-source sensing and particle size fingerprint analysis as described in any one of claims 1-4, characterized in that, The structure, including the particulate matter sensor module and the infrared sensor module, is integrated into a hardware device that can be embedded in the ceiling or wall of a public place to achieve plug-and-play functionality.
6. A smoking behavior recognition method based on multi-source sensing and particle size fingerprint analysis as described in any one of claims 1-4, characterized in that, The particulate matter sensor module includes a particulate matter laser scattering sensor, which continuously monitors the concentration of PM1, PM2.5 and PM10 particulate matter in the air at high frequency.
7. The smoking behavior recognition method based on multi-source sensing and particle size fingerprint analysis as described in claim 6, characterized in that, The particulate matter sensor module also includes a combustion heat source infrared camera and / or a chemical gas sensor.
8. A smoking behavior recognition method based on multi-source sensing and particle size fingerprint analysis as described in any one of claims 1-4, characterized in that, The infrared sensor module includes a reflective active infrared detection sensor.
9. A smoking behavior recognition method based on multi-source sensing and particle size fingerprint analysis as described in any one of claims 1-4, characterized in that, The infrared sensor module includes any one of the following: Integrate a wireless switch or sensor circuit into the door lock to automatically trigger a wireless signal based on the closing state of the lock tongue; Passive infrared receiver sensor in conjunction with millimeter-wave radar sensor; Ultrasonic ranging sensors monitor changes in the propagation path of sound waves within a confined space. By observing whether the effective space between the ultrasonic ranging sensor and the wall is occupied by a person, the presence or absence of a person can be determined.
