Elevator safety teaching training system

Abstract

The invention discloses an elevator safety teaching practical training system, and relates to the technical field of elevator safety education practical training simulation. Comprising elevator safety teaching equipment, a theory learning module, a physiology monitoring module and a teaching practical training analysis module. According to the elevator safety teaching equipment, elevator operation and emergencies are simulated through the lift car and the dynamic platform, and safety knowledge learning is carried out through the theory learning module; the physiological monitoring module collects heart rate, breathing, eye movement, skin electricity and posture signals in a physiological monitoring mode, and the concentration degree, tension degree and operation performance of a learner are analyzed through the teaching practical training analysis module in combination with image recognition. And the teaching practical training analysis module dynamically evaluates the learning effect based on the comprehensive index, adjusts the scene level, and forms a personalized practical training path. According to the invention, real experience and objective evaluation are realized, so that the safety awareness and emergency disposal capability of learners are improved.

Description

Technical Field

[0001] This invention relates to the field of elevator safety education and training simulation technology, specifically an elevator safety teaching and training system. Background Technology

[0002] With the rapid pace of urbanization, elevators have become an essential mode of transportation for daily life, and their safe use and emergency response capabilities are directly related to the safety of people and property. Currently, elevator safety education is mainly conducted through classroom lectures, video demonstrations, and promotional materials. While these methods can disseminate some safety knowledge, they have the following shortcomings:

[0003] Traditional teaching methods rely mainly on text, pictures, or videos, making it difficult for students to experience elevator operation and emergencies in a safe environment, thus limiting the learning effect.

[0004] Current teaching methods rely heavily on written tests or oral question-and-answer assessments, lacking objective monitoring and evaluation of students' attention, psychological state, and operational behavior.

[0005] In practical training, reliance on manual observation and supervision often leads to the problem of risky behaviors not being detected and corrected in a timely manner. Summary of the Invention

[0006] In order to solve the problems mentioned in the background art, the present invention proposes an elevator safety teaching and training system.

[0007] The objective of this invention can be achieved through the following technical solution: an elevator safety teaching and training system, comprising elevator teaching equipment, a theoretical learning module, a physiological monitoring module, and a teaching and training analysis module;

[0008] The elevator teaching equipment is used to simulate elevator operation and emergency scenarios through a 1:1 car, a motion platform, and a multimedia unit.

[0009] The theoretical learning module is applied to the elevator car itself to carry out theoretical learning related to elevator safety.

[0010] The physiological monitoring module acquires physiological data of learners inside the elevator car and during safety training; it analyzes the physiological data to obtain the theoretical learning completion rate of each learner; based on the obtained theoretical learning completion rate and distribution, it analyzes and outputs the safety training path for each learner.

[0011] The teaching and training analysis module receives each safe training path for training simulation: before the training simulation, the safe training path is displayed via LEDs, and the physiological monitoring module obtains training physiological data, analyzes it to obtain training physiological reception values, and obtains the initial scenario level of the simulated training based on the safe training path and training physiological reception values; based on the obtained initial scenario level of the simulated training, the training simulation is carried out, and image data of each learner is acquired during the training simulation. After analyzing the image data of each learner in the simulated training, the training completion rate of each learner is obtained; then the scenario level is dynamically adjusted according to the training completion rate.

[0012] In a preferred embodiment of the present invention, the elevator teaching device includes:

[0013] The car body is equipped with floor buttons, alarm devices, and handrail pressure sensors.

[0014] The visual unit projects real-time images of high-rise buildings onto a small-pitch LED screen, synchronized with changes in floor level.

[0015] The motion platform is used to simulate elevator falls and roof collapses.

[0016] The sound effects unit includes a hidden array of speakers mounted on the top of the car body, used for directional playback of wire breaking / metal friction sounds.

[0017] As a preferred embodiment of the present invention, the process of analyzing physiological data to obtain the theoretical learning completion rate of each learner is as follows:

[0018] Resting data from the 60 seconds prior to playing the theoretical learning content was used as the baseline segment for data parameters. The theoretical physiological data obtained from the baseline segment underwent data preprocessing. After preprocessing, the entire theoretical learning process was divided into continuous sliding time windows. Within each window, features were extracted from the preprocessed multimodal signals, including eye-tracking features, heart rate features, skin conductance features, and respiratory features, and these were denoted as a set of numerical features. These features were used as direct inputs for subsequent baseline normalization and Gaussian mapping to obtain quantized feature parameters. The extracted theoretical physiological data features were standardized using Z-scores based on the baseline and mapped to the [0, 1] interval using a Gaussian target band function. Attention level values ​​were statistically obtained based on the physiological features within each time window. The multidimensional feature attention values ​​were treated as a set, and the average attention level sequence of all windows was correlated with the time progression of the theoretical learning task. The overall learning completion rate was obtained through time integration. A probability distribution was constructed from the attention mean of each window. The attention mean was divided into several intervals, and the occurrence probability was calculated by counting the number of windows in each interval. The learning completion rate distribution was obtained based on the distribution of attention level values ​​across different intervals.

[0019] In a preferred embodiment of the present invention, the specific process of extracting features from the preprocessed multimodal signal within each window is as follows:

[0020] For eye-tracking features: the total duration of all gaze events is calculated and divided by the window length L to obtain the gaze duration percentage; then, the number of times the subject's gaze point falls into the AOI is counted and divided by the total number of gaze points to obtain the region of interest coverage; let the coordinates of all gaze points be (x... i y i ), whose mean is The gaze discreteness is obtained; blink events are detected by eyelid sensors, the number of blinks within the window is counted, and the blink frequency is obtained by dividing by the window length; let the pupil diameter time series within the window be p(t), then the pupil dilation amplitude is obtained;

[0021] For heart rate characteristics: the mean RR interval within the window. Obtain the corresponding heart rate HR (W) k Heart rate deviation f is obtained from the corresponding heart rate HR. HR-dev (W k ); in window W k Within, the RR interval sequence is {RR i} Obtain the variability f HRV (W k );

[0022] For skin conductance characteristics, the difference between the maximum and minimum values ​​of the SCR component is used to obtain the amplitude of skin conductance change.

[0023] For respiratory characteristics: the dominant respiratory rate BR(W) within the window is obtained through the peak power spectrum. k The change in respiratory rate f was obtained. BR (W k );

[0024] For head pose features, the head angular velocity vector ω(t) and stability index f are obtained. head (W k ) represents the root mean square of the head angular velocity vector; the forward gaze threshold angle θ is defined. thr Within the statistics window, the angle by which the head deviates from directly forward is less than θ. thr The time ratio is used to obtain the forward gaze ratio f front (W k ).

[0025] As a preferred embodiment of the present invention, the specific process of analyzing and outputting the safe training path for each learner based on the obtained theoretical learning completion rate and theoretical learning completion rate distribution is as follows:

[0026] Learning levels are categorized based on completion score intervals: A completion score less than 0.4 but greater than or equal to 0 is considered low; a completion score less than 0.7 but greater than or equal to 0.4 is considered medium; and a completion score less than or equal to 1.0 but greater than or equal to 0.7 is considered high. Further, the completion score distribution P... 分布 Determining learning stability: If the distribution is concentrated in the high attention interval [0.6, 1.0], the learning process is considered stable; otherwise, it is unstable. This is determined by obtaining the completion rate C and the completion rate distribution P. 分布 Then, the individual status of learners is analyzed, and personalized safety training paths are output: if the completion rate is less than 0.4 and the distribution P 分布 If the proportion of low to medium attention intervals exceeds 50%, it is considered a low-level training path; if the completion rate is less than 0.7, or the completion rate is greater than or equal to 0.4 and the distribution P... 分布 If the medium attention interval accounts for 70%, then output a medium-level training path; if the completion rate is greater than or equal to 0.7 and the distribution P 分布 If the training remains stable within the high attention range, it represents an advanced level of practical training.

[0027] As a preferred embodiment of the present invention, the specific process of obtaining the initial scenario level of the simulated training according to the safety training path and the physiological reception value of the training is as follows:

[0028] If the training physiological reception value is less than 0.4, it is judged as a low-level scenario; if the training physiological reception value is less than 0.7 and the training physiological reception value is greater than or equal to 0.4, it is judged as a medium-level scenario; if the training physiological reception value is greater than or equal to 0.7, it is judged as a high-level scenario; and the initial scenario level is recorded as the initial scenario level.

[0029] As a preferred embodiment of the present invention, the specific process of analyzing the image data of each learner in simulated training to obtain the training completion rate of each learner is as follows:

[0030] The image is processed using an image recognition algorithm to form a window W. k Feature set within: D img (Wk)={F(W k ), B(W k ), R(W k )};wherein, F(W k ) represents facial expression parameters, B(W) k R(W) represents the posture operation parameters. k The parameters for risk behavior identification are: facial expression parameters, posture operation parameters, and risk behavior identification parameters. Then, the image completion rate is calculated based on the obtained facial expression parameters, posture operation parameters, and risk behavior identification parameters.

[0031] In a preferred embodiment of the present invention, the calculation process for facial expression parameters, posture operation parameters, and risk behavior recognition parameters is as follows:

[0032] Facial expression parameters are used to extract the tension index I through a facial recognition model. tense Fatigue Index I fatigue Focus Index I focus A comprehensive calculation is performed using the formula: get;

[0033] Fatigue index extraction accuracy rate through human key point identification I acc Action completion rate I comp Attitude stability I stab A comprehensive calculation is performed using the formula: get;

[0034] Risk behavior identification parameters are extracted using a deep convolutional neural network and temporal modeling methods to capture the trajectory features of key human points and the temporal features of actions of the learner. The extracted features are compared with a pre-established standard operating behavior database. If the matching result deviates from the standard operating sequence, a risky behavior is identified. The system outputs the learner's actions within the current time window W. k The probability of an unsafe action occurring; based on the obtained probability P of the unsafe action. risk (W k The risk behavior identification parameter R(W) was obtained through analysis. k ).

[0035] As a preferred embodiment of the present invention, the specific process of dynamically adjusting the scenario level based on the completion rate of the practical training is as follows:

[0036] Then, acquire physiological data during the training simulation process, record it as process physiological data, and analyze it to obtain process physiological data status values; combine the image completion degree and physiological data status values ​​to obtain the training completion degree; dynamically adjust the scenario level according to the training completion degree. If the training completion degree is greater than or equal to 0.7, upgrade by one level and execute a new safety training path; if the training completion degree is less than or equal to 0.3, downgrade by one level and repeat the safety training.

[0037] Compared with the prior art, the beneficial effects of the present invention are:

[0038] 1. This invention constructs an immersive teaching environment using a 1:1 scale real elevator car, dynamic platform, LED display and sound system. Combined with multimodal physiological monitoring such as heart rate, respiration, eye movement, skin conductance, and posture, it quantitatively analyzes the learner's attention, tension and operational status. Compared with existing single courseware or traditional simulations, this invention enhances the learner's sense of presence and safety awareness, and ensures the objectivity and scientific nature of the teaching effect.

[0039] 2. This invention adjusts the scenario level in real time based on the learners' completion of theoretical learning and practical training, outputting personalized safety training paths; it provides progressive training from low to high levels, and repeats training at lower levels when necessary. This invention overcomes the shortcomings of traditional progress methods, making the teaching process individualized, ensuring learners safely master basic knowledge, improving emergency response capabilities, and enhancing teaching efficiency and training quality. Attached Figure Description

[0040] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0041] Figure 1 This is a schematic diagram of the principle of the present invention. Detailed Implementation

[0042] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0043] It should be understood that the terms “comprising” and “including” used in this disclosure and claims indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0044] It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this disclosure. As used in this disclosure and claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this disclosure and claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations.

[0045] Please see Figure 1 As shown, an elevator safety teaching and training system includes: elevator safety teaching equipment, a theoretical learning module, a physiological monitoring module, and a teaching and training analysis module.

[0046] Elevator safety training equipment is used to simulate elevator operation and emergency scenarios. It includes a 1:1 scale elevator car body equipped with floor buttons, alarm devices, and handrail pressure sensors. It also includes a vision unit, a motion platform, and a sound effects unit. The vision unit projects real-life images of high-rise buildings onto a small-pitch LED screen, synchronized with floor changes. The motion platform simulates the sensation of falling (0.3G acceleration) / roof collapse (vertical vibration 15Hz) using electric cylinders. The sound effects unit includes a hidden array of speakers installed on the top of the car, directionally playing the sound of steel wire breaking / metal friction (sound pressure level 85dB).

[0047] The theoretical learning module is applied within the elevator car itself to conduct theoretical learning related to elevator safety. The learning content is displayed on the LED screen built into the car, including elevator safety knowledge, riding regulations, and key points for handling emergencies, allowing learners to conduct systematic learning within the elevator environment.

[0048] The physiological monitoring module acquires physiological data of learners in the elevator car and during safety teaching and training. The physiological data includes indicators such as heart rate, respiratory rate, eye movement trajectory and body surface temperature.

[0049] The physiological data were analyzed to determine the theoretical learning completion rate of each learner, specifically:

[0050] Resting data for 60 seconds before playing the theoretical learning content is used as the baseline segment for parameters such as heart rate, skin conductance, and pupil diameter: B = [t0-60, t0), where t0 represents the time point when the theoretical learning officially begins;

[0051] Based on the baseline segment, for any original signal x(t), the mean and standard deviation on the baseline segment are defined as follows: Where median(·) is the median operation and MAD(·) is the median absolute deviation;

[0052] The theoretical learning process is divided into multiple sliding windows, denoted as: W k =[τ k , τ k +L), where L represents the window length, τ k This indicates the starting time of the k-th window, with a step size of S;

[0053] The theoretical physiological data obtained from the baseline segment will undergo data preprocessing, specifically as follows:

[0054] Eye-tracking data were resampled to a uniform frequency fe and bandpass filtered using a fourth-order Butterworth filter. Missing values ​​in the pupil diameter sequence were interpolated and smoothed using a Savitzky-Golay filter. Finally, the velocity thresholding method (I-VT) was used to determine saccades: when the eye movement angular velocity |θ| > θ thr If the time is right, it is judged as scanning; otherwise, it is judged as staring.

[0055] Heart rate signal preprocessing: The ECG signal is bandpass filtered from 0.5 to 40 Hz, followed by peak detection to extract the R-peak set {t}. i Its RR interval is defined as: RR i =t i+1 -t i ;

[0056] Preprocessing of the electrodermal signal: The acquired raw electrodermal activity signal is denoted as the raw signal EDA(t). Since the raw electrodermal signal often contains high-frequency interference components such as respiration, movement, and environmental noise, it is low-pass filtered to obtain the purified signal s. EDA (t), and the signal decomposition method was used to decompose it into slow-varying components and instantaneous components; it should be noted that: the slow-varying component reflects the baseline level of skin electrical activity, and its change trend is relatively gentle, which is used to characterize the individual's overall physiological arousal level; the instantaneous component reflects the short-term response of skin electrical activity that changes rapidly with external stimuli, and its time span is relatively short, usually manifested as a peak or sudden increase signal, which is used to characterize the immediate physiological response caused by concentration, tension or changes in cognitive load during learning or training.

[0057] Head pose preprocessing: The Euler angle data of the acquired head pose is denoted as θ(t) = [φ(t), θ(t), ψ(t)], where φ(t), θ(t), and ψ(t) represent the Euler angle values ​​around the X, Y, and Z axes, respectively. The Euler angle sequence is input into a low-pass filter to filter out noise components higher than the set cutoff frequency, resulting in a smoothed Euler angle signal, denoted as θ. LP (t)=[φ LP (t), θ LP (t), ψ LP (t)];

[0058] Furthermore, the smoothed Euler angle signal is subjected to first-order difference or numerical differentiation to calculate the head angular velocity vector ω(t):

[0059] Where ω(t) represents the rotational speed of the head around the three-dimensional coordinate axes at time t.

[0060] After data preprocessing, the entire theoretical learning process is divided into continuous sliding time windows W. k =[τ k , τ k +L); in each window W k Within this process, features are extracted from the preprocessed multimodal signals, specifically as follows:

[0061] For eye-tracking features: calculate the total duration of all gaze events ∑ i d i Divide by the window length L to get the gaze duration percentage f. fix ; Count the number of times n, the subject's gaze point falls into the AOI. AOI And divide by the total number of gaze points n total The coverage rate f of the region of interest is obtained. AOI Where AOI is the region of interest; let the coordinates of all gaze points be (x... i y i ), whose mean is Then the fixation dispersion f disp The calculation is as follows: Blinking events are detected using eye movement loss or eyelid sensors, and the number of blinks (n) within a window is counted. blink Divide by the window length to get the blink frequency f. blink Let the time series of pupil diameter within the window be p(t), then the pupil dilation amplitude f pupil for: Reflects cognitive load level;

[0062] For heart rate characteristics: the average RR interval within the window is... Obtain the corresponding heart rate HR (W) k )for: Given the corresponding heart rate HR, the heart rate deviation f HR-dev (W k ) is defined as: f HR-dev (W k ) = HR(W k )-HR baseline HR baseline Baseline heart rate; within window W k Within, the RR interval sequence is {RR i}, then the degree of variation f HRV (W k )for:

[0063] For skin conductance characteristics: in window W k Within this range, the difference between the maximum and minimum values ​​of the SCR component is used to obtain the change in skin conductance, specifically: f EDA (W k)=max(SCR(t))-min(SCR(t)), t∈W k ;

[0064] For respiratory characteristics: the dominant respiratory rate BR(W) within the window is obtained through the peak power spectrum. k ), its respiratory rate change f BR (W k ) is: f BR (W k ) = BR(W k )-BR baseline HR baseline Baseline respiratory rate;

[0065] For head pose features: given the head angular velocity vector as ω(t), the stability index is its root mean square. Define the forward gaze threshold angle θ thr Within the statistics window, the angle by which the head deviates from directly forward is less than θ. thr The time ratio is used to obtain the forward gaze ratio f front (W k ):

[0066] Through the above processing, each sliding window W k A set of numerical features can be obtained from each of the following: f(W) k )={f fix f AOI f disp f blink f pupil f HR_dev f HRV f EDA f BR f head f front These features are used as direct inputs to subsequent baseline normalization and Gaussian mapping to obtain quantized feature parameters; due to differences in individual physiological signal baseline levels, the extracted features are compared with the baseline; for the j-th feature, within window W... k The value f inside j The standardized form of (Wk) is: in, These are the median and standard deviation of the feature at the baseline, respectively. It should be noted that all features are standardized to a dimensionless relative deviation scale to ensure comparability between different individuals. The extracted theoretical physiological data features are standardized using Z-scores based on the baseline and mapped to the [0, 1] interval using a Gaussian target band function. Where c is the target center and b is the bandwidth parameter; finally, each sliding window W k The level values ​​of multiple features can be obtained: aj (W k )∈[0,1];

[0067] Based on the obtained time window W k The attention level values ​​for each physiological characteristic within the body are denoted as: A(W k )={a1(W k ), a2(W k ), ..., a m (W k )}, where m is the total number of features; in each window W k Within this framework, the multidimensional feature attention values ​​are treated as a set, and their mean μ is calculated. A (W k ) and variance σ A 2 (W k ); The attention level average sequence of all windows {μ A (W1), μ A (W2), ..., μ A (W K Corresponding to the time progression of the theoretical learning task, the overall learning completion rate C is obtained through time integration: To further characterize the dynamic features of the learning process, the mean attention value μ of each window is... A (W k Constructing the probability distribution includes:

[0068] μ A (W k Divide the data into several intervals and count the number of windows (n) within each interval. i1 And calculate the probability P that occurs. i1 P i1 =n i1 / K, based on the distribution of attention level values ​​across different intervals: P 分布 ={(I1,P1),(I2,P2),…,(I r P r )};wherein, I i This represents the i1th attention interval.

[0069] Based on the obtained theoretical learning completion rate and distribution, the safety training path for each learner is analyzed and output, as follows:

[0070] Learning levels are categorized based on the completion score C interval: if completion score C < 0.4 and completion score C ≥ 0, it indicates insufficient overall attention and is classified as low level; if completion score C < 0.7 and completion score C ≥ 0.4, it indicates moderate theoretical learning effectiveness and is classified as medium level; if completion score C ≤ 1.0 and completion score C ≥ 0.7, it indicates sufficient theoretical learning and is classified as high level; then, based on the completion score distribution P... 分布 Further assessment of learning stability: If the distribution is concentrated in the high attention interval [0.6, 1.0], the learning process is considered stable; if the distribution spans multiple intervals and accounts for a large proportion in the low attention interval, it indicates that the learning state fluctuates significantly; if the distribution is concentrated in the medium interval, it indicates that passive attention exists in the learning process and the theoretical mastery is limited.

[0071] In obtaining the completion level C and the completion distribution P 分布 Next, the individual states of learners are analyzed, and a learning state vector is defined as: S = (C, P) 分布 Then, based on the learned state vector S, a personalized security training path is output:

[0072] If the completion degree C < 0.4 and the distribution P 分布 If the proportion of low to medium attention exceeds 50%, the learner is considered to have a serious lack of theoretical mastery; the output of the safety training path is a low-level training path, including: theoretical review, basic training, and standard training, and the intensity of theoretical learning is increased.

[0073] If the completion degree C < 0.7, C ≥ 0.4, and the distribution P 分布 If the learner's attention span is moderate, then they have not mastered some of the knowledge points. The safety training path outputs a moderate level of training path, including: consolidation of key theories, scenario-based training, and emergency response training. Targeted review is arranged before the training to make up for the weak cognitive links.

[0074] If the completion degree C ≥ 0.7 and the distribution P 分布 If the learner remains stable in the high attention range, it indicates that they have a high level of theoretical preparation. The safe training path outputs a high-level training path, including: standard training, high-intensity training, and comprehensive assessment. Under the premise of ensuring safety, the learner enters the advanced training stage.

[0075] The teaching and training analysis module receives each safety training path and performs training simulations, specifically:

[0076] Before the simulation training process, a safe training path is displayed via LEDs, and the physiological monitoring module obtains training physiological data. Similarly, the physiological monitoring module collects training physiological data in real time, performs preprocessing and feature extraction, and standardizes and calculates completion based on a baseline to obtain the training physiological reception value. The elevator safety teaching equipment then determines the scenario level of the simulated training based on the safe training path and the training physiological reception value, specifically:

[0077] The elevator safety training equipment classifies the current simulated scenario into initial risk levels based on the safety training path and the physiological reception values ​​during training:

[0078] If the training physiological reception value is <0.4, it is judged as a low-level scenario; if 0.4≤training physiological reception value<0.7, it is judged as a medium-level scenario; if the training physiological reception value≥0.7, it is judged as a high-level scenario; and the initial scenario level is recorded as the initial scenario level.

[0079] The low-level scenario represents a normal state scenario where the elevator is operating normally. The teaching content focuses on basic operation demonstrations and routine safety tips, such as elevator leveling and stopping, correct entry and exit, and button operation drills. Learners mainly focus on routine observation and mastering basic emergency knowledge. The medium-level scenario represents a medium-risk scenario where common abnormal situations are introduced during elevator operation, such as sudden elevator stop, slight shaking, and delayed door opening and closing. Learners need to conduct emergency operation drills under certain stress conditions, such as correctly using alarm devices, remaining calm, and cooperating with rescue procedures. The high-level scenario represents a high-risk scenario, simulating serious elevator malfunctions or emergencies, such as elevator entrapment, stopping between floors, and emergency stop. Learners need to complete practical training tasks under high-pressure conditions, including self-rescue methods, calling for rescue, comforting passengers, and maintaining order.

[0080] Based on the initial scenario level obtained from the simulation training, a training simulation is conducted. During the simulation, image data of each learner is acquired through a camera. The image data of each learner in the simulation training is analyzed to determine the training completion rate of each learner; specifically:

[0081] The image is processed using an image recognition algorithm to form a window W. k Feature set within: D img (Wk)={F(W k ), B(W k ), R(W k )};wherein, F(W k ) represents facial expression parameters, B(W) k R(W) represents the posture operation parameters. k ) are parameters for identifying risky behaviors;

[0082] The facial expression parameter F(W) k Tension index I is extracted using a facial recognition model. tense Fatigue Index I fatigue Focus Index I focus A comprehensive calculation is performed using the formula: The tension index is obtained by using existing facial expression recognition technology to detect the muscle contraction characteristics of the learner's face and combining them with an emotion classification model to obtain a numerical representation of the tension level shown during the training process. The fatigue index is obtained by analyzing the learner's eyelid closure frequency, blink duration, and facial relaxation characteristics using existing fatigue detection technology. The focus index is obtained by combining existing fixation tracking and expression recognition methods to comprehensively judge the individual's attention concentration on the task by considering the distribution of human eye fixation points, facial orientation, and emotional stability.

[0083] The fatigue index is extracted through human key point identification with an accuracy rate of I. acc Action completion rate I comp Attitude stability I stab A comprehensive calculation is performed using the formula: Among them, the operation accuracy rate comes from existing human posture recognition and action matching technologies. By comparing the matching degree between the learner's action sequence in the training operation and the standard operation sequence, the proportion of correct operation execution is obtained and recorded as the operation accuracy rate. The action completion rate is obtained by analyzing whether the learner has completely executed the task action through existing action recognition and behavior segmentation technologies. The completeness of the action execution is obtained and recorded as the action completion rate. The posture stability is obtained by calculating the smoothness and stability of the spatial position changes of key points of the human body through existing posture balance detection technologies. The stability of the learner's body during the operation is obtained and recorded as the posture stability rate.

[0084] The risk behavior identification parameters are processed using existing deep convolutional neural networks (CNNs) and temporal modeling methods to extract the trajectory features of key human points and the temporal features of actions of the learner. The extracted features are compared with a pre-established standard operating behavior database. If the matching result deviates from the standard operating sequence, it is determined that a risky behavior exists. The classifier outputs the learner's position within the current time window W. k The probability of an unsafe action occurring is denoted as P. risk (W k Based on the probability P of the unsafe action. risk (W k The risk behavior identification parameter R(W) was obtained through analysis. k The formula is: R(W) k )=1-P risk (W k ),R(W k )∈[0,1].

[0085] Then, based on the obtained facial expression parameters, posture and operation parameters, and risk behavior recognition parameters, the image completion score C is calculated.img (W k ): Through the formula: Similarly, physiological data acquired during the simulation training process are denoted as process physiological data, and analyzed to obtain the process physiological data state value, denoted as C. phys (W k The training completion rate C is obtained by combining the image completion rate and the physiological data status value. fusion (W k ):

[0086] Finally, based on the completion rate C of the practical training... fusion (W k The scenario level is dynamically adjusted; if the training completion rate is C... fusion (W k If the completion rate is ≥0.7, then upgrade to level 1 and execute the new security training path; if the training completion rate is C... fusion (W k If the value is less than or equal to 0.3, the safety training will be downgraded by one level and repeated.

[0087] The process of upgrading to the next level and executing a new safety training path involves switching from a low-level scenario to a medium-level scenario, or from a medium-level scenario to a high-level scenario; and the teaching and training analysis module reloads the corresponding higher-level safety training path, prompting learners to enter the new training phase via an LED display screen.

[0088] The process of downgrading to a lower level and conducting repeated safety training involves switching from a high-level scenario to a medium-level scenario, or from a medium-level scenario to a low-level scenario, prompting learners to conduct repeated training, and automatically arranging basic review content.

[0089] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An elevator safety teaching and training system, comprising: Elevator teaching equipment, including a theoretical learning module, a physiological monitoring module, and a teaching and training analysis module; characterized by: Elevator teaching equipment is used to simulate elevator operation and emergency scenarios; The theoretical learning module is used for theoretical learning related to elevator safety; The physiological monitoring module acquires physiological data of learners inside the elevator car and during safety training; it analyzes the physiological data to obtain the theoretical learning completion rate of each learner; based on the obtained theoretical learning completion rate and distribution, it analyzes and outputs the safety training path for each learner. The teaching and training analysis module receives each safe training path for training simulation: before the training simulation, the safe training path is displayed via LEDs, and the physiological monitoring module acquires training physiological data, analyzes it to obtain training physiological reception values, and obtains the initial scenario level of the simulated training based on the safe training path and training physiological reception values; based on the obtained initial scenario level of the simulated training, the training simulation is carried out, and image data of each learner is acquired during the training simulation. After analyzing the image data of each learner in the simulated training, the training completion rate of each learner is obtained; then the scenario level is dynamically adjusted according to the training completion rate.

2. The elevator safety teaching and training system according to claim 1, characterized in that, Elevator teaching equipment includes: The car body is equipped with floor buttons, alarm devices, and handrail pressure sensors. The visual unit projects real-time images of high-rise buildings onto a small-pitch LED screen, synchronized with changes in floor level. The motion platform is used to simulate elevator falls and roof collapses. The sound effects unit includes a hidden array of speakers mounted on the top of the car body, used for directional playback of wire breaking / metal friction sounds.

3. The elevator safety teaching and training system according to claim 1, characterized in that, The process of analyzing physiological data to determine the theoretical learning completion rate of each learner is as follows: Resting data from 60 seconds before playing the theoretical learning content was used as the baseline segment of the data parameters. The theoretical physiological data obtained from the baseline segment were preprocessed. After the data preprocessing was completed, the entire theoretical learning process was divided into continuous sliding time windows. Within each window, features are extracted from the preprocessed multimodal signal, including eye-tracking features, heart rate features, skin conductance features, and respiratory features, and denoted as a set of numerical features. These features are used as direct inputs for subsequent baseline normalization and Gaussian mapping to obtain quantized feature parameters. The extracted theoretical physiological data features are standardized by Z-scores based on the baseline and mapped to the [0, 1] interval using a Gaussian target band function. Attention level values ​​are statistically obtained based on each physiological feature within each time window. The multidimensional feature attention values ​​are treated as a set, and the sequence of attention level averages across all windows is correlated with the time progression of the theoretical learning task. The overall learning completion rate is obtained through time integration. A probability distribution is constructed from the attention mean of each window. The attention mean is divided into several intervals, and the occurrence probability is calculated by counting the number of windows in each interval. The learning completion rate distribution is obtained based on the distribution of attention level values ​​across different intervals.

4. The elevator safety teaching and training system according to claim 3, characterized in that, The specific process of extracting features from the preprocessed multimodal signals is as follows: Eye-tracking features: The total duration of all gaze events is calculated and divided by the window length L to obtain the gaze duration percentage; the number of times the subject's gaze point falls into the AOI is counted and divided by the total number of gaze points to obtain the region of interest coverage; the coordinates of all gaze points are set, and their mean is calculated to obtain the fixation dispersion; blink events are detected by eyelid sensors, the number of blinks within the window is counted, and divided by the window length to obtain the blink frequency; the pupil diameter time series within the window is set to obtain the pupil dilation amplitude; Heart rate features: The heart rate is obtained by averaging the RR intervals within a window, and the heart rate deviation is obtained from the corresponding heart rate; within the window, the RR interval sequence is used to obtain the variability. Skin conductance characteristics are obtained by taking the difference between the maximum and minimum values ​​of the SCR component to obtain the amplitude of skin conductance change; Respiratory characteristics: The dominant respiratory rate within the window is obtained by measuring the peak value of the power spectrum to obtain the change in respiratory rate; Head posture features: The stability index is the root mean square of the head angular velocity vector, obtained by acquiring the head angular velocity vector; the forward gaze threshold angle is defined, and the forward gaze ratio is obtained by statistically analyzing the proportion of time during which the head direction deviates from the front direction by less than the forward gaze threshold angle.

5. The elevator safety teaching and training system according to claim 1, characterized in that, The specific process of analyzing and outputting the safety training path for each learner based on the obtained theoretical learning completion rate and distribution is as follows: Learning levels are categorized based on completion score ranges: a completion score less than 0.4 but greater than or equal to 0 indicates a low level; a completion score less than 0.7 but greater than or equal to 0.4 indicates a medium level; and a completion score less than or equal to 1.0 but greater than or equal to 0.7 indicates a high level. Then, based on the completion distribution P 分布 Determining learning stability: If the distribution is concentrated in the high attention interval [0.6, 1.0], the learning process is considered stable; otherwise, it is unstable. This is determined by obtaining the completion rate C and the completion rate distribution P. 分布 Then, the individual status of learners is analyzed, and personalized safety training paths are output: if the completion rate is less than 0.4 and the distribution P 分布 If the proportion of low to medium attention intervals exceeds 50%, it is considered a low-level training path; if the completion rate is less than 0.7, or the completion rate is greater than or equal to 0.4 and the distribution P... 分布 If the medium attention interval accounts for 70%, then output a medium-level training path; if the completion rate is greater than or equal to 0.7 and the distribution P 分布 If the training remains stable within the high attention range, it represents an advanced level of practical training.

6. The elevator safety teaching and training system according to claim 1, characterized in that, The specific process of obtaining the initial scenario level of the simulated training based on the safety training path and the physiological reception value of the training is as follows: If the training physiological reception value is less than 0.4, it is judged as a low-level scenario; if the training physiological reception value is less than 0.7 and the training physiological reception value is greater than or equal to 0.4, it is judged as a medium-level scenario; if the training physiological reception value is greater than or equal to 0.7, it is judged as a high-level scenario; and the initial scenario level is recorded as the initial scenario level.

7. The elevator safety teaching and training system according to claim 1, characterized in that, The specific process for analyzing the image data of each learner in the simulated training to obtain the training completion rate of each learner is as follows: The image is processed by an image recognition algorithm to form a feature set within the window; then the image completion is calculated based on the obtained facial expression parameters, posture operation parameters, and risk behavior recognition parameters.

8. The elevator safety teaching and training system according to claim 7, characterized in that, The calculation process for facial expression parameters, posture and operation parameters, and risk behavior recognition parameters is as follows: Facial expression parameters are obtained by comprehensively calculating tension index, fatigue index, and concentration index extracted by a facial recognition model. The fatigue index is calculated by comprehensively considering the accuracy of human body key point recognition and extraction operations, the degree of action completion, and the stability of posture. The risk behavior identification parameters extract the trajectory features of human body key points and the temporal features of actions of learners through deep convolutional neural networks and temporal modeling methods; the extracted features are compared with a pre-established standard operation behavior library, and if the matching result deviates from the standard operation sequence, it is determined that there is a risk behavior; The probability of a learner engaging in an unsafe action within the current time window is output; risk behavior identification parameters are obtained by analyzing the obtained unsafe action probabilities.

9. The elevator safety teaching and training system according to claim 8, characterized in that, The specific process of dynamically adjusting the scenario level based on the completion rate of the practical training is as follows: Then, acquire physiological data during the training simulation process, record it as process physiological data, and analyze it to obtain process physiological data status values; combine the image completion degree and physiological data status values ​​to obtain the training completion degree; dynamically adjust the scenario level according to the training completion degree. If the training completion degree is greater than or equal to 0.7, upgrade by one level and execute a new safety training path; if the training completion degree is less than or equal to 0.3, downgrade by one level and repeat the safety training.

Images