Safety behavior capability assessment method and device under time pressure, and electronic equipment

By analyzing the safety behavioral capability data and physiological signals of subway drivers, training neural network models to predict their safety behavioral capability index, solving the problem that the driver's status cannot be quantitatively evaluated in the existing technology and improving operational safety.

CN119025904AInactive Publication Date: 2024-11-26SHANGHAI UNIV OF ENG SCI
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410909049.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-08
Publication Date
2024-11-26
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The prior art is difficult to quantitatively evaluate the status of subway drivers under time pressure, resulting in possible accidents and affecting operational safety.

Method used

By obtaining driver's safety behavioral ability data and physiological signal data, analyzing the correlation and time lag between physiological indicators and safety behavioral ability index, training neural network models to predict driver's safety behavioral ability index, providing a time pressure safety behavioral capacity assessment method and device.

Benefits of technology

A quantitative assessment of the driver's status is achieved, and potential hazards can be discovered in a timely manner, reducing the probability of accidents, and improving operational safety.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119025904A_ABST
    Figure CN119025904A_ABST
Patent Text Reader

Abstract

The invention discloses a safety behavior capability assessment method and device under time pressure and electronic equipment, and belongs to the technical field of safe driving, and the method comprises the steps: obtaining safety behavior capability data and physiological signal data of a driver under different time pressures; based on the safety behavior capability data and the physiological signal data, correlation and time lag between the physiological indexes and safety behavior capability indexes are analyzed; and obtaining a safety behavior capability index prediction model according to the correlation between the physiological indexes and the safety behavior capability indexes and the time delay training neural network model. According to the safety behavior capability evaluation scheme under the time pressure provided by the invention, the safety behavior capability index of the driver can be evaluated according to the safety behavior capability index prediction model obtained through training and the collected physiological signal of the driver, so that the driving safety of the driver is further evaluated.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of safe driving technology, and in particular to a method and device for evaluating safe behavior ability under time pressure, and electronic equipment. Background Art

[0002] The lives and safety of urban rail transit passengers directly depend on the drivers' operations and decisions. When emergencies occur, they must be solely responsible for troubleshooting, emergency recovery, passenger evacuation, and coordinate with other departments such as dispatching to perform multiple emergency processing tasks such as information confirmation and reporting. Driver operating errors and inattention are important factors in the occurrence of traffic accidents. Subway drivers often suffer from fatigue, decreased vigilance, and insufficient mental load during driving, which has a negative impact on the driver's performance in emergency situations. At the same time, because the subway's operating interval is short, the full-day operating interval is usually between 2-10 minutes, and only 2-5 minutes during peak hours. Drivers need to complete emergency troubleshooting within a tight time, less than 3 minutes. Therefore, it is necessary to construct an evaluation method for safety behavior capabilities based on load and performance under different time pressures.

[0003] Existing studies have shown that time pressure is a psychological sense of urgency caused by insufficient time to complete a task. Time pressure is a key factor affecting subway drivers' behavior and decision-making. When time is tight, excessive mental workload will affect the driver's ability to obtain information and increase the probability of road accidents. At the same time, time pressure may also lead to psychological stress, thereby affecting people's judgment and evaluation.

[0004] At present, although there have been studies on the identification of driver status during driving, status identification is only the first step in safety monitoring and rarely involves quantitative evaluation of driver status. If the driver's status cannot be accurately evaluated, accidents may occur when an emergency occurs. How to effectively monitor the driving ability of subway drivers will help improve operational safety. Summary of the invention

[0005] The purpose of the embodiments of the present invention is to provide a method and device for evaluating safety behavior ability under time pressure, and an electronic device, which can solve the problem in the prior art that the driver's status cannot be quantitatively evaluated.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] An embodiment of the present invention provides a method for evaluating safety behavior capability under time pressure, wherein the method comprises:

[0008] Obtain the driver's safety behavior ability data and physiological signal data under different time pressures;

[0009] Analyzing the correlation and time lag between the physiological index and the safety behavior ability index based on the safety behavior ability data and the physiological signal data;

[0010] According to the correlation and time lag between physiological indicators and safety behavior ability index, the neural network model is trained to obtain the safety behavior ability index prediction model;

[0011] The safety behavior ability index prediction model is used to predict the driver's safety behavior ability index based on the input driver's physiological signals.

[0012] Optionally, the safety behavior capability data includes: load and performance indicators under different time pressures, and the step of obtaining the driver's safety behavior capability data and physiological signal data includes:

[0013] Collect performance indicators of drivers under different time pressures;

[0014] Collect drivers' subjective load scores in different scenarios;

[0015] Filter out the load and performance indicators in different time periods that meet the preset conditions;

[0016] The electrocardiogram (ECG) signal and the electroencephalogram (EEG) signal of the driver under different time pressures and different scenarios are collected through a physiological data collector; wherein the physiological signal data includes: the electrocardiogram (ECG) signal and the electroencephalogram (EEG) signal.

[0017] Optionally, after the step of collecting the driver's electrocardiogram (ECG) signals and electroencephalogram (EEG) signals under different time pressures and different scenarios through a physiological data collector, the method further comprises:

[0018] The recorded EEG signals were filtered in upper and lower frequency bands using high- and low-pass filters;

[0019] Independent component analysis was used to remove blink and eye movement components from the filtered EEG signals, and the common average reference was used as the reference point of the data to obtain the bands of each EEG channel.

[0020] For each band of the EEG channel, power spectrum density, approximate entropy, sample entropy and fuzzy entropy are taken as indicators for analyzing EEG signals.

[0021] Optionally, after the step of collecting the driver's electrocardiogram (ECG) signals and electroencephalogram (EEG) signals under different time pressures and different scenarios through a physiological data collector, the method further comprises:

[0022] Perform multi-layer decomposition on ECG signals and remove low-frequency noise components;

[0023] The detail components of the 1st to 3rd layers after soft threshold processing and the detail components of the 4th to 7th layers without processing are reconstructed by wavelet to obtain a smooth signal after removing the baseline drift and power frequency interference noise.

[0024] Analyzing the smooth signal using a preset peak detection algorithm to obtain a relative risk interval;

[0025] The time domain, frequency domain and nonlinearity are extracted as indicators for analyzing ECG signals.

[0026] Optionally, the step of training a neural network model based on the correlation and time lag between the physiological index and the safety behavior ability index to obtain a safety behavior ability index prediction model includes:

[0027] The correlation and time lag between the driver's resting physiological indicators and safety behavior ability index are used as training data sets;

[0028] The correlation and time lag between the driver's physiological indicators in a non-resting state and the safety behavior ability index are used as a test data set;

[0029] A safety behavior ability index prediction model based on physiological signals and a safety behavior ability index prediction model incorporating time lag considerations were established to obtain a preset neural network model;

[0030] The preset neural network model is trained according to the training data set, and the trained neural network model is tested according to the test data set, until the trained neural network model meets the preset standard and is determined as the safety behavior ability index prediction model.

[0031] Optionally, the step of collecting the driver's electrocardiogram (ECG) and electroencephalogram (EEG) signals under different time pressures and scenarios by means of a physiological data collector includes:

[0032] The driver's EEG signals in different time periods and scenarios are collected by an electrode cap worn on the driver's head, wherein the electrode cap comprises a plurality of electrodes, each electrode is injected with EEG paste and contacts the scalp, and the sampling frequency of the EEG signals is 1000 Hz;

[0033] The driver's heart signals under different time pressures and different scenarios are collected through a wireless ECG device worn by the driver, wherein the ECG device includes: data recording software, an ECG sensor and a wireless signal receiver.

[0034] An embodiment of the present invention provides a device for evaluating safety behavior ability under time pressure, wherein the device includes:

[0035] An acquisition module, used to acquire the driver's safety behavior ability data and physiological signal data under different time pressures;

[0036] An analysis module, used to analyze the correlation and time lag between the physiological index and the safety behavior ability index based on the safety behavior ability data and the physiological signal data;

[0037] A training module is used to train a neural network model based on the correlation and time lag between the physiological index and the safety behavior ability index to obtain a safety behavior ability index prediction model;

[0038] The safety behavior ability index prediction model is used to predict the driver's safety behavior ability index based on the input driver's physiological signals.

[0039] Optionally, the acquisition module includes:

[0040] The first submodule is used to collect the performance indicators of drivers under different time pressures;

[0041] The second submodule is used to collect the driver's subjective load score in different scenarios;

[0042] The third submodule is used to screen out load and performance indicators under different time pressures that meet preset conditions; the safety behavior capability data includes: load and performance indicators under different time pressures;

[0043] The fourth submodule is used to collect the driver's electrocardiogram (ECG) and electroencephalogram (EEG) signals under different time pressures and scenarios through a physiological data collector; wherein the physiological signal data includes: electrocardiogram (ECG) signals and electroencephalogram (EEG) signals.

[0044] Optionally, the device further comprises:

[0045] A filtering module, used for performing upper and lower frequency band filtering on the recorded EEG signals through high and low pass filters after the acquisition module collects the EEG signals and EEG signals of the driver under different time pressures and different scenes through the physiological data collector;

[0046] The brain signal processing module is used to use the independent component analysis method to remove the blinking and eye movement components of the filtered EEG signal, and use the common average reference as the reference point of the data to obtain the band of each EEG channel; for the band of each EEG channel, the power spectral density, approximate entropy, sample entropy and fuzzy entropy are taken as indicators for analyzing the EEG signal.

[0047] Optionally, the device further comprises:

[0048] A decomposition module, used for performing multi-layer decomposition on the ECG signal and removing low-frequency noise components after the acquisition module collects the ECG signal and EEG signal of the driver under different time pressures and different scenes through the physiological data collector;

[0049] The ECG signal processing module is used to perform wavelet reconstruction on the detail components of the 1st to 3rd layers after soft threshold processing and the detail components of the 4th to 7th layers without processing to obtain a smooth signal after removing baseline drift and power frequency interference noise; use a preset peak detection algorithm to analyze the smooth signal to obtain a relative risk interval; and extract time domain, frequency domain and nonlinearity as indicators for analyzing ECG signals.

[0050] Optionally, the training module includes:

[0051] The fifth submodule is used to use the correlation and time lag between the driver's physiological indicators in a resting state and the safety behavior ability index as a training data set;

[0052] The sixth submodule is used to use the correlation and time lag between the driver's physiological indicators in a non-resting state and the safety behavior ability index as a test data set;

[0053] The eighth submodule is used to respectively establish a safety behavior ability index prediction model based on physiological signals and a safety behavior ability index prediction model incorporating time lag considerations to obtain a preset neural network model;

[0054] The seventh submodule is used to train the preset neural network model according to the training data set, and test the trained neural network model according to the test data set, until the trained neural network model meets the preset standard and is determined as the safety behavior ability index prediction model.

[0055] Optionally, the fourth submodule is specifically used for:

[0056] The driver's EEG signals under different time pressures and scenarios are collected through an electrode cap worn on the driver's head, wherein the electrode cap comprises a plurality of electrodes, each electrode is injected with EEG paste and contacts the scalp, and the sampling frequency of the EEG signal is 1000 Hz;

[0057] The driver's heart signals under different time pressures and different scenarios are collected through a wireless ECG device worn by the driver, wherein the ECG device includes: data recording software, an ECG sensor and a wireless signal receiver.

[0058] An embodiment of the present invention provides an electronic device, which includes a processor, a memory, and a program or instruction stored in the memory and executable on the processor. When the program or instruction is executed by the processor, the steps of any one of the above-mentioned methods for assessing safety behavior capabilities under time pressure are implemented.

[0059] An embodiment of the present invention provides a computer-readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of any one of the above-mentioned methods for assessing safety behavior capabilities under time pressure are implemented.

[0060] The safety behavior ability evaluation scheme under time pressure provided in the embodiment of the present application obtains the safety behavior ability data and physiological signal data of the driver under different time pressures; based on the safety behavior ability data and physiological signal data, analyzes the correlation and time lag between the physiological indicators and the safety behavior ability index; trains the neural network model based on the correlation and time lag between the physiological indicators and the safety behavior ability index to obtain a safety behavior ability index prediction model, and the obtained safety behavior ability index prediction model is used to predict the driver's safety behavior ability index based on the input driver's physiological signals. The safety behavior ability evaluation scheme under time pressure provided in the embodiment of the present application can provide a comprehensive evaluation of the driver's behavior ability and evaluate the abstract behavior ability; based on the evaluation results, a safety behavior ability index prediction model is constructed using physiological indicators to provide a basis for the development of the driver's state detection system and driving assistance system. The scheme can quantitatively evaluate the driver's state, which helps to timely discover potential dangerous situations caused by stress to the driver, thereby reducing the probability of accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 is a flowchart showing the steps of a method for evaluating safety behavior ability under time pressure in an embodiment of the present application;

[0062] Figure 2 is a schematic diagram showing the time lag between physiological signals and safety behavior capabilities at different times under 1NTP in an embodiment of the present application;

[0063] Figure 3 is a schematic diagram showing the time lag between physiological signals and safety behavior capabilities at different times under 2MTP of an embodiment of the present application;

[0064] Figure 4 is a schematic diagram showing the time lag between physiological signals and safety behavior capabilities at different times under 3TP in an embodiment of the present application;

[0065] Figure 5 is a flowchart showing a safety behavior capability index construction under time pressure in an embodiment of the present application;

[0066] Figure 6 It is a graph representing the PCA-SVDD training model;

[0067] Figure 7 It represents the change of NASA-TLX in the embodiment of the present application with driving time;

[0068] Figure 8 It represents the performance of the embodiment of the present application changing with driving time;

[0069] Fig. 9 is a schematic diagram showing the relationship between the safety behavior ability index, load and performance under different time pressures in an embodiment of the present application;

[0070] Fig.10 is an index of safety behavior ability under three time pressures in the embodiment of the present application;

[0071] Fig.11 is a graph showing the changing trend of the safety behavior capability index under different time pressures in the embodiment of the present application;

[0072] Fig.12 is a GA-BP neural network model framework representing an embodiment of the present application;

[0073] Fig.13 is a schematic diagram showing an electrocardiogram signal before and after filtering according to an embodiment of the present application;

[0074] Fig.14 is a neural network prediction result after optimization by genetic algorithm in the embodiment of the present application;

[0075] Fig.15 It represents the prediction result of the GA-BP model under 4NTP of the embodiment of the present application;

[0076] Fig.16 It represents the prediction result of the GA-BP model under 5MTP of the embodiment of the present application;

[0077] Fig.17 It represents the prediction result of the GA-BP model under 6TP in the embodiment of the present application;

[0078] Fig.18 It represents the prediction result of the DARNN model under NTP of the embodiment of the present application;

[0079] Fig.19 It represents the prediction result of the DARNN model under MTP of the embodiment of the present application;

[0080] Fig. 20 It represents the prediction result of the DARNN model under TP of the embodiment of the present application;

[0081] Fig.21 It is a structural block diagram of a device for evaluating safety behavior ability under time pressure according to an embodiment of the present application. DETAILED DESCRIPTION

[0082] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0083] The embodiment of the present application provides a method for evaluating safety behavior ability based on load and performance under different time pressures, and proposes a safety behavior ability index prediction model based on physiological data, which can improve the recognition of the driver's state during driving and help to quantitatively evaluate the driving level of subway drivers.

[0084] The following, in conjunction with the accompanying drawings, describes in detail the safety behavior ability assessment solution under time pressure provided by the embodiment of the present application through specific embodiments and their application scenarios.

[0085] As attached Figure 1 As shown, the method for evaluating safety behavior ability under time pressure in the embodiment of the present application includes the following steps:

[0086] Step 101: Acquire the driver's safety behavior ability data and physiological signal data under different time pressures.

[0087] The method for evaluating the ability to conduct safe behaviors under time pressure provided in the embodiment of the present application can be applied to electronic devices. The storage medium of the electronic device stores a computer program corresponding to the method for evaluating the ability to conduct safe behaviors under time pressure. The processor executes the computer program to evaluate the ability to conduct safe behaviors of the driver during driving. The safety behavior ability data includes: load and performance indicators under different time pressures. A feasible way to obtain the driver's safety behavior ability data and physiological signal data may include the following sub-steps:

[0088] Sub-step 1: Collecting performance indicators of drivers under different time pressures;

[0089] The driver can be a subway driver, a bus driver, etc. In the embodiment of the present application, the driver is a subway driver as an example. The performance data of the subway driver under different time pressures are collected, and three experimental methods are used to test the performance data of the subway driver under different time pressures based on the actual working scenes of the subway driver, specifically: the selection reaction based on the oddball experiment, the discrimination reaction based on the Go / No-go and Stroop experimental paradigms, and the call reaction.

[0090] The specific selection response task is as follows: a group of pictures will be presented on the screen, and the subjects will observe the font color in the middle of the screen and the meaning of the words in the four boxes. They are required to select the meaning of the words that match the color displayed in the middle font from the four boxes. For example, if the central font color is red, then the subjects need to select the meaning of the words that represent red from the four borders and press the corresponding key. Specifically, corresponding to the meanings of "yellow", "red", "green" and "blue", press the "w", "a", "s" and "e" keys on the keyboard respectively.

[0091] The specific task of discrimination response is that the subjects need to accurately discriminate the presented stimuli according to specific conditions and make corresponding responses. Specifically, when observing that "red" lights up and hearing "high pitch", the subjects need to press the "h" key on the keyboard; and when observing that "green" lights up and hearing "low pitch", the subjects need to press the "l" key on the keyboard. In other cases, the subjects do not need to respond.

[0092] The specific task of the call response is: the subject will hear a paragraph and be asked to repeat it completely. If the subject cannot repeat it completely, the staff will play the paragraph again for the subject until the subject repeats it completely. For example, the script used for the test can be set as: Due to AST failure, the Yishan Road-Yan'an West Road driving mode is changed to SM mode, and the monitoring of operational parking is strengthened.

[0093] Sub-step 2: Collecting drivers' subjective load scores in different scenarios;

[0094] In the actual implementation process, the NASA-TLX task load scale can be used to measure the workload by verbally expressing the test subjects, avoiding noise, talking to people, and improving accuracy. The change of NASA-TLX with driving time is shown in the attached figure. Figure 7 shown.

[0095] The NASA-TLX load scale is shown in Table 1:

[0096] Table 1

[0097]

[0098]

[0099] Sub-step three: Filter out the load and performance indicators in different time periods that meet the preset conditions.

[0100] In this optional embodiment, load and performance indicators that can characterize the subway driver under different time pressures are screened out. The score of the NASA-TLX scale is used as the load indicator of the subway driver. Combined with the methodology of signal detection theory, the performance indicators of the subway driver are constructed, including hit rate, missed alarm rate, false alarm rate, correct rejection rate, discrimination index and response bias. Among them, the performance changes with driving time are shown in the attached figure. Figure 8 shown.

[0101] The performance indicators based on signal detection theory are shown in Table 2:

[0102] Table 2

[0103]

[0104] Sub-step 4: Collect the driver's electrocardiogram (ECG) and electroencephalogram (EEG) signals under different time pressures and scenarios through a physiological data collector.

[0105] Among them, physiological signal data includes: electrocardiogram signals and electroencephalogram signals.

[0106] An optional method of collecting the driver's electrocardiogram and electroencephalogram signals under different time pressures and different scenarios through a physiological data collector may be as follows:

[0107] The driver's EEG signals in different time periods and scenarios are collected through the electrode cap worn on the driver's head. The electrode cap contains multiple electrodes, each electrode is injected with EEG paste and contacts the scalp, and the sampling frequency of the EEG signal is 1000Hz. The number of electrodes contained in the electrode cap can vary depending on the model of the electrode cap. For example, an electrode cap containing 59 electrodes can be selected. Each electrode is injected with EEG paste and contacts the scalp to reduce resistance.

[0108] The driver's heart signals under different time pressures and scenarios are collected through a wireless ECG device worn by the driver. The ECG device includes: data recording software, ECG sensor and wireless signal receiver. In the actual implementation process, the T-Sens ECG sensor can be selected.

[0109] In a preferred embodiment, after collecting the electrocardiogram (ECG) signals and electroencephalogram (EEG) signals of the driver under different time pressures and in different scenarios, the collected ECG signals and EEG signals need to be preprocessed.

[0110] The preprocessing of EEG signals can be carried out in the following manner, including the following sub-steps:

[0111] Sub-step 1: Perform upper and lower frequency band filtering on the recorded EEG signal using high-pass and low-pass filters; specifically, the data can be filtered in the upper and lower frequency bands of 1 Hz to 30 Hz using high-pass and low-pass filters, respectively.

[0112] Sub-step 2: Using independent component analysis to remove blink and eye movement components from the filtered EEG signal, and using the common average reference as the reference point of the data to obtain the band of each EEG channel;

[0113] Sub-step 2: For each band of the EEG channel, take the power spectral density, approximate entropy, sample entropy and fuzzy entropy as indicators for analyzing the EEG signal.

[0114] The preprocessing of the ECG signal can be carried out in the following manner, including the following sub-steps:

[0115] Sub-step 1: Perform multi-layer decomposition on the ECG signal and remove the low-frequency noise component; specifically, the ECG signal can be decomposed into three components using the db6 wavelet.

[0116] Sub-step 2: Use the soft threshold to process the detail components of the 1st to 3rd layers and the unprocessed detail components of the 4th to 7th layers to perform wavelet reconstruction to obtain a smooth signal after removing the baseline drift and power frequency interference noise; the soft threshold is the variable threshold.

[0117] Sub-step 3: Use the preset peak detection algorithm to analyze the smooth signal to obtain the relative risk interval; specifically, the Pan-Tompkins peak detection algorithm can be used to obtain the RR interval, i.e., the relative risk interval. The schematic diagram of the ECG signal before and after filtering is shown in the attached figure. Fig.13 shown.

[0118] Sub-step 4: Extract time domain, frequency domain and nonlinearity as indicators for analyzing ECG signals.

[0119] Step 102: Based on the safety behavior ability data and the physiological signal data, the correlation and time lag relationship between the physiological index and the safety behavior ability index are analyzed.

[0120] The relationship between physiological indicators and safety behavior ability index is analyzed, and the relevant indicators are selected to predict the safety behavior ability index using the GA-BP neural network model to achieve intelligent monitoring of safety behavior ability. Based on the safety behavior ability data and physiological signal data, the Pearson correlation analysis is performed on the indicators of analyzing EEG signals and analyzing ECG signals and the safety behavior ability index, respectively, to obtain the relationship between physiological indicators and safety behavior ability index, which includes the correlation and time lag between the two.

[0121] This step can be used to analyze under what time pressure the driver's safety behavior ability can be achieved optimally. The influence of time pressure on physiological signals is analyzed. By performing a difference test on the physiological data of subway drivers under different time pressures, the results confirm that time pressure has a significant impact on physiological signals; the intrinsic connection between physiological signals and safety behavior ability is further analyzed, and physiological signal indicators with correlation are selected; from the perspective of the mutual control of physiology and action, the time lag relationship between the two is analyzed. The results show that the change of physiological signals precedes the change of behavior. In addition, the time lag relationship between the two at different times is explored. The results show that the mutual correlation between the two at different times presents different changing trends, and generally maintains the time lag relationship that the change of physiological signals precedes the change of behavior.

[0122] The correlation analysis between EEG index (i.e., EEG signal index) and safety behavior ability index is shown in Table 3:

[0123] Table 3

[0124] EEG indicators Correlation coefficient (significance level P) Power Spectral Density -0.8197(0.0323) Approximate Entropy 0.0842(0.6957) Fuzzy Entropy 0.08041(0.7088) Sample Entropy -0.7532(0.0475)

[0125] The correlation analysis between ECG index (i.e., electrocardiogram index) and safety behavior ability index is shown in Table 4:

[0126] Table 4

[0127]

[0128]

[0129] Analyze the time lag between physiological signals and safety behavior ability under time pressure. Use the cross-correlation function to measure the time lag relationship between the two. In order to understand the correlation between physiological signals and safety behavior ability at different time points in detail, three moments were selected to further analyze the intrinsic relationship between the two at different moments. The cross-correlation between the safety behavior ability index and the physiological signal index and the cross-correlation between the two at different moments were analyzed to obtain the time delay between the two. The results show that the physiological signal index changes ahead of human behavior, which provides information support for the safety behavior ability index prediction model below. Figure 2 is a schematic diagram showing the time lag between physiological signals and safety behavior capabilities at different times under 7NTP in an embodiment of the present application; Figure 3 is a schematic diagram showing the time lag between physiological signals and safety behavior capabilities at different times under 8 MTP of an embodiment of the present application; Figure 4 It is a schematic diagram showing the time lag between physiological signals and safety behavior capabilities at different times under 9TP in an embodiment of the present application.

[0130] Step 103: training a neural network model based on the correlation and time lag between the physiological index and the safety behavior ability index to obtain a safety behavior ability index prediction model.

[0131] In this step, the SVDD can be used to construct a safety behavior capability index. An exemplary flow chart for constructing a safety behavior capability index under time pressure is shown in the attached figure. Figure 5 As shown in FIG, the specific steps include: Step 1: First, the load and performance indicators (NASA-TLX, hit rate, false alarm rate, discrimination, reaction time 2, and voice performance) of subway drivers in a resting state are collected, and the weighted fusion is completed with the help of the PCA method to obtain the fusion feature indicators; Step 2: The processed sample data in the resting state is used as training data and sent to the SVDD model for training to obtain the center a and the radius R of the sphere; Step 3: The data to be tested is also processed using the PCA algorithm to obtain the feature indicators; Step 4: The feature indicators of the data to be tested are input into the trained SVDD model to obtain the distance D between the sample point and the hypersphere, which is used as the basis for evaluating the safety behavior capability index; Step 5: D is compared with R. If D and R are different, the safety behavior capability can be evaluated.

[0132] During construction, the acquired data is divided into a training set and a test set. The data of subway drivers in resting state is the training set, and the data in other states is the test set. Among them, SVDD Support Vector Data Description (Support Vector Data Description) is a single-value classification algorithm that can distinguish target samples from non-target samples.

[0133] The factors affected by time pressure are summarized, including NASA-TLX, hit rate, false alarm rate, discrimination, reaction time 2, and speech performance. The PCA algorithm is used to analyze the contribution of the above influencing factors in the data sample. The key factors extracted by the PCA algorithm are shown in Table 5:

[0134] Table 5

[0135] name Variance explained (%) Cumulative variance explained (%) Weight (%) Principal component 1 0.749 74.870 75.085 Principal component 2 0.228 97.667 22.863 Principal component 3 0.013 98.985 1.322 Principal component 4 0.007 99.713 0.730

[0136] It can be seen from the table that the cumulative contribution rate of principal component 1 and principal component 2 is greater than 95%, so the first two principal components are selected as comprehensive indicators for evaluating the safety behavior ability index. The SVDD model is constructed using the data of the subjects in the resting state as the training set, and the penalty coefficient C and the kernel parameter are set within the range for grid optimization. The radius of the hypersphere R is calculated to be 1.031, and the data in other states are input into the model to obtain D. D and R are brought into the PCA-SVDD training model to obtain the safety behavior ability index. Figure 6 shown.

[0137] Fig. 9 is a schematic diagram showing the relationship between the safety behavior ability index, load and performance under different time pressures in an embodiment of the present application;

[0138] Among them, Figure (a) shows the law between the safety behavior ability index and the load and performance indicators under low time pressure; Figure (b) shows the law between the safety behavior ability index and the load and performance indicators under medium time pressure; Figure (c) shows the law between the safety behavior ability index and the load and performance indicators under high time pressure. The safety behavior ability index under three time pressures is shown in the attached figure. Fig.10 As shown in the attached figure, the trend of safety behavior ability index changes under different time pressures is shown in the attached figure. Fig.11 As shown. Fig.10 , 11 It can be seen that the safety behavior ability index is different under different time pressures. Finally, the changes in the safety behavior ability index under different time pressures are analyzed, and it is concluded that the safety behavior ability of subway drivers is stronger under NTP.

[0139] In an optional embodiment, a neural network model such as a GA-BP neural network model is trained based on the relationship between the physiological index and the safety behavior ability index to obtain a safety behavior ability index prediction model as follows:

[0140] The relationship between the driver's physiological indicators in a resting state and the safety behavior ability index is used as a training data set; the relationship between the driver's physiological indicators in a non-resting state and the safety behavior ability index is used as a test data set; the preset neural network model is trained based on the training data set, and the trained neural network model is tested based on the test data set until the trained neural network model meets the preset standards and is determined as a safety behavior ability index prediction model.

[0141] Among them, the safety behavior ability index prediction model is used to predict the driver's safety behavior ability index based on the input driver's physiological signals. The prediction results of the neural network optimized by the genetic algorithm (i.e., the safety behavior ability index prediction model after training) are shown in the attached figure. Fig.14 shown.

[0142] The safety behavior capability index prediction model uses power spectral density, sample entropy, RMSSD, NN50, SD1, and CVI as input indicators, and the safety behavior capability index as the output indicator.

[0143] The data is preprocessed using the following formula:

[0144]

[0145] Among them, x' is the data after variable normalization. is the mean value of a variable in the sample, x is the original data of the variable, and δ is the standard deviation of this variable in the sample.

[0146] Among them, the number of hidden layers and neurons is one of the important factors affecting the accuracy of the model. If the number is too large, it will lead to overfitting. Therefore, the structural design of the hidden layer is particularly important.

[0147] Since the essence of GA-BP neural network is based on BP neural network, different numbers of neurons are used for training comparison through trial and error. The final model structure is 5-5-1, that is, the number of neurons in the input layer is 5, the number of neurons in the hidden layer is 5, and the number of neurons in the output layer is 1. The GA-BP neural network model framework of the embodiment of the present application is shown in the attached figure. Fig.12 shown.

[0148] The parameter settings of the safety behavior ability index prediction model are shown in Table 6:

[0149] Table 6

[0150]

[0151] In an optional embodiment, there is a close correlation between physiological signals and safety behavior capabilities, but after analysis, it is found that the two are not in real-time correspondence, but there is a certain time lag. This time lag may be caused by the delay of physiological reactions, the time difference between signal acquisition and processing, or the process of formation and display of safety behavior capabilities. Deep Autoregressive Neural Network (DARNN) has significant advantages in dealing with time series lag, which makes it perform well in capturing and modeling dynamic features in time series data. The method of obtaining a safety behavior capability index prediction model based on the DARNN model can be as follows:

[0152] To build a DARNN model, the steps are as follows:

[0153] The first step is to determine the number of lags required. The autoregressive structure of the time series, that is, the number of lags that need to be considered, is determined by analyzing the autocorrelation function of the time series data. The determined number of lags is used as a feature to construct a lag matrix. Each row of the matrix represents a feature vector at a time point, which contains the observed values ​​of that time point and several lag time points before it. For example, when the number of lags is 2, the time series data can be constructed into a lag matrix as follows:

[0154] [X1 X2 X3]

[0155] t-2 t-1 t

[0156] Among them, represents the observed value of the time series at different time points, where is the input feature of the model, and the input features are different under different time pressures, representing the time point. It is used to predict the observed value at time point t. In this section, the input of DARNN is established according to different time pressures. The physiological indicators that are correlated with the safety behavior ability index under the corresponding time pressure are selected, and the output vector is the safety behavior ability index. Specifically, under NTP, the input vectors are Beta / Alpha, Theta / Alpha, SDNN, RMSSD, NN50, PNN50, VLF, SD1, SD2, CSI, CVI; under MTP, the input vectors are Beta / Alpha, Theta / Alpha, RMSSD, NN50, SD1, CVI; under TP, the input vectors are Theta / Alpha, SDNN, RMSSD, PNN50, HF, SD1, SD2, CSI, CVI.

[0157] The second step is model construction. The physiological signals at each time point in the lag matrix and the previous few time points are used as input features to predict the safety behavior ability index at the next time point. The feature data and target data are divided into training sets and test sets respectively. The data are usually arranged in chronological order to ensure the generalization ability of the model on future data. When dividing the data, the chronological relationship between the training set and the test set should be maintained. The prepared training set and test set data are loaded into the deep autoregressive neural network model for model training and evaluation. Common evaluation indicators such as MAE, MSE, RMSE, and MAPE are used to evaluate the prediction effect of the model.

[0158] Step 3: Compare different lags. Repeat steps 1 and 2 for different numbers of lags and compare the prediction performance of the model with different numbers of lags. Analyze the prediction error of the model with different numbers of lags and select the number of lags that provides the best prediction results.

[0159] Step 4: Model application: Select the optimal number of lag terms and model parameters, retrain the deep autoregressive neural network, and apply it to the actual time series forecasting task.

[0160] By introducing a moderate time lag term, the model can more accurately predict observations at future times.

[0161] Table 7 shows the parameter settings of the DARNN model, and Table 8 shows the prediction results of the DARNN model:

[0162] Table 7

[0163]

[0164]

[0165] Table 8

[0166]

[0167]

[0168] In the case of NTP, when the hysteresis term is 3, the model prediction result is closer to the actual value. In the case of MTP, when the hysteresis term is 4, the model prediction result is closer to the actual value. In the case of TP, when the hysteresis term is 3, the model prediction result is closer to the actual value. Fig.18 It represents the prediction result of the DARNN model under NTP of the embodiment of the present application; Fig.19 It represents the prediction result of the DARNN model under MTP of the embodiment of the present application; Fig. 20 It represents the prediction result of the DARNN model under TP of the embodiment of the present application.

[0169] The method for evaluating the safety behavior ability under time pressure provided by the embodiment of the present application obtains the safety behavior ability data and physiological signal data of the driver under different time pressures; constructs the driver's safety behavior ability index based on load and performance data, analyzes under what time pressure the driver's safety behavior ability index is optimal, and analyzes the correlation and time lag between the physiological index and the safety behavior ability index based on the safety behavior ability data and physiological signal data; trains the neural network model based on the relationship between the physiological index and the safety behavior ability index to obtain a safety behavior ability index prediction model, and the obtained safety behavior ability index prediction model is used to predict the driver's safety behavior ability index based on the input driver's physiological signal. Combined with the time lag of the physiological signal, the prediction effect of the model is optimized. The safety behavior ability evaluation scheme under time pressure provided by the embodiment of the present application can provide a comprehensive evaluation of the driver's behavior ability and evaluate the abstract behavior ability; based on the evaluation results, the safety behavior ability index prediction model is constructed using physiological indicators to provide a basis for the development of the driver's state detection system and driving assistance system. The scheme can quantitatively evaluate the driver's state, which helps to timely discover potential dangerous situations caused by stress to the driver, thereby reducing the probability of accidents.

[0170] Fig.21 A structural block diagram of a device for evaluating safety behavior ability under time pressure according to an embodiment of the present application.

[0171] The safety behavior ability assessment device under time pressure provided in the embodiment of the present application includes the following functional modules:

[0172] An acquisition module 1001 is used to acquire the driver's safety behavior ability data and physiological signal data under different time pressures;

[0173] An analysis module 1002 is used to analyze the correlation and time lag between the physiological index and the safety behavior ability index based on the safety behavior ability data and the physiological signal data;

[0174] A training module 1003 is used to train a neural network model based on the relationship between the physiological index and the safety behavior ability index to obtain a safety behavior ability index prediction model;

[0175] The safety behavior ability index prediction model is used to predict the driver's safety behavior ability index based on the input driver's physiological signals.

[0176] Optionally, the acquisition module includes:

[0177] The first submodule is used to collect the performance indicators of drivers under different time pressures;

[0178] The second submodule is used to collect the driver's subjective load score in different scenarios;

[0179] The third submodule is used to screen out load and performance indicators under different time pressures that meet preset conditions; the safety behavior capability data includes: load and performance indicators under different time pressures;

[0180] The fourth submodule is used to collect the driver's electrocardiogram (ECG) and electroencephalogram (EEG) signals under different time pressures and scenarios through a physiological data collector; wherein the physiological signal data includes: electrocardiogram (ECG) signals and electroencephalogram (EEG) signals.

[0181] Optionally, the device further comprises:

[0182] A filtering module, used for performing upper and lower frequency band filtering on the recorded EEG signals through high and low pass filters after the acquisition module collects the EEG signals and EEG signals of the driver under different time pressures and different scenes through the physiological data collector;

[0183] The brain signal processing module is used to use the independent component analysis method to remove the blinking and eye movement components of the filtered EEG signal, and use the common average reference as the reference point of the data to obtain the band of each EEG channel; for the band of each EEG channel, the power spectral density, approximate entropy, sample entropy and fuzzy entropy are taken as indicators for analyzing the EEG signal.

[0184] Optionally, the device further comprises:

[0185] A decomposition module, used for performing multi-layer decomposition on the ECG signal and removing low-frequency noise components after the acquisition module collects the ECG signal and EEG signal of the driver under different time pressures and different scenes through the physiological data collector;

[0186] The ECG signal processing module is used to perform wavelet reconstruction on the detail components of the 1st to 3rd layers after soft threshold processing and the detail components of the 4th to 7th layers without processing to obtain a smooth signal after removing baseline drift and power frequency interference noise; use a preset peak detection algorithm to analyze the smooth signal to obtain a relative risk interval; and extract time domain, frequency domain and nonlinearity as indicators for analyzing ECG signals.

[0187] Optionally, the training module includes:

[0188] The fifth submodule is used to use the correlation and time lag between the driver's physiological indicators in a resting state and the safety behavior ability index as a training data set;

[0189] The sixth submodule is used to use the correlation and time lag between the driver's physiological indicators in a non-resting state and the safety behavior ability index as a test data set;

[0190] The eighth submodule is used to respectively establish a safety behavior ability index prediction model based on physiological signals and a safety behavior ability index prediction model incorporating time lag considerations to obtain a preset neural network model;

[0191] The seventh submodule is used to train the preset neural network model according to the training data set, and test the trained neural network model according to the test data set, until the trained neural network model meets the preset standard and is determined as the safety behavior ability index prediction model.

[0192] Optionally, the fourth submodule is specifically used for:

[0193] The driver's EEG signals under different time pressures and scenarios are collected through an electrode cap worn on the driver's head, wherein the electrode cap comprises a plurality of electrodes, each electrode is injected with EEG paste and contacts the scalp, and the sampling frequency of the EEG signal is 1000 Hz;

[0194] The driver's heart signals under different time pressures and different scenarios are collected through a wireless ECG device worn by the driver, wherein the ECG device includes: data recording software, an ECG sensor and a wireless signal receiver.

[0195] The safety behavior ability assessment device under time pressure provided in the embodiment of the present application can provide a comprehensive assessment of the driver's behavior ability and evaluate the abstract behavior ability; based on the assessment results, a safety behavior ability index prediction model is constructed using physiological indicators to provide a basis for the development of the driver's state detection system and driving assistance system. The device can quantitatively assess the driver's state, which helps to timely discover potential dangerous situations caused by the driver's stress, thereby reducing the probability of accidents.

[0196] In the present application embodiment Fig.21 The safety behavior ability assessment device under time pressure shown is set in an electronic device or a server. The electronic device or server provided with the device can be a device with an operating system. The operating system can be an Android operating system, an iOS operating system, or other possible operating systems, which are not specifically limited in the embodiments of the present application.

[0197] The embodiments of the present application provide Fig.21 The safety behavior ability assessment device under time pressure shown can achieve Figure 1 To avoid repetition, the various processes implemented by the method embodiment are not described here.

[0198] Optionally, an embodiment of the present application also provides an electronic device, including a processor, a memory, and a program or instruction stored in the memory and executable on the processor. When the program or instruction is executed by the processor, the various processes performed by the safety behavior capability assessment device under the above-mentioned time pressure are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0199] It should be noted that the electronic device in the embodiment of the present application includes the server described above.

[0200] The processor is a processor in the electronic device described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0201] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0202] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for evaluating safety behavior ability under time pressure, characterized in that: include: Obtain the driver's safety behavior ability data and physiological signal data under different time pressures; Based on the safety behavior ability data and the physiological signal data, analyzing the correlation and time lag between the physiological index and the safety behavior ability index; According to the correlation and time lag between physiological indicators and safety behavior ability index, the neural network model is trained to obtain the safety behavior ability index prediction model; The safety behavior ability index prediction model is used to predict the driver's safety behavior ability index based on the input driver's physiological signals.

2. The method according to claim 1, characterized in that The safety behavior capability data includes: load and performance indicators under different time pressures. The step of obtaining the driver's safety behavior capability data and physiological signal data includes: Collect performance indicators of drivers under different time pressures; Collect drivers' subjective load scores in different scenarios; Filter out the load and performance indicators in different time periods that meet the preset conditions; The electrocardiogram (ECG) signal and the electroencephalogram (EEG) signal of the driver under different time pressures and different scenarios are collected through a physiological data collector; wherein the physiological signal data includes: the electrocardiogram (ECG) signal and the electroencephalogram (EEG) signal.

3. The method according to claim 2, characterized in that After the step of collecting the driver's electrocardiogram and electroencephalogram signals under different time pressures and different scenarios through the physiological data collector, the method further includes: The recorded EEG signals were filtered in upper and lower frequency bands using high- and low-pass filters; Independent component analysis was used to remove blink and eye movement components from the filtered EEG signals, and the common average reference was used as the reference point of the data to obtain the bands of each EEG channel. For each band of the EEG channel, power spectrum density, approximate entropy, sample entropy and fuzzy entropy are taken as indicators for analyzing EEG signals.

4. The method according to claim 3, characterized in that After the step of collecting the driver's electrocardiogram and electroencephalogram signals under different time pressures and different scenarios through the physiological data collector, the method further includes: Perform multi-layer decomposition on ECG signals and remove low-frequency noise components; The detail components of the 1st to 3rd layers after soft threshold processing and the detail components of the 4th to 7th layers without processing are reconstructed by wavelet to obtain a smooth signal after removing the baseline drift and power frequency interference noise. Analyzing the smooth signal using a preset peak detection algorithm to obtain a relative risk interval; The time domain, frequency domain and nonlinearity are extracted as indicators for analyzing ECG signals.

5. The method according to claim 1, characterized in that The steps of training a neural network model based on the correlation and time lag between the physiological index and the safety behavior ability index to obtain a safety behavior ability index prediction model include: The correlation and time lag between the driver's resting physiological indicators and safety behavior ability index are used as training data sets; The correlation and time lag between the driver's physiological indicators in a non-resting state and the safety behavior ability index are used as a test data set; A safety behavior ability index prediction model based on physiological signals and a safety behavior ability index prediction model incorporating time lag considerations were established to obtain a preset neural network model; The preset neural network model is trained according to the training data set, and the trained neural network model is tested according to the test data set, until the trained neural network model meets the preset standard and is determined as the safety behavior ability index prediction model.

6. The method according to claim 3, characterized in that The step of collecting the driver's electrocardiogram (ECG) and electroencephalogram (EEG) signals under different time pressures and scenarios through a physiological data collector includes: The driver's EEG signals in different time periods and scenarios are collected by an electrode cap worn on the driver's head, wherein the electrode cap comprises a plurality of electrodes, each electrode is injected with EEG paste and contacts the scalp, and the sampling frequency of the EEG signals is 1000 Hz; The driver's heart signals under different time pressures and different scenarios are collected through a wireless ECG device worn by the driver, wherein the ECG device includes: data recording software, an ECG sensor and a wireless signal receiver.

7. A device for evaluating safety behavior ability under time pressure, the device being used to implement the method according to any one of claims 1 to 6, characterized in that: include: An acquisition module, used to acquire the driver's safety behavior ability data and physiological signal data under different time pressures; An analysis module, used to analyze the correlation and time lag between the physiological index and the safety behavior ability index based on the safety behavior ability data and the physiological signal data; A training module is used to train a neural network model based on the correlation and time lag between the physiological index and the safety behavior ability index to obtain a safety behavior ability index prediction model; The safety behavior ability index prediction model is used to predict the driver's safety behavior ability index based on the input driver's physiological signals.

8. An electronic device, characterized in that: The electronic device comprises: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are loaded and executed by the processor, the method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program code, which can be called by a processor to execute the method according to any one of claims 1 to 6.

Citation Information

Cited By

  • EEG (electroencephalogram) classification method based on multi-domain feature fusion

    CN120892917A