A firefighter training effectiveness evaluation method and system

Abstract

The present invention discloses a method for evaluating the training effectiveness of firefighters. First, the corresponding human factor parameters of the firefighters to be evaluated are collected through wireless sensors; the human factor parameters include electrocardiogram signals, electromyography signals, pulse signals and respiratory signals; the collected human factor parameters are denoised using flexible analytical wavelet transform (FAWT) and salp sea squirt algorithm (SSA); the time domain of the denoised signal is analyzed to extract the corresponding eigenvalues; the extracted eigenvalues ​​include six eigenvalues: rectified mean, variance, waveform factor, peak factor, skewness and kurtosis; the training effectiveness evaluation value of the firefighters to be evaluated is calculated using a BP neural network based on the eigenvalues ​​extracted in step 3, and training suggestions are given based on the evaluation value. The method and system combine human factor engineering with signal denoising, data dimensionality reduction and other technologies to solve the problem that the training effectiveness of firefighters is difficult to evaluate.

Description

Technical Field

[0001] The present invention relates to the technical field of training effectiveness evaluation, and in particular to a firefighter training effectiveness evaluation method and system. Background Art

[0002] When major disasters and accidents occur, the emergency rescue capabilities of firefighters are particularly important. Firefighters' on-the-spot emergency rescue capabilities are accumulated through daily technical and tactical training. However, in daily training, there are often problems such as too high or too low training intensity and too much emphasis on quantity. These will make the firefighters' training fail to achieve the expected results. In severe cases, it may also lead to excessive fatigue and even muscle or joint injuries. How to correctly and scientifically evaluate the effectiveness of firefighter training is one of the problems that need to be solved urgently.

[0003] The existing performance evaluation methods include expert scoring method, analytic hierarchy process, analytical calculation method, statistical test method, etc. However, the expert scoring method and analytic hierarchy process have the disadvantage that the subjective factors account for too large a proportion, the analytical calculation method considers fewer factors, and the statistical test method requires a large amount of actual combat and exercise data to be obtained in advance. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for evaluating the training effectiveness of firefighters. The method and system combine ergonomics with signal denoising, data dimensionality reduction and other technologies to solve the problem that it is difficult to evaluate the training effectiveness of firefighters.

[0005] The objective of the present invention is achieved through the following technical solutions:

[0006] A method for evaluating the effectiveness of firefighter training, the method comprising:

[0007] Step 1: Collect the corresponding human factor parameters of the firefighters to be evaluated through wireless sensors; the human factor parameters include electrocardiogram signals, electromyography signals, pulse signals and respiratory signals;

[0008] Step 2, using flexible analytical wavelet transform (FAWT) and salp sea squamous algorithm (SSA) to denoise the collected human factor parameters;

[0009] Step 3, analyze the time domain of the signal after denoising and extract the corresponding eigenvalues; the extracted eigenvalues ​​include six eigenvalues: rectified mean, variance, waveform factor, peak factor, skewness, and kurtosis;

[0010] Step 4: Use the BP neural network to calculate the training effectiveness evaluation value of the firefighter to be evaluated based on the characteristic values ​​extracted in step 3, and give training suggestions based on the evaluation value.

[0011] A firefighter training effectiveness evaluation system, the system comprising:

[0012] A human factor parameter acquisition module, used to acquire corresponding human factor parameters of the firefighters to be evaluated through wireless sensors; the human factor parameters include electrocardiogram signals, electromyography signals, pulse signals and respiratory signals;

[0013] The human factor parameter denoising module is used to denoise the collected human factor parameters using the flexible analytical wavelet transform FAWT and the salp sea squirt algorithm SSA;

[0014] The eigenvalue extraction module is used to analyze the time domain of the signal after denoising and extract the corresponding eigenvalues; the extracted eigenvalues ​​include six eigenvalues: rectified mean, variance, waveform factor, peak factor, skewness, and kurtosis;

[0015] The effectiveness evaluation module is used to calculate the training effectiveness evaluation value of the firefighter to be evaluated based on the eigenvalues ​​extracted by the eigenvalue extraction module using the BP neural network, and give training suggestions based on the evaluation value.

[0016] It can be seen from the technical solution provided by the present invention that the above method and system combine ergonomics with signal denoising, data dimensionality reduction and other technologies, which solves the problem of difficulty in evaluating the effectiveness of firefighter training, improves the accuracy of evaluation, and provides a visual evaluation system for the direction and effect of firefighter training, which is conducive to improving the enthusiasm of firefighter training, strengthening the construction of firefighting talent team, and improving the rescue capabilities of firefighters. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0018] Figure 1 A schematic diagram of a flow chart of a method for evaluating the effectiveness of firefighter training provided by an embodiment of the present invention;

[0019] Figure 2 It is a structural schematic diagram of the firefighter training effectiveness evaluation system according to an embodiment of the present invention;

[0020] Figure 3 A schematic diagram of a system interface of an example according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments, which does not constitute a limitation of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] like Figure 1 The figure is a flow chart of a method for evaluating the effectiveness of firefighter training provided by an embodiment of the present invention, the method comprising:

[0023] Step 1: Collect the corresponding human factor parameters of the firefighters to be evaluated through wireless sensors;

[0024] In this step, the human factors parameters collected include ECG signals, EMG signals, pulse signals and respiratory signals. ECG signals are closely related to the generation and disappearance of human fear reactions; the strength of EMG signals quantifies human muscle strength, explosive power, etc.; the waveform characteristic changes of photoelectric volume pulse signals are an important basis for evaluating the physiological and pathological state of the human cardiovascular system; changes in respiratory signals can reflect changes in human alertness from the side.

[0025] Specifically, the electrocardiogram (ECG) is a very weak physiological low-frequency electrical signal. According to the standards established by the American Electrocardiogram Association, the amplitude range of a typical ECG signal is 10μV to 4mV, the standard value is 1mV, and the spectrum energy is mostly concentrated in 0.25Hz to 35Hz. ECG signals can reflect the electrical activity process of cardiac excitement. They can not only be used to analyze and identify various arrhythmias, but also reflect the degree and development of myocardial damage, as well as the characterization of fear reactions. It is a physiological indicator that can well reflect the effectiveness of firefighter training.

[0026] Surface electromyography (sEMG) is a non-stationary micro-electric AC signal generated by muscle contraction. It is generally proportional to muscle strength, with an amplitude of 0 to 1.5 mV, a useful signal frequency of 0 to 500 Hz, and the main energy concentrated in 20 Hz to 150 Hz. sEMG can distinguish between neurogenic and myogenic injuries by measuring the duration and amplitude of motor unit potentials, whether there is spontaneous electrical activity in a quiet state, and the waveform and amplitude of muscle contraction. In addition, electromyography can also be used to track the force-generating parts of the human body and the physical recovery process, which is of reference value for evaluating the effectiveness of firefighter training.

[0027] Photoplethysmography (PPG) is a method of detecting the heart rate of human exercise using photoplethysmography. The principle is to detect the difference in the intensity of reflected light absorbed by human blood and tissues through a photoelectric sensor, record the changes in blood pressure during the cardiac cycle, and calculate the heart rate based on the pulse waveform. The change in pulse frequency is an external reflection of important information such as the state of the heart and blood vessels, and is also an important physiological indicator reflecting the health of the human body.

[0028] Respiratory signal (RESP) is a human electrical signal generated by the periodic deformation of the airway, chest and abdomen after the conversion of exhalation and inhalation. As the degree of fatigue deepens, the human breathing amplitude decreases and the cycle frequency is delayed. In addition, as the individual physiological state changes, the human respiratory signal will also change accordingly. If the human body stops breathing, it will cause hypoxia in the short term. If the stop time is too long, it will cause severe acidosis and even suffocation. At the same time, the human body's reaction ability and alertness will also decrease accordingly. Therefore, the respiratory signal is also an important physiological indicator for evaluating the effectiveness of firefighter training.

[0029] Step 2, using flexible analytical wavelet transform (FAWT) and salp sea squamous algorithm (SSA) to denoise the collected human factor parameters;

[0030] In this step, relevant constraints are added on the basis of the flexible analytical wavelet transform FAWT to improve it into a denoising model with the collected human factor parameters as input and the signal-to-noise ratio of the decomposed signal as output;

[0031] Then, the Salp Sea Squid algorithm (SSA) in the swarm intelligence optimization algorithm is added to adjust the quality factor and redundancy in the flexible analytic wavelet transform (FAWT), and the optimized model is used to effectively denoise the low-frequency signals in the collected human factor parameters.

[0032] In a specific implementation, the frequency responses of the low-pass filter H(w) and the high-pass filter G(w) of the flexible analytical wavelet transform FAWT are shown in Formula 1 and Formula 2 respectively:

[0033]

[0034]

[0035] In formula 1, p and q control the up-sampling rate and down-sampling rate of the low-pass filter H(w) respectively; w s and w P represents the frequency band parameter of the low-pass filter; while r and s in Formula 2 determine the up-sampling rate and down-sampling rate of the high-pass filter G(w), w 1 、w 2 、w 3 and w 4 As the frequency band parameter of the high-pass filter;

[0036] By changing the up- and down-sampling parameters p, q, r and s of the filter and the constant β, the quality factor Q and the redundancy R of the flexible analytical wavelet transform FAWT can be adjusted, thereby achieving accurate analysis of different oscillation component signals and precise positioning of mutation components;

[0037] When the constant β is fixed and the up and down sampling parameters p, q, r, s are changed, the center frequency and filter gain of each filter will change accordingly. The relevant constraints are added as shown in Equation 3:

[0038]

[0039] Therefore, the flexible analytical wavelet transform FAWT uses 3 filters to process the signal. When the decomposition layer number of the flexible analytical wavelet transform FAWT is 1, the signal will be decomposed into 3 frequency bands; when the decomposition layer number is m, the signal sequence will be decomposed into 2m+1 sub-frequency bands, and the fineness is equivalent to 2m+1 times of the original signal. As the value of the decomposition layer m increases, the fineness of the signal will increase linearly.

[0040] The parameters of the flexible analytical wavelet transform FAWT can be flexibly adjusted. The quality factor Q is increased when processing an oscillating signal, and the quality factor Q is decreased when processing a stable signal with less noise. Flexible parameter settings can enable FAWT to better process different types of signals.

[0041] In addition, in the introduced salp algorithm SSA, the salp population is divided into two parts: leaders and followers; the leader is located at the front of the chain, it leads the followers to move in the D-dimensional space and find food, and the other individuals are followers;

[0042] The leader's position update method is shown in Formula 4:

[0043]

[0044] is the position of the leader in the j-dimensional space; F j is the position of the food in the j-dimensional space; max j is the upper boundary of the value of the j-dimensional space; min j is the lower boundary of the value of the j-th dimension space; c 1 is the convergence factor, which is used to control the exploration and development capabilities of the entire population. It is related to the number of iterations of the current population. The calculation method is shown in Formula 5:

[0045]

[0046] Among them, t is the current number of generations; T max is the maximum number of generations; c 2 and c3 is a random number generated in the interval [0,1], c 2 What determines the length of the move, c 3 What is determined is the positive or negative direction of movement;

[0047] The followers move with the leader. The position of the i-th follower in the next iteration is determined by its own position in the current generation and the position of the i-1th salp. Therefore, the position update formula of the follower is shown in Formula 6:

[0048]

[0049] in, represents the coordinates of the i-th salp in the j-dimensional space at the t-th generation;

[0050] In the specific implementation, since the selection of up and down sampling factors in the transformation process of FAWT has a great influence on the signal processing effect, this embodiment first converts FAWT into a multi-input single-output model, and uses the introduced Salp Sea Squid algorithm SSA to select the optimal parameters. Specifically:

[0051] After introducing the signal of the human factor parameter, the relevant parameters are set, the upper and lower sampling parameters of the filter in the flexible analytical wavelet transform FAWT are initialized, and the upper and lower sampling parameters of the filter are set to the primary chromosome with an initial value of 1, and the initial signal-to-noise ratio, the leader position and the follower position are calculated by the initial sampling parameters, and the food source position is updated to determine whether the individual position of the salp exceeds the upper and lower limits of the search space;

[0052] If the upper and lower limits are exceeded, the positions of the salps and the food source are recalculated according to the position update formula of the leader and the follower, and new sampling parameters are calculated accordingly, and the new sampling parameters are brought into the flexible analytical wavelet transform FAWT to calculate a new signal-to-noise ratio;

[0053] If the upper and lower limits are not exceeded, the new generation of sampling parameters are brought into the flexible analytical wavelet transform FAWT, which performs denoising processing and outputs the denoised human factor parameter signal.

[0054] Step 3, analyze the time domain of the signal after denoising and extract the corresponding eigenvalues; the extracted eigenvalues ​​include six eigenvalues: rectified mean, variance, waveform factor, peak factor, skewness, and kurtosis;

[0055] In this step, (1) the rectified average value is expressed as the average value of the integral of the absolute value of the signal, which is used to describe the overall mean characteristics of the AC signal, where T is the sampling period, x(t) is the input signal, and t 0 is the initial time, as shown in formula (7):

[0056]

[0057] (2) Variance describes the degree of deviation of a random variable from its mathematical expectation, as shown in formula (8), where n is the sampling rate, x i is the input signal, is the signal mean, and the variance is used to measure the volatility of sample data. The smaller the variance, the more stable the sample:

[0058]

[0059] (3) The form factor is the ratio of the effective value (RMS) to the rectified average value, where x rms is the effective value of the signal, x arv is the rectified average value of the signal, see formula (9):

[0060]

[0061] (4) The peak factor is the ratio of the signal peak value to the effective value, which represents the extreme degree of the peak value in the waveform, as shown in formula (10):

[0062]

[0063] where x peak is the signal peak value; x rms is the effective value, n is the sampling rate, x i is the input signal, see formula (11):

[0064]

[0065] (5) Skewness, also known as skewness, is the ratio of the third central moment to the cube of the standard deviation, as shown in formula (12). Skewness reflects the symmetry of the distribution:

[0066]

[0067] Where X is the input signal, μ is the mean, σ is the standard deviation, and E is the mean operation.

[0068] If we simply understand the physical meaning, for a unimodal distribution, negative skewness means that the "head" of the distribution line is on the right and the "tail" is on the left. Skewness can reflect the symmetry of the distribution, and kurtosis reflects the sharpness of the image: the greater the kurtosis, the sharper the center point on the image.

[0069] (6) Kurtosis is a statistic that reflects the distribution characteristics of vibration signals. Skewness and kurtosis are related to a certain extent. Kurtosis is the ratio of the fourth-order central moment to the fourth power of the standard deviation, as shown in formula (13), where n is the sampling rate, x is i is the input signal, is the signal mean:

[0070]

[0071] Step 4: Use the BP neural network to calculate the training effectiveness evaluation value of the firefighter to be evaluated based on the characteristic values ​​extracted in step 3, and give training suggestions based on the evaluation value.

[0072] In this step, using BP neural network to calculate the training effectiveness evaluation value mainly includes two steps: input incentive and weight correction. The specific process is:

[0073] Firstly, the extracted human factor parameter characteristic value sample set is input into the BP neural network model to obtain the corresponding response result, and the response result is compared with the training target value to obtain the BP neural network error;

[0074] According to the obtained BP neural network error, the error factors with different weights are obtained by multiplying the input and the error;

[0075] According to the steepest descent method, the local gradient of the error to the initial weight is calculated, that is, the product of the error signal and the derivative of the transfer function, and the weight correction is obtained;

[0076] The hidden layer before the output layer is corrected according to the weight correction amount, and then the weight of the previous layer is corrected according to the error of the hidden layer. By correcting the weights of all hidden layers and input layers, the training result is closest to the training target in the human factor parameter characteristic value sample set, and the final training effectiveness evaluation value is calculated.

[0077] Based on the above method, the embodiment of the present invention also provides a firefighter training effectiveness evaluation system, such as Figure 2 FIG. 1 is a schematic diagram of the structure of a system according to an embodiment of the present invention, wherein the system comprises:

[0078] A human factor parameter acquisition module, used to acquire corresponding human factor parameters of the firefighters to be evaluated through wireless sensors; the human factor parameters include electrocardiogram signals, electromyography signals, pulse signals and respiratory signals;

[0079] The human factor parameter denoising module is used to denoise the collected human factor parameters using the flexible analytical wavelet transform FAWT and the salp sea squirt algorithm SSA;

[0080] The eigenvalue extraction module is used to analyze the time domain of the signal after denoising and extract the corresponding eigenvalues; the extracted eigenvalues ​​include six eigenvalues: rectified mean, variance, waveform factor, peak factor, skewness, and kurtosis;

[0081] The effectiveness evaluation module is used to calculate the training effectiveness evaluation value of the firefighter to be evaluated based on the eigenvalues ​​extracted by the eigenvalue extraction module using the BP neural network, and give training suggestions based on the evaluation value.

[0082] The specific implementation of each of the above modules is described in the above method embodiment.

[0083] For example, the embodiment of the present invention can design a firefighter training effectiveness evaluation system based on human factor parameters through the App Designer platform of MATLAB, such as Figure 3 The system interface diagram of the example of the embodiment of the present invention is shown. The interface of the system is divided into three columns. The first column is the button operation column. The top is the import signal button, and the bottom is the signal preprocessing operation box, including classical modal decomposition and wavelet transform. After signal denoising, it is also necessary to extract the eigenvalue, calculate the training effectiveness evaluation value based on the eigenvalue, and evaluate it through the BP neural network.

[0084] The second column is the view column. After the ECG signal, EMG signal, pulse signal, and respiratory signal data are imported, they will be presented in this view column. Click the "Import Signal" button to present the original signal in this view frame; click the "Classic Modal Decomposition" button to refresh the view frame and present the signal after EMD denoising; click the "Wavelet Transform" button to refresh the view frame again and present the signal after wavelet transform denoising. Finally, click the "Extract Eigenvalues" button, and the extracted eigenvalue data table will be displayed below the view column;

[0085] The third column is the result column, which mainly displays the training effectiveness evaluation results and related suggestion text boxes.

[0086] After extracting the characteristic values, the corresponding training effectiveness score will be calculated according to different analysis methods. That is, the firefighter's alertness, fatigue, muscle strength, fear reaction, etc. will be evaluated based on the imported electrocardiogram (ECG) signal, electromyography (EMG) signal, pulse signal, and respiratory signal data, and then the comprehensive evaluation value of their training effectiveness will be calculated. Then, corresponding training suggestions will be given based on the scoring system.

[0087] It is worth noting that the contents not described in detail in the embodiments of the present invention belong to the prior art known to professional and technical personnel in the field.

[0088] In order to more clearly demonstrate the technical solutions and technical effects provided by the present invention, the method and system provided by the embodiments of the present invention are described in detail with specific embodiments. This example uses ErgoLAB wireless sensors to collect the corresponding human factors parameters of firefighters. The system can use any indicator sensor and automatically identify the type of signal, and can monitor the human body's respiratory rate (RESP), pulse (PPG), electrocardiogram (ECG), electromyography (EMG) and other physiological indicators in real time. In order to simulate the emergencies that firefighters may encounter during the rescue process, this example cooperates deeply with the Xihongmen Special Service Station of the Daxing District Fire Rescue Detachment to measure human factors parameters in three training scenarios: high-altitude rappelling, 10-meter rock climbing, and weighted climbing. According to the measured human factors parameters, the training effectiveness of firefighters is evaluated.

[0089] In order to analyze the denoising effect of the modified algorithm, this example takes the sEMG of firefighters as an example for analysis. When collecting sEMG, the positive and negative electrodes of the sensor are placed on the extensor carpi longus of the brachialis muscle to measure the brachialis muscle, and the reference electrode is placed on the bony landmark without muscle. The direction of electrode placement is along the direction of muscle fibers. Then the sEMG of the test subjects is collected when they make a series of brachialis muscle-related actions such as clenching fists, extending palms, and weightlifting. Since weightlifting is the most commonly used action by emergency rescue personnel during emergency rescue, and it can also effectively reflect the muscle strength and professional ability of emergency rescue personnel, this example collects sEMG of 8 firefighters during weightlifting as experimental signals.

[0090] Taking one of the firefighters as an example, this example first intercepts the valid signal fragment as input, sets the decomposition layer number of FAWT to 6 layers, and the initial up and down sampling parameters are p = 4, q = 6, r = 9, s = 10, k = 0.8. After calculation, the initial signal-to-noise ratio of sEMG is 10.611629, and the signal-to-noise ratio after FAWT denoising is 11.605031. As a rational expansion wavelet transform, FAWT can adjust the quality factor Q and redundancy R by changing the up and down sampling parameters p, q, r, s and constant β of the filter group, so as to achieve accurate analysis of signals of different oscillation components and precise positioning of mutation components. Compared with the WT algorithm, the signal-to-noise ratio of the FAWT algorithm after denoising is not significantly improved. Therefore, it is necessary to add an optimization algorithm to reasonably select sampling parameters so that the algorithm can not only effectively remove noise, but also retain the effective characteristics of the signal to the maximum extent.

[0091] This embodiment adds the Salp Sea SSA algorithm to the existing FAWT to optimize its up and down sampling parameters, and uses the comprehensive evaluation index as the objective function for denoising, thereby obtaining a better denoising effect. The initial signal-to-noise ratio is 10.611629, and the signal-to-noise ratio after the improved FAWT transformation is 12.456929. From the results, it can be seen that the signal-to-noise ratio of the signal after denoising using the improved FAWT algorithm is greatly improved, and it has a better denoising effect. In summary, when the improved FAWT algorithm processes different oscillation signals, by adopting the Salp Sea SSA algorithm, it can be optimized from the signal-to-noise ratio according to the characteristics of the signal, thereby improving the overall denoising effect of the algorithm.

[0092] Table 1 Characteristic values ​​of human factors parameters of firefighters in high-altitude rappelling projects

[0093]

[0094] After extracting the eigenvalues ​​of human factor parameters, the eigenvalues ​​can also be used to conduct a more detailed analysis of the training effectiveness of firefighters. For example, in the high-altitude rappelling project, the peak factor of the sEMG signal of Firefighter No. 2 is the highest, which shows that No. 2's muscle explosiveness is relatively strong; Firefighter No. 3's signal rectified average value is the smallest, which shows that No. 3's overall muscle strength is weak, so it takes longer; Firefighter No. 4's signal rectified average value is the largest, and the variance is the largest, so his overall muscle strength is relatively strong.

[0095] Table 2 Comparison of firefighter training effectiveness evaluation results

[0096]

[0097]

[0098] In this example, the human factor parameter characteristic values ​​of 8 firefighters measured multiple times in different training scenarios are used as training samples, and one set of data is used as a test sample to obtain the training effectiveness evaluation results of firefighters. In order to verify the effectiveness of the evaluation results, the study collected the training effectiveness of the 8 firefighters participating in the test according to the expert scoring method. At the same time, the system divides the training effectiveness of firefighters into four levels according to the scores of the evaluation results, among which scores above 85 points are excellent; 75-85 points are rated as good; 65-75 points are rated as medium; and below 65 points are rated as poor. After comparison and verification, it can be seen that the firefighter training effectiveness evaluation results based on human factor parameters are more accurate and effective than the traditional expert scoring evaluation results.

[0099] Traditional training effectiveness evaluation methods can only evaluate based on the time spent by firefighters on a certain project, and cannot provide scientific analysis and guidance for a specific firefighter. However, when using human factor parameters for analysis, targeted analysis and training guidance can be given to each firefighter. Taking the sEMG signals collected by firefighters in high-altitude rappelling projects as an example, after analysis, it can be found that:

[0100] Firefighter No. 1's muscle burst in the rappelling event is mainly at the beginning and end of the rappelling, Firefighter No. 2's muscle burst is mainly in the middle of the rappelling, Firefighter No. 3's force is relatively even throughout the event, and Firefighter No. 4's force is at the end. The comprehensive project time shows that Firefighter No. 1 takes the least time, followed by No. 2 and No. 4, and No. 3 takes the longest time. From the above analysis, it can be seen that in the rappelling event, if you want to improve the training efficiency of firefighters, you can first start from the perspective of force. In the rappelling process, exerting force mainly at the beginning and end can effectively improve the efficiency of rappelling.

[0101] The study also input the human factor parameters of firefighter No. 1 before denoising into the neural network, and the evaluation result was 83.6 points; the result of the human factor parameters evaluated by the neural network after denoising was 92.1 points; according to the daily evaluation results of firefighter No. 1 by the fire brigade, it can be known that his training effectiveness is about 90 points. Therefore, the results evaluated by the neural network after denoising are more in line with the actual training effectiveness of the firefighter, so it can be known that effective denoising is needed before evaluation.

[0102] In summary, the method and system described in the embodiment of the present invention combine human factor parameters with signal processing technology, design different experimental scenarios and collect various human factor parameters for analysis and processing, establish the relationship between human factor parameter changes and training effectiveness, scientifically and effectively evaluate the training effectiveness of firefighters, and provide a visual evaluation system for the direction and effect of firefighter training, which is conducive to improving the enthusiasm of firefighter training, strengthening the construction of firefighting talent team, and improving the rescue capabilities of firefighters.

[0103] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with the technical field within the technical scope disclosed in the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims. The information disclosed in the background technology section of this article is only intended to deepen the understanding of the overall background technology of the present invention, and should not be regarded as an admission or in any form that the information constitutes prior art known to those skilled in the art.

Claims

1. A firefighter training effectiveness evaluation method, It is characterized in that The method comprises: Step 1: Collect the corresponding human factor parameters of the firefighters to be evaluated through wireless sensors; the human factor parameters include electrocardiogram signals, electromyography signals, pulse signals and respiratory signals; Step 2, using flexible analytical wavelet transform FAWT and salp sea squamous algorithm SSA to denoise the collected human factor parameters; in step 2, specifically, on the basis of flexible analytical wavelet transform FAWT, relevant constraints are added to improve it into a denoising model with the collected human factor parameters as input and the signal-to-noise ratio of the decomposed signal as output; Then, the Salp Sea Squid algorithm (SSA) in the swarm intelligence optimization algorithm is added to adjust the quality factor and redundancy in the flexible analytical wavelet transform (FAWT), and the optimized model is used to effectively denoise the low-frequency signals in the collected human factor parameters. In the introduced Salp Algorithm SSA, the Salp population is divided into two parts: leaders and followers; the leader is at the front of the chain, leading the followers to move in the D-dimensional space and find food, and the other individuals are followers; The leader’s position update method is shown in formula (4): is the position of the leader in the j-dimensional space; F j is the position of the food in the j-dimensional space; max j is the upper boundary of the value of the j-dimensional space; min j is the lower boundary of the value of the j-th dimension space; c 1 is the convergence factor, which is used to control the exploration and development capabilities of the entire population. It is related to the number of iterations of the current population. The calculation method is shown in formula (5): Among them, t is the current number of generations; T max is the maximum number of generations; c 2 and c 3 is a random number generated in the interval [0,1], c 2 What determines the length of the move, c 3 What is determined is the positive or negative direction of movement; The followers move along with the leader. The position of the i-th follower in the next iteration is determined by its own position in the current generation and the position of the i-1th salp. Therefore, the position update formula of the follower is shown in formula (6): in, represents the coordinates of the i-th salp in the j-dimensional space at the t-th generation; In the process of denoising the collected human factors parameters: After introducing the signal of the human factor parameter, the relevant parameters are set, the upper and lower sampling parameters of the filter in the flexible analytical wavelet transform FAWT are initialized, and the upper and lower sampling parameters of the filter are set to the primary chromosome with an initial value of 1, and the initial signal-to-noise ratio, the leader position and the follower position are calculated by the initial sampling parameters, and the food source position is updated to determine whether the individual position of the salp exceeds the upper and lower limits of the search space; If the upper and lower limits are exceeded, the positions of the salps and the food source are recalculated according to the position update formula of the leader and the follower, and new sampling parameters are calculated accordingly, and the new sampling parameters are brought into the flexible analytical wavelet transform FAWT to calculate a new signal-to-noise ratio; If the upper and lower limits are not exceeded, the new generation of sampling parameters are brought into the flexible analytical wavelet transform FAWT, and the flexible analytical wavelet transform FAWT performs denoising processing and outputs the denoised human factor parameter signal; Step 3, analyze the time domain of the signal after denoising and extract the corresponding eigenvalues; the extracted eigenvalues ​​include six eigenvalues: rectified mean, variance, waveform factor, peak factor, skewness, and kurtosis; Step 4: Calculate the training effectiveness evaluation value of the firefighter to be evaluated based on the characteristic value extracted in step 3 by using the BP neural network, and give training suggestions based on the evaluation value; In step 4, the process of using the BP neural network to calculate the training effectiveness evaluation value of the firefighter to be evaluated based on the characteristic value extracted in step 3 is as follows: Firstly, the extracted human factor parameter characteristic value sample set is input into the BP neural network model to obtain the corresponding response result, and the response result is compared with the training target value to obtain the BP neural network error; According to the obtained BP neural network error, the error factors with different weights are obtained by multiplying the input and the error; According to the steepest descent method, the local gradient of the error to the initial weight is calculated, that is, the product of the error signal and the derivative of the transfer function, and the weight correction is obtained; The hidden layer before the output layer is corrected according to the weight correction amount, and then the weight of the previous layer is corrected according to the error of the hidden layer. By correcting the weights of all hidden layers and input layers, the training result is closest to the training target in the human factor parameter characteristic value sample set, and the final training effectiveness evaluation value is calculated.

2. The firefighter training effectiveness evaluation method according to claim 1, It is characterized in that The frequency responses of the low-pass filter H(w) and the high-pass filter G(w) of the flexible analytical wavelet transform FAWT are shown in equations (1) and (2), respectively: In formula (1), p and q respectively control the up-sampling rate and down-sampling rate of the low-pass filter H(w); w s and w P represents the frequency band parameter of the low-pass filter; while r and s in formula (2) determine the up-sampling rate and down-sampling rate of the high-pass filter G(w), w 1 、w 2 、w 3 and w 0 As the frequency band parameter of the high-pass filter; By changing the up- and down-sampling parameters p, q, r and s of the filter and the constant β, the quality factor Q and the redundancy R of the flexible analytical wavelet transform FAWT can be adjusted, thereby achieving accurate analysis of different oscillation component signals and precise positioning of mutation components; When the constant β is fixed and the up- and down-sampling parameters p, q, r, and s are changed, the center frequency and filter gain of each filter will change accordingly. The relevant constraints are added as shown in formula (3): When the decomposition level of the flexible analytical wavelet transform FAWT is 1, the signal will be decomposed into 3 frequency bands; when the decomposition level is m, the signal sequence will be decomposed into 2m+1 sub-frequency bands, and the fineness is equivalent to 2m+1 times of the original signal. As the value of the decomposition level m increases, the fineness of the signal will increase linearly; The parameters of the flexible analytical wavelet transform FAWT can be flexibly adjusted. When processing an oscillating signal, the quality factor Q is increased, and when processing a stable signal with less noise, the quality factor Q is decreased.

3. The firefighter training effectiveness evaluation method according to claim 1, It is characterized in that In step 3, (1) The rectified mean value is expressed as the average value of the integral of the absolute value of the signal, which is used to describe the overall mean characteristics of the AC signal, where T is the sampling period, x(t) is the input signal, and t 0 is the initial time, as shown in formula (7): (2) Variance describes the degree of deviation of a random variable from its mathematical expectation, as shown in formula (8), where n is the sampling rate, x i is the input signal, is the signal mean, and the variance is used to measure the volatility of sample data. The smaller the variance, the more stable the sample: (3) The form factor is the ratio of the effective value to the rectified average value, where x rms is the effective value of the signal, x arv is the rectified average value of the signal, see formula (9): (4) The peak factor is the ratio of the signal peak value to the effective value, which represents the extreme degree of the peak value in the waveform, as shown in formula (10): where x peak is the signal peak value; x rms is the effective value, n is the sampling rate, x i is the input signal, see formula (11): (5) Skewness, also known as skewness, is the ratio of the third central moment to the cube of the standard deviation, as shown in formula (12). Skewness reflects the symmetry of the distribution: Where X is the input signal, μ is the mean, σ is the standard deviation, and E is the mean operation. (6) Kurtosis is a statistic that reflects the distribution characteristics of vibration signals. Kurtosis is the ratio of the fourth-order central moment to the fourth power of the standard deviation, as shown in formula (13), where n is the sampling rate, x is i is the input signal, is the signal mean:

4. A firefighter training effectiveness evaluation system, It is characterized in that The system comprises: A human factor parameter acquisition module, used to acquire corresponding human factor parameters of the firefighters to be evaluated through wireless sensors; the human factor parameters include electrocardiogram signals, electromyography signals, pulse signals and respiratory signals; The human factor parameter denoising module is used to denoise the collected human factor parameters using the flexible analytical wavelet transform FAWT and the Salp Sea Squib algorithm SSA; specifically, on the basis of the flexible analytical wavelet transform FAWT, relevant constraints are added to improve it into a denoising model that takes the collected human factor parameters as input and the signal-to-noise ratio of the decomposed signal as output; Then, the Salp Sea Squid algorithm (SSA) in the swarm intelligence optimization algorithm is added to adjust the quality factor and redundancy in the flexible analytical wavelet transform (FAWT), and the optimized model is used to effectively denoise the low-frequency signals in the collected human factor parameters. In the introduced Salp Algorithm SSA, the Salp population is divided into two parts: leaders and followers; the leader is at the front of the chain, leading the followers to move in the D-dimensional space and find food, and the other individuals are followers; The leader’s position update method is shown in formula (4): is the position of the leader in the j-dimensional space; F j is the position of the food in the j-dimensional space; max j is the upper boundary of the value of the j-dimensional space; min j is the lower boundary of the value of the j-th dimension space; c 1 is the convergence factor, which is used to control the exploration and development capabilities of the entire population. It is related to the number of iterations of the current population. The calculation method is shown in formula (5): Among them, t is the current number of generations; T max is the maximum number of generations; c 2 and c 3 is a random number generated in the interval [0,1], c 2 What determines the length of the move, c 3 What is determined is the positive or negative direction of movement; The followers move along with the leader. The position of the i-th follower in the next iteration is determined by its own position in the current generation and the position of the i-1th salp. Therefore, the position update formula of the follower is shown in formula (6): in, represents the coordinates of the i-th salp in the j-dimensional space at the t-th generation; In the process of denoising the collected human factors parameters: After introducing the signal of the human factor parameter, the relevant parameters are set, the upper and lower sampling parameters of the filter in the flexible analytical wavelet transform FAWT are initialized, and the upper and lower sampling parameters of the filter are set to the primary chromosome with an initial value of 1, and the initial signal-to-noise ratio, the leader position and the follower position are calculated by the initial sampling parameters, and the food source position is updated to determine whether the individual position of the salp exceeds the upper and lower limits of the search space; If the upper and lower limits are exceeded, the positions of the salps and the food source are recalculated according to the position update formula of the leader and the follower, and new sampling parameters are calculated accordingly, and the new sampling parameters are brought into the flexible analytical wavelet transform FAWT to calculate a new signal-to-noise ratio; If the upper and lower limits are not exceeded, the new generation of sampling parameters are brought into the flexible analytical wavelet transform FAWT, and the flexible analytical wavelet transform FAWT performs denoising processing and outputs the denoised human factor parameter signal; The eigenvalue extraction module is used to analyze the time domain of the signal after denoising and extract the corresponding eigenvalues; the extracted eigenvalues ​​include six eigenvalues: rectified mean, variance, waveform factor, peak factor, skewness, and kurtosis; The effectiveness evaluation module is used to calculate the training effectiveness evaluation value of the firefighter to be evaluated based on the eigenvalues ​​extracted by the eigenvalue extraction module using the BP neural network, and give training suggestions based on the evaluation value; the process of calculating the training effectiveness evaluation value of the firefighter to be evaluated based on the extracted eigenvalues ​​using the BP neural network is specifically as follows: Firstly, the extracted human factor parameter characteristic value sample set is input into the BP neural network model to obtain the corresponding response result, and the response result is compared with the training target value to obtain the BP neural network error; According to the obtained BP neural network error, the error factors with different weights are obtained by multiplying the input and the error; According to the steepest descent method, the local gradient of the error to the initial weight is calculated, that is, the product of the error signal and the derivative of the transfer function, and the weight correction is obtained; The hidden layer before the output layer is corrected according to the weight correction amount, and then the weight of the previous layer is corrected according to the error of the hidden layer. By correcting the weights of all hidden layers and input layers, the training result is closest to the training target in the human factor parameter characteristic value sample set, and the final training effectiveness evaluation value is calculated.

Images