A student concentration non-inductive monitoring method, system and device based on gas sensing and a storage medium

Abstract

This invention discloses a non-invasive method, system, device, and storage medium for monitoring student attention based on gas sensing, belonging to the field of non-invasive sensing technology. The method includes: calculating a net physiological gas signal based on environmental baseline data and initial gas concentration data of the target student; calculating physiological characteristic indicators of the target student based on the net physiological gas signal, and determining the target student's metabolic rate level based on the physiological characteristic indicators; obtaining a state time-series signal based on the net physiological gas signal according to a preset discrimination criterion; inputting the net physiological gas signal and the state time-series signal into a pre-trained attention monitoring model to obtain the probability distribution of the target student's attention state; and determining the target student's attention state based on the target student's metabolic rate level and the probability distribution of the attention state. This invention assesses the cognitive state by capturing the gas released by the student, achieving non-invasive monitoring.

Description

Technical Field

[0001] This invention belongs to the field of non-invasive sensing technology, specifically relating to a non-intrusive monitoring method, system, device, and storage medium for student attention based on gas sensing. Background Technology

[0002] Student attention monitoring is a key technology for achieving personalized educational intervention and influencing learning outcomes. Its core lies in accurately determining the level of student concentration by collecting and analyzing various data generated during the learning process. Current mainstream attention monitoring technologies primarily rely on wearable sensors for physiological information collection and computer vision-based behavioral feature acquisition and analysis.

[0003] Traditional models, such as behavioral observation scales based on teacher observation, have laid the basic framework for assessment in this field, but they are highly subjective. With technological advancements, IoT sensing technology has been introduced to capture physiological data related to students and their learning, leading to various solutions based on wearable sensing devices and environmental awareness. Simultaneously, some research has begun to utilize computer vision and other technologies to capture explicit features such as facial expressions and eye movements during student learning for image analysis, attempting to improve the objectivity, rigor, and automation of attention monitoring. These methods have improved the accuracy of predictions to some extent.

[0004] Computer vision methods are highly dependent on the environment and lack privacy risks: existing technologies, such as CN120932278A, require the use of cameras to capture images of students' faces and bodies, which poses a significant risk of privacy breaches and could easily provoke resistance from students and parents. Furthermore, this technology is extremely sensitive to environmental conditions; changes in lighting and obstructions can affect the capture and analysis of facial features, leading to distorted attention scores and insufficient robustness.

[0005] Wearable sensors are highly invasive and lack universality: Existing technologies, such as CN120131020A, rely on functional near-infrared spectroscopy (fNIRS) head-mounted devices and other sensors, requiring students to wear specialized equipment. Prolonged wear increases discomfort for students, especially younger students, and can interfere with normal learning. Furthermore, the deployment and maintenance costs of these devices are high, hindering large-scale application; their physical intrusion is a fundamental and difficult-to-overcome drawback.

[0006] However, achieving truly seamless attention monitoring faces two major challenges: First, the difficulty of effectively capturing individual exhaled gases in an open classroom environment. The concentrations of marker gases such as acetone and isoprene exhaled by students are extremely low and rapidly diluted by free diffusion. Traditional sensors deployed in the corners of desks cannot distinguish individual signals from environmental noise, let alone isolate interference from classmates. Second, there is confusion between physiological indicators and cognitive states. For example, elevated acetone levels may stem from post-exercise metabolism, while elevated isoprene levels may reflect anxiety rather than focus. Therefore, without multimodal contextual verification, misjudgments are highly likely. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the prior art and provide a non-intrusive monitoring method, system, device and storage medium for student attention based on gas sensing. It can accurately extract attention-related features from weak and non-linear physiological gas signals to achieve high-precision, non-intrusive and real-time monitoring of individual student attention.

[0008] This invention provides the following technical solution:

[0009] In the first aspect, a method for monitoring student attention based on gas sensing is provided, including: calculating net physiological gas signals based on pre-acquired environmental baseline data and initial gas concentration data of the target student, wherein the net physiological gas signals include net carbon dioxide concentration sequence, net acetone concentration sequence and net isoprene concentration sequence;

[0010] Based on the net physiological gas signal, the physiological characteristic indicators of the target student are calculated, and the metabolic rate level of the target student is determined according to the physiological characteristic indicators.

[0011] Based on the net physiological gas signal, the state time sequence signal is obtained according to the preset discrimination criteria;

[0012] The net physiological gas signal and state time sequence signal are input into a pre-trained attention monitoring model to obtain the probability distribution of the target student's attention state;

[0013] The target students' attention levels are determined based on their metabolic rate levels and the probability distribution of their attention states.

[0014] The attention monitoring model includes a sequence fusion module, a temporal attention network module, and a fully connected classification network module.

[0015] The net physiological gas signal and the state time series signal are input into the sequence fusion module to obtain the multimodal fused time series signal;

[0016] The multimodal fused temporal signal is input into the temporal attention network module and encoded to obtain a deep feature vector;

[0017] The deep feature vector is input into the fully connected classification network module to obtain the probability distribution of the target student's attention state.

[0018] As a preferred embodiment of the present invention, the step of calculating the net physiological gas signal based on pre-acquired environmental baseline data and initial gas concentration data of the target student includes:

[0019] The environmental baseline data includes carbon dioxide concentration sequences, acetone concentration sequences, and isoprene concentration sequences in the target environment when no one is present.

[0020] The initial gas concentration data of the target student includes the carbon dioxide concentration sequence, acetone concentration sequence, and isoprene concentration sequence exhaled by the target student.

[0021] The net physiological gas signals are calculated and expressed as follows:

[0022] ;

[0023] ;

[0024] ;

[0025] in, This represents a sequence of net carbon dioxide concentrations. This represents the sequence of carbon dioxide concentrations exhaled by the target student. This represents the carbon dioxide concentration sequence in the target environment. This represents the net acetone concentration sequence. This represents the acetone concentration sequence exhaled by the target student. This represents the acetone concentration sequence in the target environment. This represents the net isoprene concentration sequence. This represents the isoprene concentration sequence exhaled by the target student. This represents the isoprene concentration sequence in the target environment.

[0026] As a preferred embodiment of the present invention, the step of calculating the physiological characteristic indicators of the target student based on net physiological gas signals and determining the metabolic rate level of the target student based on the physiological characteristic indicators includes:

[0027] The mean and standard deviation of the net acetone concentration sequence and the net isoprene concentration sequence were calculated respectively, and the results were used as physiological characteristic indicators of the target students.

[0028] Students with a mean of net acetone concentration less than 1 ppb and a mean of net isoprene concentration less than 0.5 ppb, or a mean of net acetone concentration less than 1 ppb and a standard deviation of net acetone concentration less than 60% of a pre-obtained reference value, or a mean of net isoprene concentration less than 0.5 ppb and a standard deviation of net isoprene concentration less than 60% of a pre-obtained reference value, are considered to have a low metabolic rate.

[0029] Students whose mean net acetone concentration sequence is in [1,5] and whose mean net isoprene concentration sequence is in [0.5,3], and whose standard deviations of the net acetone concentration sequence and net isoprene concentration sequence are in the range of [60%, 140%] of the corresponding reference values, are judged to have normal metabolic rate.

[0030] Students whose net acetone concentration sequence means are greater than 5 ppb, whose net isoprene concentration sequence means are greater than 3 ppb, and whose standard deviations for both net acetone and net isoprene concentration sequences are greater than 140% of the corresponding reference values ​​are considered to have high metabolic rates.

[0031] As a preferred embodiment of the present invention, the step of obtaining the state time sequence signal based on the net physiological gas signal according to a preset discrimination criterion includes:

[0032] The state time-series signals include high-confidence focus, motor metabolic interference, anxiety interference, and normal state;

[0033] If the standard deviation of the net carbon dioxide concentration sequence is less than the preset first threshold, the linear regression slope of the net acetone concentration sequence and the net isoprene concentration sequence is greater than zero, and the percentage of sampling points in the net acetone concentration sequence and the net isoprene concentration sequence that are in the same direction of concentration change exceeds 70% of the total sampling points, then it is judged as high confidence focus.

[0034] If only the growth slope of the net acetone concentration sequence is greater than the preset second threshold, and the standard deviation of the net carbon dioxide concentration sequence is greater than the preset first threshold, then it is judged as exercise metabolic interference.

[0035] If only the net isoprene concentration sequence is greater than the preset third threshold, and the standard deviation of the net acetone concentration sequence is less than the preset fourth threshold, then it is judged as anxiety interference;

[0036] If the standard deviations of the net carbon dioxide concentration sequence, net acetone concentration sequence, and net isoprene concentration sequence are all less than the corresponding preset thresholds, then it is judged to be in a normal state.

[0037] As a preferred embodiment of the present invention, the step of inputting the net physiological gas signal and the state time series signal into the sequence fusion module to obtain the multimodal fused time series signal includes:

[0038] The original gas time series signal was obtained by concatenating the net carbon dioxide concentration sequence, net acetone concentration sequence, and net isoprene concentration sequence, as follows:

[0039] ;

[0040] in, Represents the time-series signal of the original gas. This represents a sequence of net carbon dioxide concentrations. This represents the net acetone concentration sequence. Represents the net isoprene concentration sequence;

[0041] The state timing signal is numerically encoded to obtain the state timing encoded signal. (t);

[0042] The original gas time series signal and state timing encoded signal (t) is concatenated into a multimodal fused time-series signal, represented as:

[0043] ;

[0044] in, This represents a multimodal fusion timing signal. This indicates a splicing operation.

[0045] As a preferred embodiment of the present invention, the step of inputting the multimodal fused temporal signal into a temporal attention network module and encoding it to obtain a deep feature vector includes:

[0046] The multimodal fusion timing signal is divided into several overlapping time windows of fixed length;

[0047] Within each time window, the attention weights are calculated and represented as follows:

[0048] ;

[0049] ;

[0050] in, Represents the query vector. Represents the key vector. Represents a value vector. Indicates the first Time series data within a time window The weight matrix represents the query vector. The weight matrix represents the key vector. The weight matrix represents the value vector. Indicates the scaling factor. Indicates transpose. Represents the normalization function. Indicates attention weight;

[0051] The first The time series data within each time window and the corresponding attention weights are weighted to obtain the first... Window features ;

[0052] The Transformer shift window method is used to fuse the window features of all time windows, and finally the depth feature vector is obtained.

[0053] As a preferred embodiment of the present invention, determining the focus state of the target student based on the probability distribution of the target student's metabolic rate level and focus state includes:

[0054] Obtain the attention state threshold corresponding to the metabolic rate level of the target student;

[0055] Based on the attention state threshold and the probability distribution of attention states, the attention state of the target student is determined, expressed as:

[0056] ;

[0057] ;

[0058] in, Indicates the first The student's level of concentration This represents the probability distribution of attentional states. Represents the probability distribution of a state of focus. The value of the state of maximum focus. Indicates the first Classification confidence of each student This indicates taking the maximum value.

[0059] Secondly, a gas-sensing-based student attention monitoring system is provided, comprising:

[0060] The net data calculation module is used to calculate the net physiological gas signal based on the pre-acquired environmental baseline data and the initial gas concentration data of the target students. The net physiological gas signal includes the net carbon dioxide concentration sequence, the net acetone concentration sequence, and the net isoprene concentration sequence.

[0061] The metabolic rate level judgment module is used to calculate the physiological characteristic indicators of the target student based on the net physiological gas signal, and to judge the metabolic rate level of the target student based on the physiological characteristic indicators.

[0062] The discrimination module is used to determine the state time sequence signal according to the net physiological gas signal and a preset discrimination criterion.

[0063] The prediction module is used to input the net physiological gas signal and state time series signal into a pre-trained attention monitoring model to obtain the probability distribution of the target student's attention state.

[0064] The output module is used to determine the target student's attention state based on the probability distribution of the target student's metabolic rate level and attention state.

[0065] The attention monitoring model includes a sequence fusion module, a temporal attention network module, and a fully connected classification network module.

[0066] The net physiological gas signal and the state time series signal are input into the sequence fusion module to obtain the multimodal fused time series signal;

[0067] The multimodal fused temporal signal is input into the temporal attention network module and encoded to obtain a deep feature vector;

[0068] The deep feature vector is input into the fully connected classification network module to obtain the probability distribution of the target student's attention state.

[0069] Thirdly, a gas-sensing-based non-intrusive monitoring device for student attention is provided, comprising a processor and a storage medium; the storage medium is used to store instructions.

[0070] The processor is configured to operate according to the instructions to execute the gas-sensor-based non-intrusive monitoring method for student attention as described in the first aspect.

[0071] Fourthly, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the gas-sensor-based non-intrusive monitoring method for student attention as described in the first aspect.

[0072] Compared with the prior art, the beneficial effects of the present invention are:

[0073] This invention provides a non-invasive method for monitoring student attention based on gas sensing. It detects attention by using the gas exhaled by students, enabling non-invasive attention assessment with zero privacy risks. It accurately extracts attention-related features from weak and non-linear physiological gas signals, achieving high-precision, non-invasive, and real-time monitoring of individual student attention. Attached Figure Description

[0074] Figure 1 This is a schematic diagram illustrating the change of acetone concentration over time in an embodiment of the present invention;

[0075] Figure 2This is a schematic diagram illustrating the fluctuation of isoprene concentration under stress in an embodiment of the present invention;

[0076] Figure 3 This is a schematic diagram of the cognitive states of acetone and isoprene under different cognitive states in an embodiment of the present invention;

[0077] Figure 4 This is a flowchart of a gas-sensor-based method for non-intrusive monitoring of student attention in an embodiment of the present invention. Detailed Implementation

[0078] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0079] Example 1

[0080] During periods of focused cognitive activity, brain activity intensifies, leading to a significant increase in energy demand. When glucose supply is insufficient to meet this demand, the body accelerates fat breakdown for energy, a process that produces ketone bodies. Acetone is the most volatile component of ketone bodies and is released through respiration. Therefore, the concentration of acetone in a student's exhaled breath is positively correlated with the intensity of their brain's energy metabolism. Higher levels of focus result in more active fat breakdown and a higher concentration of acetone in exhaled breath; conversely, in states of distraction or fatigue, cognitive activity decreases, metabolic levels decline, and acetone concentration tends to decrease. Figure 2 As shown, the concentration of exhaled acetone during classroom cognitive tasks exhibits a dynamic change of "increasing when focused and decreasing when fatigued," reflecting the differences in energy metabolism under different cognitive states.

[0081] Isoprene is a natural byproduct of cholesterol synthesis in the human body, and its release rate is closely related to sympathetic nerve excitation and skeletal muscle tension. When students face cognitive challenges or psychological stress, the sympathetic nervous system is activated, triggering accelerated blood flow and micro-muscle contractions, thereby promoting the production and release of isoprene. A controlled chamber study in the published literature "Impact of Cognitive Tasks on CO2 and Isoprene Emissions from Humans" clearly showed through controlled experiments that subjects performing high-intensity cognitive tasks experienced a significant increase in isoprene emission rates compared to a relaxed state. This provides a solid experimental basis for using isoprene sensing to quantify students' cognitive engagement and stress levels in the classroom. Figure 2 As shown, the concentration of isoprene in the subjects was significantly increased under stress.

[0082] Furthermore, while CO2 concentration is easily influenced by the environment, at the individual level, its short-term fluctuations are closely related to respiratory rhythm and depth, indirectly reflecting an individual's state of relaxation or tension. Combining signals from volatile organic compounds (VOCs) such as acetone and isoprene can help determine respiratory metabolic patterns and signal validity. Post-exercise metabolism is often accompanied by rapid breathing, leading to abnormally high CO2 emission rates and concentrations in the short term. Breathing during cognitive tasks is usually relatively stable. Therefore, if acetone levels are significantly elevated while CO2 remains stable or fluctuates only slightly, it is more likely to originate from cognitive activity. Simultaneously, in a state of true focus, the brain's energy metabolism (acetone) and neural activation (isoprene) often occur in parallel. Figure 3 As shown, when students enter a state of high concentration, acetone and isoprene exhibit a synergistic upward trend. Conversely, if only isoprene rises sharply while acetone remains unchanged or changes asynchronously, it is more likely to indicate non-cognitive stress or anxiety.

[0083] Based on this, this embodiment provides a non-intrusive method for monitoring student attention based on gas sensing. For example... Figure 4 As shown, the specific steps include the following:

[0084] Step 1: Calculate the net physiological gas signal based on the pre-acquired environmental baseline data and the initial gas concentration data of the target students.

[0085] In this embodiment, a miniature multi-gas sensing unit is deployed in the inner left or right corner of the front edge of each student's desk (5–10 cm from the front edge of the desk and 5–10 cm from the side). Its air intake is positioned within a thermal plume enrichment zone 10–25 cm in front of the student's mouth and nose in a natural sitting posture, at a height of 20–25 mm. Combined with a micro-negative pressure suction mechanism, this efficiently captures the temporal concentration signals of specific volatile organic compounds (acetone, isoprene) and carbon dioxide (CO2) in the individual's exhaled breath, which are closely related to cognitive state. The acetone monitoring sensor uses a CCS811 MOX sensor with a molecular sieve filter membrane added to its inlet. The isoprene detection sensor uses a cross-sensitive array of TGS2600 and TGS2602 sensors. The CO2 detection sensor is a Senseair S8 LP sensor based on the NDIR principle.

[0086] The environmental baseline data includes the carbon dioxide concentration sequence, acetone concentration sequence, and isoprene concentration sequence in the target environment when no one is present. During confirmed periods of no teaching activity and when the classroom is empty (verified by CO2 concentration ≤ 450 ppm), the baseline data acquisition process is automatically initiated and lasts for 5 minutes. The median value is then used as the current environmental concentration sequence for carbon dioxide, acetone, and isoprene. A dynamic update algorithm is used to update the environmental baseline data. =0.95* +0.05* , This indicates the updated environmental baseline data. This represents the environmental baseline data before the update. This represents the environmental baseline data from the new test.

[0087] Net physiological gas signals include net carbon dioxide concentration sequences, net acetone concentration sequences, and net isoprene concentration sequences.

[0088] The initial gas concentration data of the target student includes the carbon dioxide concentration sequence, acetone concentration sequence, and isoprene concentration sequence exhaled by the target student.

[0089] The net physiological gas signals are calculated and expressed as follows:

[0090] ;

[0091] ;

[0092] ;

[0093] in, This represents a sequence of net carbon dioxide concentrations. This represents the sequence of carbon dioxide concentrations exhaled by the target student. This represents the carbon dioxide concentration sequence in the target environment. This represents the net acetone concentration sequence. This represents the acetone concentration sequence exhaled by the target student. This represents the acetone concentration sequence in the target environment. This represents the net isoprene concentration sequence. This represents the isoprene concentration sequence exhaled by the target student. This represents the isoprene concentration sequence in the target environment.

[0094] Step 2: Based on the net physiological gas signal, calculate the physiological characteristic indicators of the target student, and determine the metabolic rate level of the target student based on the physiological characteristic indicators.

[0095] To address sudden, transient non-physiological disturbances (such as sudden perfume or food odors), a statistical anomaly detection and repair mechanism is introduced. A sliding window of length M seconds is maintained for each gas channel, and the average value μ and standard deviation σ of the signal within the window are calculated in real time. A dynamic threshold K is set; if the signal value at the current time t satisfies | If -μ|>Kσ, then the point is determined to be a sudden interference point. This represents the net physiological gas signal. For the marked interference points, linear interpolation is performed using N normal data points before and after them, effectively eliminating interference from sudden outliers.

[0096] If both conditions are met simultaneously—that the carbon dioxide concentration sequence exhaled by the target student is consistently close to or lower than the carbon dioxide concentration sequence in the target environment, and that the standard deviation of the acetone concentration sequence is lower than a threshold—and this condition is maintained for a preset duration (e.g., 60 seconds), then the student's sensing unit is determined to be in a physically obstructed state. Once physical obstruction is determined, the student's attention calculation and output will be suspended.

[0097] The mean and standard deviation of the net acetone concentration sequence and the net isoprene concentration sequence were calculated respectively, and the results were used as physiological characteristic indicators of the target students.

[0098] Students with a net acetone concentration sequence mean less than 1 ppb and a net isoprene concentration sequence mean of 0.5 ppb, or a net acetone concentration sequence mean less than 1 ppb and a standard deviation of the net acetone concentration sequence less than 60% of a pre-obtained reference value, or a net isoprene concentration sequence mean less than 0.5 ppb and a standard deviation of the net isoprene concentration sequence less than 60% of a pre-obtained reference value, were classified as having a low metabolic rate. The reference value is the arithmetic mean of the standard deviations of the corresponding gas concentration signals for all students in the current class.

[0099] Students whose mean net acetone concentration sequence is in the range of [1, 5] and net isoprene concentration sequence is in the range of [0.5, 3], and whose standard deviations of net acetone concentration sequence and net isoprene concentration sequence are in the range of [60%, 140%] of the corresponding reference values, are judged to have normal metabolic rate.

[0100] Students whose net acetone concentration sequence means are greater than 5 ppb, whose net isoprene concentration sequence means are greater than 3 ppb, and whose standard deviations for both net acetone and net isoprene concentration sequences are greater than 140% of the corresponding reference values ​​are considered to have high metabolic rates.

[0101] Step 3: Based on the net physiological gas signal, the state timing signal is obtained according to the preset discrimination criteria.

[0102] The state timing signals include high-confidence focus, exercise-metabolic interference, anxiety interference, and normal state.

[0103] If the standard deviation of the net carbon dioxide concentration sequence is less than the preset first threshold, the linear regression slope of the net acetone concentration sequence and the net isoprene concentration sequence is greater than zero, and the percentage of sampling points in the net acetone concentration sequence and the net isoprene concentration sequence that are in the same direction of concentration change exceeds 70% of the total sampling points, then it is judged as high confidence focus.

[0104] If only the growth slope of the net acetone concentration sequence is greater than the preset second threshold, and the standard deviation of the net carbon dioxide concentration sequence is greater than the preset first threshold, then it is judged as exercise metabolic interference.

[0105] If only the net isoprene concentration sequence is greater than the preset third threshold, and the standard deviation of the net acetone concentration sequence is less than the preset fourth threshold, then it is judged as anxiety interference.

[0106] If the standard deviations of the net carbon dioxide concentration sequence, net acetone concentration sequence, and net isoprene concentration sequence are all less than the corresponding preset thresholds, then it is judged to be in a normal state.

[0107] Step 4: Input the net physiological gas signal and state time sequence signal into the pre-trained attention monitoring model to obtain the probability distribution of the target student's attention state.

[0108] The attention monitoring model includes a sequence fusion module, a temporal attention network module, and a fully connected classification network module.

[0109] The net physiological gas signal and state time-series signal are input into the sequence fusion module to obtain a multimodal fused time-series signal. This includes:

[0110] The original gas time series signal was obtained by concatenating the net carbon dioxide concentration sequence, net acetone concentration sequence, and net isoprene concentration sequence, as follows:

[0111] ;

[0112] in, Represents the time-series signal of the original gas. This represents a sequence of net carbon dioxide concentrations. This represents the net acetone concentration sequence. This represents the net isoprene concentration sequence.

[0113] The state timing signal is numerically encoded to obtain the state timing encoded signal. (t). In this embodiment, the high-confidence focus candidate is encoded as [1, 0, 0, 0], the exercise-metabolic interference is encoded as [0, 1, 0, 0], the anxiety interference is encoded as [0, 0, 1, 0], and the normal baseline is encoded as [0, 0, 0, 1].

[0114] The original gas time series signal and state timing encoded signal (t) is concatenated into a multimodal fused time-series signal, represented as:

[0115] ;

[0116] in, This represents a multimodal fusion timing signal. This indicates a splicing operation.

[0117] The multimodal fused temporal signal is input into a temporal attention network module and encoded to obtain a deep feature vector. This includes:

[0118] After representing the multimodal fused timing signal in matrix form, it is divided into several overlapping time windows of fixed length.

[0119] Within each time window, the attention weights are calculated and represented as follows:

[0120] ;

[0121] ;

[0122] in, Represents the query vector. Represents the key vector. Represents a value vector. Indicates the first Time series data within a time window The weight matrix represents the query vector. The weight matrix represents the key vector. The weight matrix represents the value vector. Indicates the scaling factor. Indicates transpose. Represents the normalization function. This represents the attention weight.

[0123] The first The time series data within each time window and the corresponding attention weights are weighted to obtain the first... Window features .

[0124] The Transformer method with a shifted window is used to fuse window features from all time windows, ultimately obtaining a depth feature vector. Specifically, in the next layer, the window is shifted to the lower right. Each time step allows adjacent windows to overlap, thus enabling information interaction.

[0125] The deep feature vector is input into a fully connected classification network module to obtain the probability distribution of the target student's attention state. This includes:

[0126] The probability distribution of the target student's attention state is represented as follows:

[0127] ;

[0128] ;

[0129] in, This represents the initial score for each level of focus. Indicates weight, Indicates bias. This represents the probability distribution of attentional states.

[0130] Step 5: Determine the target student's attention level based on the probability distribution of the target student's metabolic rate and attention state.

[0131] Obtain the attention state threshold corresponding to the target student's metabolic rate level. Attention state includes highly focused, normal, distracted, and fatigued. The attention state threshold varies at different metabolic rate levels.

[0132] Based on the attention state threshold and the probability distribution of attention states, the attention state of the target student is determined, expressed as:

[0133] ;

[0134] ;

[0135] in, Indicates the first The student's level of concentration This represents the probability distribution of attentional states. Represents the probability distribution of a state of focus. The value of the state of maximum focus. Indicates the first Classification confidence of each student This indicates taking the maximum value.

[0136] Example 2

[0137] This embodiment provides a gas-sensor-based non-intrusive monitoring system for student attention, including:

[0138] The net data calculation module is used to calculate the net physiological gas signal based on the pre-acquired environmental baseline data and the initial gas concentration data of the target students. The net physiological gas signal includes the net carbon dioxide concentration sequence, the net acetone concentration sequence, and the net isoprene concentration sequence.

[0139] The metabolic rate level judgment module is used to calculate the physiological characteristic indicators of the target student based on the net physiological gas signal, and to judge the metabolic rate level of the target student based on the physiological characteristic indicators.

[0140] The discrimination module is used to determine the state time sequence signal according to the net physiological gas signal and a preset discrimination criterion.

[0141] The prediction module is used to input the net physiological gas signal and state time series signal into a pre-trained attention monitoring model to obtain the probability distribution of the target student's attention state.

[0142] The output module is used to determine the target student's attention state based on the probability distribution of the target student's metabolic rate level and attention state.

[0143] The attention monitoring model includes a sequence fusion module, a temporal attention network module, and a fully connected classification network module.

[0144] The net physiological gas signal and the state time series signal are input into the sequence fusion module to obtain the multimodal fused time series signal;

[0145] The multimodal fused temporal signal is input into the temporal attention network module and encoded to obtain a deep feature vector;

[0146] The deep feature vector is input into the fully connected classification network module to obtain the probability distribution of the target student's attention state.

[0147] Example 3

[0148] This embodiment provides a non-intrusive monitoring device for student attention based on gas sensing, including a processor and a storage medium; the storage medium is used to store instructions;

[0149] The processor is configured to operate according to the instructions to execute the gas-sensor-based non-intrusive monitoring method for student attention as described in the first aspect.

[0150] Example 4

[0151] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the gas-sensor-based non-intrusive monitoring method for student attention as described in Embodiment 1.

[0152] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0153] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0154] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0155] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0156] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for monitoring student concentration without awareness based on gas sensing, characterized in that, include: Based on the pre-acquired environmental baseline data and the initial gas concentration data of the target students, the net physiological gas signal is calculated. The net physiological gas signal includes the net carbon dioxide concentration sequence, the net acetone concentration sequence, and the net isoprene concentration sequence. Based on the net physiological gas signal, the physiological characteristic indicators of the target student are calculated, and the metabolic rate level of the target student is determined according to the physiological characteristic indicators. Based on the net physiological gas signal, the state time sequence signal is obtained according to the preset discrimination criteria; The net physiological gas signal and state time sequence signal are input into a pre-trained attention monitoring model to obtain the probability distribution of the target student's attention state; The target students' attention levels are determined based on their metabolic rate levels and the probability distribution of their attention states. The attention monitoring model includes a sequence fusion module, a temporal attention network module, and a fully connected classification network module. The net physiological gas signal and the state time series signal are input into the sequence fusion module to obtain the multimodal fused time series signal; The multimodal fused temporal signal is input into the temporal attention network module and encoded to obtain a deep feature vector; The deep feature vector is input into the fully connected classification network module to obtain the probability distribution of the target student's attention state.

2. The student concentration unobtrusive monitoring method based on gas sensing of claim 1, wherein, The calculation of net physiological gas signals based on pre-acquired environmental baseline data and initial gas concentration data of the target students includes: The environmental baseline data includes carbon dioxide concentration sequences, acetone concentration sequences, and isoprene concentration sequences in the target environment when no one is present. The initial gas concentration data of the target student includes the carbon dioxide concentration sequence, acetone concentration sequence, and isoprene concentration sequence exhaled by the target student. The net physiological gas signals are calculated and expressed as follows: ； ； ； in, This represents a sequence of net carbon dioxide concentrations. This represents the sequence of carbon dioxide concentrations exhaled by the target student. This represents the carbon dioxide concentration sequence in the target environment. This represents the net acetone concentration sequence. This represents the acetone concentration sequence exhaled by the target student. This represents the acetone concentration sequence in the target environment. This represents the net isoprene concentration sequence. This represents the isoprene concentration sequence exhaled by the target student. This represents the isoprene concentration sequence in the target environment.

3. The non-intrusive student attention monitoring method based on gas sensing according to claim 1, characterized in that, The process of calculating the target student's physiological characteristic indicators based on net physiological gas signals and determining the target student's metabolic rate level based on these indicators includes: The mean and standard deviation of the net acetone concentration sequence and the net isoprene concentration sequence were calculated respectively, and the results were used as physiological characteristic indicators of the target students. Students with a mean of net acetone concentration less than 1 ppb and a mean of net isoprene concentration less than 0.5 ppb, or a mean of net acetone concentration less than 1 ppb and a standard deviation of net acetone concentration less than 60% of a pre-obtained reference value, or a mean of net isoprene concentration less than 0.5 ppb and a standard deviation of net isoprene concentration less than 60% of a pre-obtained reference value, are considered to have a low metabolic rate. Students whose mean net acetone concentration sequence is in [1,5] and whose mean net isoprene concentration sequence is in [0.5,3], and whose standard deviations of the net acetone concentration sequence and net isoprene concentration sequence are in the range of [60%, 140%] of the corresponding reference values, are judged to have normal metabolic rate. Students whose net acetone concentration sequence means are greater than 5 ppb, whose net isoprene concentration sequence means are greater than 3 ppb, and whose standard deviations for both net acetone and net isoprene concentration sequences are greater than 140% of the corresponding reference values ​​are considered to have high metabolic rates.

4. The non-intrusive student attention monitoring method based on gas sensing according to claim 1, characterized in that, The process of obtaining the state time-series signal based on the net physiological gas signal and according to a preset discrimination criterion includes: The state time-series signals include high-confidence focus, motor metabolic interference, anxiety interference, and normal state; If the standard deviation of the net carbon dioxide concentration sequence is less than the preset first threshold, the linear regression slope of the net acetone concentration sequence and the net isoprene concentration sequence is greater than zero, and the percentage of sampling points in the net acetone concentration sequence and the net isoprene concentration sequence that are in the same direction of concentration change exceeds 70% of the total sampling points, then it is judged as high confidence focus. If only the growth slope of the net acetone concentration sequence is greater than the preset second threshold, and the standard deviation of the net carbon dioxide concentration sequence is greater than the preset first threshold, then it is judged as exercise metabolic interference. If only the net isoprene concentration sequence is greater than the preset third threshold, and the standard deviation of the net acetone concentration sequence is less than the preset fourth threshold, then it is judged as anxiety interference; If the standard deviations of the net carbon dioxide concentration sequence, net acetone concentration sequence, and net isoprene concentration sequence are all less than the corresponding preset thresholds, then it is judged to be in a normal state.

5. The non-intrusive student attention monitoring method based on gas sensing according to claim 1, characterized in that, The step of inputting net physiological gas signals and state time-series signals into a sequence fusion module to obtain a multimodal fused time-series signal includes: The original gas time series signal was obtained by concatenating the net carbon dioxide concentration sequence, net acetone concentration sequence, and net isoprene concentration sequence, as follows: ； in, Represents the time-series signal of the original gas. This represents a sequence of net carbon dioxide concentrations. This represents the net acetone concentration sequence. Represents the net isoprene concentration sequence; The state timing signal is numerically encoded to obtain the state timing encoded signal. (t); The original gas time series signal and state timing encoded signal (t) is concatenated into a multimodal fused time-series signal, represented as: ； in, This represents a multimodal fusion timing signal. This indicates a splicing operation.

6. The non-intrusive student attention monitoring method based on gas sensing according to claim 1, characterized in that, The step of inputting the multimodal fused temporal signal into the temporal attention network module and encoding it to obtain a deep feature vector includes: The multimodal fusion timing signal is divided into several overlapping time windows of fixed length; Within each time window, the attention weights are calculated and represented as follows: ； ； in, Represents the query vector. Represents the key vector. Represents a value vector. Indicates the first Time series data within a time window The weight matrix represents the query vector. The weight matrix represents the key vector. The weight matrix represents the value vector. Indicates the scaling factor. Indicates transpose. Represents the normalization function. Indicates attention weight; The first The time series data within each time window and the corresponding attention weights are weighted to obtain the first... Window features ; The Transformer shift window method is used to fuse the window features of all time windows, and finally the depth feature vector is obtained.

7. The non-intrusive student attention monitoring method based on gas sensing according to claim 3, characterized in that, Determining the target student's attention state based on the probability distribution of the target student's metabolic rate level and attention state includes: Obtain the attention state threshold corresponding to the metabolic rate level of the target student; Based on the attention state threshold and the probability distribution of attention states, the attention state of the target student is determined, expressed as: ； ； in, Indicates the first The student's level of concentration This represents the probability distribution of attentional states. Represents the probability distribution of a state of focus. The value of the state of maximum focus. Indicates the first Classification confidence of each student This indicates taking the maximum value.

8. A non-intrusive student attention monitoring system based on gas sensing, characterized in that, include: The net data calculation module is used to calculate the net physiological gas signal based on the pre-acquired environmental baseline data and the initial gas concentration data of the target students. The net physiological gas signal includes the net carbon dioxide concentration sequence, the net acetone concentration sequence, and the net isoprene concentration sequence. The metabolic rate level judgment module is used to calculate the physiological characteristic indicators of the target student based on the net physiological gas signal, and to judge the metabolic rate level of the target student based on the physiological characteristic indicators. The discrimination module is used to determine the state time sequence signal according to the net physiological gas signal and a preset discrimination criterion. The prediction module is used to input the net physiological gas signal and state time series signal into a pre-trained attention monitoring model to obtain the probability distribution of the target student's attention state. The output module is used to determine the target student's attention state based on the probability distribution of the target student's metabolic rate level and attention state. The attention monitoring model includes a sequence fusion module, a temporal attention network module, and a fully connected classification network module. The net physiological gas signal and the state time series signal are input into the sequence fusion module to obtain the multimodal fused time series signal; The multimodal fused temporal signal is input into the temporal attention network module and encoded to obtain a deep feature vector; The deep feature vector is input into the fully connected classification network module to obtain the probability distribution of the target student's attention state.

9. A non-intrusive monitoring device for student attention based on gas sensing, characterized in that, Including processor and storage media; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the gas-sensor-based non-intrusive monitoring method for student attention as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the gas-sensor-based non-intrusive monitoring method for student attention as described in any one of claims 1-7.

Images