Risk assessment management system and method based on neural recognition technology

Abstract

This application discloses a risk assessment management system and method based on neural recognition technology, which analyzes the grayscale image of the EEG signal frequency domain by using data processing and image analysis algorithms based on artificial intelligence and deep learning, so as to capture the semantic features of the EEG signal frequency domain image, so as to monitor the emotional state of the operating object in real time and detect anomalies. If an abnormal emotion is detected, an alarm is immediately triggered to remind relevant personnel to pay attention and take corresponding measures. In this way, the emotional state or perceived stress level of an individual can be automatically estimated by analyzing the EEG signal of the operating object, and then potential operational risks can be identified and warned, so as to achieve more intelligent operational risk assessment management and provide more accurate and effective support for operational risk management in high-risk scenarios.

Description

Technical Field

[0001] The present application relates to the field of intelligent management, and more specifically, to a risk assessment management system and method based on neural recognition technology. Background Art

[0002] In today's highly interconnected and rapidly changing economic environment, risk management has become one of the key factors for enterprises and organizations to ensure continued success. Especially in the financial industry, risk management is even more critical. Financial institutions need to face various internal and external risks, including market risk, credit risk, operational risk, etc. Among them, operational risk refers to the risk of loss caused by imperfect or failed internal processes, personnel, systems or external events. In actual operations, non-financial factors such as employees' emotional state and psychological pressure may also become important causes of operational risks.

[0003] However, many institutions rely on a set of pre-set rules to monitor and manage operational risks, such as setting transaction limits and requiring double-checking. However, this approach lacks flexibility, is difficult to adapt to complex and changing operating environments, and cannot effectively capture unstructured information such as employee mood swings. In addition, some financial institutions use historical data to build statistical models to predict future operational risks. Although this approach can provide certain forward-looking guidance, it focuses mainly on quantifiable risk indicators and ignores the emotional and psychological factors in human behavior, which are often unpredictable and difficult to quantify. Moreover, although regular mental health assessments and job satisfaction surveys can help understand the mental state of employees and help identify potential problems, it is difficult to achieve real-time monitoring due to its high subjectivity and low frequency.

[0004] Therefore, an optimized risk assessment management system is desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a risk assessment management system and method based on neural recognition technology, which analyzes the grayscale image of the EEG signal frequency domain by utilizing data processing and image analysis algorithms based on artificial intelligence and deep learning, so as to capture the semantic features of the EEG signal frequency domain image, thereby performing real-time monitoring of the emotional state of the operating object and abnormality detection. If abnormal emotions are detected, an alarm is immediately triggered to remind relevant personnel to pay attention and take corresponding measures. In this way, the emotional state or perceived stress level of the individual can be automatically estimated by analyzing the EEG signal of the operating object, and then potential operational risks can be identified and warned, achieving more intelligent operational risk assessment management, and providing more accurate and effective support for operational risk management in high-risk scenarios.

[0006] According to one aspect of the present application, a risk assessment management system based on neural recognition technology is provided, which includes:

[0007] An electroencephalogram (EEG) signal acquisition module is used to acquire the EEG signal of the subject through an EEG device;

[0008] An EEG signal preprocessing module, used for obtaining the EEG signal of the operation object and preprocessing the EEG signal to obtain a grayscale image of the EEG signal in the frequency domain;

[0009] An EEG frequency domain feature extraction module is used to extract EEG frequency domain features from the EEG signal frequency domain grayscale image to obtain EEG signal frequency domain image semantic coding features;

[0010] An EEG signal frequency domain feature enhancement module is used to perform spatial modulation feature enhancement processing on the EEG signal frequency domain image semantic coding feature to obtain the EEG signal frequency domain image semantic enhancement coding feature;

[0011] The operation risk assessment management module is used to perform emotion recognition based on the semantic enhancement coding features of the EEG signal frequency domain image to determine whether the emotion of the operation object is abnormal, and determine whether to generate an operation risk warning prompt.

[0012] According to another aspect of the present application, a risk assessment management method based on neural recognition technology is provided, which includes:

[0013] Collecting the EEG signal of the subject through an EEG device;

[0014] After obtaining the EEG signal of the operation object, preprocessing the EEG signal to obtain a grayscale image of the EEG signal in the frequency domain;

[0015] Extracting EEG frequency domain features from the EEG signal frequency domain grayscale image to obtain EEG signal frequency domain image semantic coding features;

[0016] Performing spatial modulation feature enhancement processing on the semantic coding feature of the EEG signal frequency domain image to obtain semantic enhancement coding feature of the EEG signal frequency domain image;

[0017] Emotion recognition is performed based on the semantic enhancement coding features of the EEG signal frequency domain image to determine whether the emotions of the operating object are abnormal, and to determine whether to generate an operation risk warning prompt.

[0018] Compared with the prior art, the risk assessment management system and method provided by the present application, which is based on neural recognition technology, analyzes the grayscale image of the EEG signal frequency domain by using data processing and image analysis algorithms based on artificial intelligence and deep learning, so as to capture the semantic features of the EEG signal frequency domain image, thereby performing real-time monitoring of the emotional state of the operating object and abnormality detection. If abnormal emotions are detected, an alarm is immediately triggered to remind relevant personnel to pay attention and take corresponding measures. In this way, the emotional state or perceived stress level of an individual can be automatically estimated by analyzing the EEG signal of the operating object, and then potential operational risks can be identified and warned, achieving more intelligent operational risk assessment management, and providing more accurate and effective support for operational risk management in high-risk scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0020] Figure 1 is a block diagram of a risk assessment management system based on neural recognition technology according to an embodiment of the present application;

[0021] Figure 2 A data flow diagram of a risk assessment management system based on neural recognition technology according to an embodiment of the present application;

[0022] Figure 3 It is a block diagram of an EEG signal frequency domain feature enhancement module in a risk assessment management system based on neural recognition technology according to an embodiment of the present application;

[0023] Figure 4 Flow chart of a risk assessment management method based on neural recognition technology according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0025] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0026] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0027] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0028] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0029] It should be understood that a person's emotional state has a significant impact on his or her behavior. For example, in the process of financial transactions, traders' emotional fluctuations (such as overconfidence, panic, etc.) may lead to irrational decisions, which in turn may cause market fluctuations or personal asset losses. Therefore, timely and accurate monitoring of the operator's abnormal emotional state and taking corresponding intervention measures are important for preventing operational risks and ensuring the stable operation of the system.

[0030] Electroencephalogram (EEG) is a physiological signal that directly reflects brain activity. Its changes can reveal an individual's emotional state, cognitive process, and mental health. In recent years, with the development of brain-computer interface technology, the research on emotion recognition using EEG signals has gradually become a hot topic. By analyzing specific patterns in EEG signals, such as energy distribution in different frequency bands and time-frequency characteristics, the individual's emotional state, such as anxiety, tension, relaxation, etc., can be effectively estimated.

[0031] Based on this, in the technical solution of this application, a risk assessment management system based on neural recognition technology is proposed, which can use neural recognition technology and artificial intelligence technology to collect and analyze the EEG signals of the operating object, so as to realize real-time monitoring of the emotional state of the operating process and early warning of operational risks. Specifically, firstly, the original EEG signal is collected by using EEG equipment; then, the data is preprocessed by using advanced signal processing technology, including signal decomposition, filtering and other steps to remove noise interference and retain useful information; then, the processed EEG signal is converted into a two-dimensional time-frequency spectrum diagram, and further converted into an EEG signal frequency domain grayscale diagram. Furthermore, the data processing and image analysis algorithm based on artificial intelligence and deep learning is used to analyze the EEG signal frequency domain grayscale diagram, so as to capture the semantic features of the EEG signal frequency domain image, so as to carry out real-time monitoring of the emotional state of the operating object and abnormality detection. If abnormal emotions are detected, an alarm is immediately triggered to remind relevant personnel to pay attention and take corresponding measures. In this way, the emotional state or stress level of the individual can be automatically estimated by analyzing the EEG signals of the operating object, and then potential operational risks can be identified and warned, achieving more intelligent operational risk assessment and management, and providing more accurate and effective support for operational risk management in high-risk scenarios.

[0032] Specifically, in the technical solution of the present application, a risk assessment management system based on neural recognition technology is proposed. Figure 1 4 is a block diagram of a risk assessment management system based on neural recognition technology according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of risk assessment management system based on neural recognition technology according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to an embodiment of the present application, a risk assessment and management system 300 based on neural recognition technology includes: an EEG signal acquisition module 310, which is used to acquire the EEG signal of the operating object through an EEG device; an EEG signal preprocessing module 320, which is used to preprocess the EEG signal after acquiring the EEG signal of the operating object to obtain an EEG signal frequency domain grayscale image; an EEG frequency domain feature extraction module 330, which is used to extract EEG frequency domain features from the EEG signal frequency domain grayscale image to obtain EEG signal frequency domain image semantic coding features; an EEG signal frequency domain feature enhancement module 340, which is used to perform spatial modulation feature enhancement processing on the EEG signal frequency domain image semantic coding features to obtain EEG signal frequency domain image semantic enhancement coding features; an operation risk assessment management module 350, which is used to perform emotion recognition based on the EEG signal frequency domain image semantic enhancement coding features to determine whether there is any abnormality in the emotion of the operating object, and determine whether to generate an operation risk warning prompt.

[0033] In particular, the EEG signal acquisition module 310 is used to collect EEG signals of the operating object through an EEG device. It should be understood that EEG signals are formed by the weak current changes generated by the activity of brain neurons. They reflect the functional state of the brain, and changes in emotional state will be reflected in different frequency bands of EEG signals. By collecting and analyzing these signals in real time, the operator's psychological state, such as fatigue, stress, anxiety, etc., can be understood in a timely manner. Therefore, in the technical solution of the present application, it is expected that by collecting and analyzing the EEG signals of the operating object, real-time monitoring of the emotional state of the operating process and early warning of operational risks can be achieved.

[0034] In particular, the EEG signal preprocessing module 320 is used to obtain the EEG signal of the operating object and preprocess the EEG signal to obtain a grayscale image of the EEG signal frequency domain. Specifically, in a specific example of the present application, first, after obtaining the EEG signal of the operating object, the EEG signal is subjected to signal decomposition and filtering to obtain a set of filtered EEG signal components; it should be understood that the original EEG signal usually contains a large amount of noise, which may come from poor electrode contact, muscle activity, electromagnetic interference, etc. If not processed, these noises will interfere with the subsequent emotion recognition process and reduce the accuracy of the system. In addition, the EEG signal contains information in multiple frequency bands, and different frequency bands correspond to different brain functional areas and activity patterns. Therefore, by performing signal decomposition (such as wavelet transform, Fourier transform, etc.) and filtering on the EEG signal, the characteristics of each frequency band can be better analyzed while filtering out noise interference, thereby improving the accuracy of subsequent analysis. Next, the set of components of the filtered EEG signal is converted into a two-dimensional time-spectrogram, and the two-dimensional time-spectrogram is grayed to obtain the EEG signal frequency domain grayscale image. It should be understood that, since traditional frequency domain analysis can only provide frequency information, time-frequency domain analysis provides information in both time and frequency dimensions. This is particularly important for analyzing non-stationary signals (such as EEG signals) because the frequency components of EEG signals change over time. The time-spectrogram can intuitively display the energy distribution of the signal at different times and frequencies, which is convenient for observation and analysis. For example, a sudden increase in the energy of a certain frequency band in a certain time period can be observed through a two-dimensional time-spectrogram, which may be a manifestation of emotional fluctuations. In particular, graying can standardize the information in the two-dimensional time-spectrogram, which is convenient for subsequent feature extraction and operation of abnormal emotion detection tasks of the object. In addition, the data volume of the grayscale image is relatively small, which is conducive to reducing the consumption of computing resources.

[0035] In particular, the EEG frequency domain feature extraction module 330 is used to extract EEG frequency domain features from the EEG signal frequency domain grayscale image to obtain semantic coding features of the EEG signal frequency domain image. In a specific example of the present application, the EEG signal frequency domain grayscale image is input into an EEG frequency domain feature extractor based on the GoogLeNet network to obtain an EEG signal frequency domain image semantic coding feature map as the EEG signal frequency domain image semantic coding feature. That is, by inputting the EEG signal frequency domain grayscale image into an EEG frequency domain feature extractor based on the GoogLeNet network for feature mining, the frequency domain image semantic feature information in the EEG signal frequency domain grayscale image is extracted, thereby obtaining an EEG signal frequency domain image semantic coding feature map. It is worth mentioning that the EEG signal frequency domain image semantic coding feature map contains high-level EEG signal abstract features, which have stronger semantic meanings. For example, some features may represent specific emotional states (such as anxiety, relaxation, etc.).

[0036] In particular, the EEG signal frequency domain feature enhancement module 340 is used to perform spatial modulation feature enhancement processing on the semantic coding features of the EEG signal frequency domain image to obtain the semantic enhancement coding features of the EEG signal frequency domain image. It should be understood that in the frequency domain image of the EEG signal, the image semantic features of different regions are of different importance to the emotion recognition of the operating object. Some regions may contain more key information about the emotional state, while other regions may contain noise or irrelevant background information. However, in the case of complex backgrounds, traditional feature extraction methods may be difficult to distinguish the semantic features of the EEG signal frequency domain image in different spaces. Therefore, in the technical solution of the present application, the semantic coding features of the EEG signal frequency domain image are further subjected to spatial modulation feature enhancement processing to obtain the semantic enhancement coding features of the EEG signal frequency domain image. In particular, the spatial modulation feature enhancement processing method can extract the fine features of the semantics of the EEG signal frequency domain image, give different levels of attention to different regions in the feature map, help the model to more accurately locate and identify specific key information semantics, while suppressing unimportant areas, thereby improving the recognition accuracy of the operator's emotional state and abnormalities. That is to say, the spatial modulation feature enhancement processing can also reduce the influence of non-critical parts that are irrelevant to the subsequent emotion recognition and operation risk warning tasks to a certain extent, while highlighting the key EEG signal frequency domain image semantic coding features that are relatively stable even under adverse conditions, thereby enhancing the model's ability to cope with various complex situations and generalization ability, and providing a more accurate basis for the task of identifying abnormal emotions of the operating object. Specifically, in a specific example of the present application, such as Figure 3As shown, the EEG signal frequency domain feature enhancement module 340 includes: an image pixel granularity decoupling unit 341, which is used to perform feature fine-grained decoupling of the EEG signal frequency domain image semantic features to obtain a set of EEG signal frequency domain image semantic pixel granularity features; a pixel granularity semantic association calculation unit 342, which is used to calculate the EEG signal frequency domain image semantic spatial attenuation pixel granularity semantic association score between every two EEG signal frequency domain image semantic pixel granularity features in the set of EEG signal frequency domain image semantic pixel granularity features to obtain a set of EEG signal frequency domain image semantic spatial attenuation pixel granularity semantic association score representations; a spatial modulation enhancement unit 343, which is used to perform spatial modulation enhancement on the EEG signal frequency domain image semantic features based on the set of EEG signal frequency domain image semantic spatial attenuation pixel granularity semantic association score representations to obtain the EEG signal frequency domain image semantic enhancement coding features.

[0037] Specifically, the image pixel granularity decoupling unit 341 is used to perform feature fine-grained decoupling on the semantic features of the EEG signal frequency domain image to obtain a set of semantic pixel granularity features of the EEG signal frequency domain image. In an embodiment of the present application, the EEG signal frequency domain image semantic coding feature map is subjected to feature fine-grained spatial decoupling along the channel dimension to obtain a set of EEG signal frequency domain image semantic pixel granularity feature vectors as the set of EEG signal frequency domain image semantic pixel granularity features. That is, by performing feature fine-grained spatial decoupling along the channel dimension on the EEG signal frequency domain image semantic coding feature map, more refined EEG signal frequency domain image semantic coding feature information is obtained. This process is similar to decomposing an EEG signal frequency domain grayscale image into different frequency components, so that the fine features and local details of each local discharge point can be analyzed independently.

[0038] Specifically, the pixel granularity semantic association calculation unit 342 is used to calculate the semantic association score between the semantic spatial attenuation pixel granularity of the EEG signal frequency domain image between each two semantic pixel granularity features of the EEG signal frequency domain image in the set of semantic pixel granularity features of the EEG signal frequency domain image to obtain a set of semantic association scores represented by the semantic spatial attenuation pixel granularity of the EEG signal frequency domain image. In an embodiment of the present application, first, the spatial position of each EEG signal frequency domain image semantic pixel granularity feature vector in the set of semantic pixel granularity feature vectors of the EEG signal frequency domain image is annotated to obtain a set of semantic spatial position data of the EEG signal frequency domain image; it should be understood that in order to maintain the spatial relationship between the pixel granularity features of these local details of the EEG signal frequency domain image and the original EEG signal frequency domain image, it is necessary to add the exact position information of each pixel-level feature vector in the original image. This position annotation is very important for understanding the relative spatial relationship between the pixel granularity features of the EEG signal frequency domain image, because even if two pixels have similar features, if they are located in different positions, they may represent completely different meanings. Accordingly, maintaining good spatial structure information helps to accurately capture the subtle differences between the pixel granularity features of these EEG signal frequency domain images in subsequent processing, and provides a basis for the enhanced expression of key EEG signal frequency domain image pixel granularity features. Next, any two EEG signal frequency domain image semantic pixel granularity feature vectors in the set of EEG signal frequency domain image semantic pixel granularity feature vectors are input into the semantic association score measurement network to obtain a set of semantic association score vectors between EEG signal frequency domain image semantic pixel granularities; and based on the EEG signal frequency domain image semantic spatial position data of any two EEG signal frequency domain image semantic pixel granularity feature vectors in the set of EEG signal frequency domain image semantic spatial position data, the spatial span values ​​of the any two pixel granularity feature vectors are determined to obtain a set of EEG signal frequency domain image semantic spatial span values; here, by using a specially trained network to estimate the semantic connection strength between any two EEG signal frequency domain image pixel granularities, this network learns to identify which types of EEG signal frequency domain image local features tend to appear together in the same type of emotional state. At the same time, it is also necessary to consider the actual distance between the two atlas pixel granularity features in the feature map space, that is, to calculate their spatial span values. Combining the semantic association score with the spatial span value can help filter out the pixel granularity parts of the EEG signal frequency domain image that look similar but are actually not related, further improving the expressiveness of the feature map and its ability to recognize specific emotional states.Further, based on each EEG signal frequency domain image semantic space span value in the set of EEG signal frequency domain image semantic space span values, each EEG signal frequency domain image semantic pixel granularity semantic association score vector in the set of EEG signal frequency domain image semantic pixel granularity semantic association score vectors is subjected to semantic association attenuation adaptive attenuation modulation to obtain a set of EEG signal frequency domain image semantic space attenuation pixel granularity semantic association score vectors as a set of EEG signal frequency domain image semantic space attenuation pixel granularity semantic association score representations. That is, when two EEG signal frequency domain image pixel granularity features show strong semantic correlation and are close to each other, their importance will be maximized; on the contrary, if the two have high semantic similarity but are far apart, their influence will be appropriately weakened. Such dynamic adjustment ensures that the semantic association weight map between the EEG signal frequency domain image space attenuation pixel granularity finally generated can accurately reflect the most noteworthy part of the EEG signal frequency domain grayscale image, providing a basis for subsequent abnormal emotion recognition and operation risk warning of the operation object.

[0039] Among them, the specific process of inputting any two EEG signal frequency domain image semantic pixel granularity feature vectors in the set of EEG signal frequency domain image semantic pixel granularity feature vectors into the semantic association score measurement network includes: cascading any two EEG signal frequency domain image semantic pixel granularity feature vectors in the set of EEG signal frequency domain image semantic pixel granularity feature vectors to obtain the EEG signal frequency domain image semantic pixel granularity cascade feature vector; multiplying the EEG signal frequency domain image semantic pixel granularity cascade feature vector with the weight matrix, and then adding the bias vector by position to obtain the semantic association score vector between the EEG signal frequency domain image semantic pixel granularities.

[0040] More specifically, based on each EEG signal frequency domain image semantic space span value in the set of EEG signal frequency domain image semantic space span values, the specific process of performing semantic association attenuation adaptive attenuation modulation on each EEG signal frequency domain image semantic pixel granularity semantic association score vector in the set of EEG signal frequency domain image semantic pixel granularity semantic association score vector includes: dividing the EEG signal frequency domain image semantic space span value by a double semantic association attenuation hyperparameter to obtain an EEG signal frequency domain image semantic space span semantic association attenuation modulation factor; calculating a natural exponential function value with a natural constant e as the base and the negative of the EEG signal frequency domain image semantic space span semantic association attenuation modulation factor as the exponential power to obtain an EEG signal frequency domain image semantic space span class support modulation factor; calculating the positional division between each group of corresponding EEG signal frequency domain image semantic pixel granularity semantic association score vectors and the EEG signal frequency domain image semantic space span class support modulation factor to obtain an EEG signal frequency domain image semantic space attenuation pixel granularity semantic association score vector.

[0041] Specifically, the spatial modulation enhancement unit 343 is used to perform spatial modulation enhancement on the semantic features of the EEG signal frequency domain image based on the set of semantic association scores between pixel granularities of the semantic spatial attenuation of the EEG signal frequency domain image to obtain the semantic enhancement coding features of the EEG signal frequency domain image. In an embodiment of the present application, first, after the set of semantic association score vectors between pixel granularities of the semantic spatial attenuation of the EEG signal frequency domain image is arranged into a feature map of semantic association scores between pixel granularities of the semantic spatial attenuation of the EEG signal frequency domain image, the feature map of semantic association scores between pixel granularities of the semantic spatial attenuation of the EEG signal frequency domain image is input into a channel dimension modulation layer based on a point convolution layer and a weight distribution network based on a Sigmoid function to obtain a semantic association weight map between pixel granularities of the semantic spatial attenuation of the EEG signal frequency domain image; through the application of the Sigmoid function, it is ensured that all weights are within the range of 0 to 1, which is convenient for direct application to the original feature map. Specifically, through element-by-element multiplication operations, important pixel granularity feature information of the EEG signal frequency domain image is selectively amplified and insignificant or noisy pixel granularity areas are suppressed. This not only retains all the advantages of the original EEG signal frequency domain image semantic coding feature map, but also further highlights those EEG signal frequency domain image pixel granularity information fragments that are considered to be the most critical to the current operation object's emotion recognition task. Especially in complex EEG signals, this mechanism can effectively distinguish the feature areas that are closely related to the emotional state of the operation object, providing a basis for subsequent operation risk identification and early warning. Finally, the semantic association weight map between the attenuated pixel granularity of the EEG signal frequency domain image semantic space and the EEG signal frequency domain image semantic coding feature map are calculated by multiplying the position points to obtain the EEG signal frequency domain image semantic enhancement coding feature map as the EEG signal frequency domain image semantic enhancement coding feature.

[0042] In summary, in the above embodiment, the semantic coding features of the EEG signal frequency domain image are subjected to spatial modulation feature enhancement processing to obtain the semantic enhancement coding features of the EEG signal frequency domain image, including: performing spatial modulation feature enhancement processing on the semantic coding features of the EEG signal frequency domain image using the following feature attention enhancement formula to obtain the semantic enhancement coding features of the EEG signal frequency domain image; wherein the feature attention enhancement formula is:

[0043]

[0044]

[0045]

[0046]

[0047]

[0048]

[0049]

[0050]

[0051] A = [ S ']

[0052]

[0053]

[0054]

[0055]

[0056] in, is the semantic coding feature map of the EEG signal frequency domain image, It is the feature fine-grained spatial decoupling process. is each EEG signal frequency domain image semantic pixel granularity feature vector in the set of EEG signal frequency domain image semantic pixel granularity feature vectors, , is the number of channels of the semantic coding feature map of the EEG signal frequency domain image, is a real number, for Spatial location data, , is the spatial position coordinate represented by the spatial position data, is a collection of spatial location data, and are any two semantic pixel granularity feature vectors of the EEG signal frequency domain image in the set of semantic pixel granularity feature vectors of the EEG signal frequency domain image, respectively. For vector concatenation, is the weight matrix, is the bias vector, for and The semantic association score vector between the EEG signal frequency domain image semantic pixel granularity, is the set of semantic association score vectors between semantic pixel granularities of EEG signal frequency domain images, for Spatial location data, for Spatial location data, for and The space span value between is a set of spatial span values, is the semantic relevance decay hyperparameter, for and The semantic association score vector between the EEG signal frequency domain image semantic space attenuation pixel granularity, is the set of semantic association score vectors between attenuated pixel granularities in the semantic space of the EEG signal frequency domain image, [ ∙ ] is a feature graph arrangement of a vector set, is a semantic association score feature map of attenuated pixel granularity of the semantic space of the EEG signal frequency domain image obtained by arranging a set of semantic association score vectors of attenuated pixel granularity of the semantic space of the EEG signal frequency domain image, It is the point convolution channel modulation processing, The semantic association score feature map between pixel granularity of the semantic space attenuation of the EEG signal frequency domain image modulated by the channel dimension is is the Sigmoid function, is the semantic association weight map between attenuated pixel granularities in the semantic space of the EEG signal frequency domain image, For point-by-position multiplication, It is the semantic enhancement coding feature map of EEG signal frequency domain image.

[0057] In particular, the operational risk assessment management module 350 is used to perform emotion recognition based on the semantic enhancement coding features of the EEG signal frequency domain image to determine whether the emotions of the operating object are abnormal, and determine whether to generate an operational risk warning prompt. In a specific example of the present application, the semantic enhancement coding feature map of the EEG signal frequency domain image is input into the classifier-based emotion recognizer to obtain a recognition result, and the recognition result is used to indicate whether the emotions of the operating object are abnormal. That is, the spatial modulation of the EEG signal frequency domain image is used to enhance the semantic representation information for classification processing, so as to perform emotional recognition of the operating object. Then, in response to the recognition result that the emotions of the operating object are abnormal, an operational risk warning prompt is generated to remind relevant personnel to pay attention and take corresponding measures. In this way, the emotional state or the level of stress felt by the individual can be automatically estimated by analyzing the EEG signal of the operating object, and then the potential risk of operation can be identified and warned, so as to achieve more intelligent operational risk assessment management and provide more accurate and effective support for operational risk management in high-risk scenarios.

[0058] Preferably, the step of passing the EEG signal frequency domain image semantic enhancement coding feature map through a classifier-based emotion recognizer to obtain a recognition result comprises:

[0059] Determine the feature mean and feature variance of the EEG signal frequency domain image semantic enhancement coding feature map, and divide the feature mean by the feature variance to obtain the EEG signal frequency domain image semantic enhancement coding probability statistics value:

[0060]

[0061] in, represents the feature mean of the semantic enhancement coding feature map of the EEG signal frequency domain image, represents the feature variance of the semantic enhancement coding feature map of the EEG signal frequency domain image, represents the probability statistic value of semantic enhancement coding of the EEG signal frequency domain image;

[0062] The EEG signal frequency domain image semantic enhancement coding feature map is point-multiplied with the inverse of the maximum eigenvalue in the EEG signal frequency domain image semantic enhancement coding feature map to obtain the EEG signal frequency domain image semantic enhancement coding probability constraint map:

[0063]

[0064] in, is the semantic enhancement coding feature map of the EEG signal frequency domain image, represents the inverse of the maximum eigenvalue in the semantic enhancement coding feature map of the EEG signal frequency domain image, It means point multiplication by position. is the probability constraint graph of semantic enhancement coding of the EEG signal frequency domain image;

[0065] After performing point addition on the EEG signal frequency domain image semantic enhancement coding probability constraint graph and the EEG signal frequency domain image semantic enhancement coding probability statistical value, a logarithmic value with base 2 is calculated to obtain an EEG signal frequency domain image semantic enhancement coding information interaction graph:

[0066] in, Indicates adding by position point. represents the logarithm operation with base 2, is the semantic enhancement coding information interaction diagram of the EEG signal frequency domain image;

[0067] After performing point subtraction on the EEG signal frequency domain image semantic enhancement coding probability constraint map with the EEG signal frequency domain image semantic enhancement coding probability statistics, the inverse of each eigenvalue is calculated to obtain the EEG signal frequency domain image semantic enhancement coding sequence constraint map:

[0068]

[0069] in, It means to subtract by position point. is the semantic enhancement coding sequence constraint diagram of the EEG signal frequency domain image;

[0070] The EEG signal frequency domain image semantic enhancement coding information interaction graph is interpolated with the EEG signal frequency domain image semantic enhancement coding sequence constraint graph to obtain an optimized EEG signal frequency domain image semantic enhancement coding feature graph;

[0071] The optimized EEG signal frequency domain image semantic enhancement coding feature map is passed through a classifier-based emotion recognizer to obtain a recognition result.

[0072] That is, in the case where the EEG signal frequency domain image semantic enhancement coding feature map represents the EEG signal frequency domain semantic coding features of the EEG signal frequency domain grayscale map determined by frequency domain analysis of the EEG signal of the acquisition operation object, when performing feature attention enhancement based on spatial flexible modulation, the EEG signal frequency domain image semantic enhancement coding feature map will also have distribution diversity due to differences in spatial weight distribution. As a result, when the EEG signal frequency domain image semantic enhancement coding feature map is classified by a classifier, the accuracy of the classification result is affected.

[0073] Therefore, considering that when the weight matrix of the classifier acts on the EEG signal frequency domain image semantic enhancement coding feature vector after the EEG signal frequency domain image semantic enhancement coding feature map is expanded, the distribution diversity of the EEG signal frequency domain image semantic enhancement coding feature map causes the uncertainty of the weight stimulation parameters of the weight matrix of the classifier, resulting in the lack of posterior probability density inference of the EEG signal frequency domain image semantic enhancement coding feature map through the weight matrix, which will affect the accuracy of the classification result.

[0074] In this case, the probability statistical characteristics of the EEG signal frequency domain image semantic enhancement coding feature map are used to simulate the mesoscale interaction structure under probability constraints between the eigenvalue scale and the feature map scale of the EEG signal frequency domain image semantic enhancement coding feature map, so as to construct a bidirectional latent variable motif based on the mesoscale short sequence relative to the probability statistical value to perform mesoscale bidirectional migration, and perform posterior recovery based on the sequence constraint on the interactive information, thereby improving the convergence effect in the probability density domain and improving the accuracy of the recognition result obtained by the classifier-based emotion recognizer of the EEG signal frequency domain image semantic enhancement coding feature map. In this way, the emotional state of the operating object can be more accurately monitored in real time and the potential risk of operation can be detected. If abnormal emotions are detected, an alarm will be triggered immediately to remind relevant personnel to pay attention and take corresponding measures, thereby realizing more intelligent operation risk assessment and management.

[0075] As described above, the risk assessment management system 300 based on neural recognition technology according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a risk assessment management algorithm based on neural recognition technology. In one possible implementation, the risk assessment management system 300 based on neural recognition technology according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the risk assessment management system 300 based on neural recognition technology can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the risk assessment management system 300 based on neural recognition technology can also be one of the many hardware modules of the wireless terminal.

[0076] Alternatively, in another example, the risk assessment and management system 300 based on neural recognition technology and the wireless terminal may also be separate devices, and the risk assessment and management system 300 based on neural recognition technology may be connected to the wireless terminal via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0077] Furthermore, a risk assessment management method based on neural recognition technology is also provided.

[0078] Figure 4 FIG. 1 is a flow chart of a risk assessment management method based on neural recognition technology according to an embodiment of the present application. Figure 4 As shown, according to the risk assessment management method based on neural recognition technology of the embodiment of the present application, the steps include: S1, collecting the EEG signal of the operating object through an EEG device; S2, after obtaining the EEG signal of the operating object, preprocessing the EEG signal to obtain a grayscale image of the EEG signal in the frequency domain; S3, extracting EEG frequency domain features from the EEG signal frequency domain grayscale image to obtain semantic coding features of the EEG signal frequency domain image; S4, performing spatial modulation feature enhancement processing on the semantic coding features of the EEG signal frequency domain image to obtain semantic enhancement coding features of the EEG signal frequency domain image; S5, performing emotion recognition based on the semantic enhancement coding features of the EEG signal frequency domain image to determine whether there is any abnormality in the emotions of the operating object, and determine whether to generate an operation risk warning prompt.

[0079] In summary, the risk assessment management method based on neural recognition technology according to the embodiment of the present application is explained, which analyzes the grayscale image of the EEG signal frequency domain by using data processing and image analysis algorithms based on artificial intelligence and deep learning, so as to capture the semantic features of the EEG signal frequency domain image, so as to perform real-time monitoring of the emotional state of the operating object and abnormality detection. If abnormal emotions are detected, an alarm is immediately triggered to remind relevant personnel to pay attention and take corresponding measures. In this way, the emotional state or perceived stress level of the individual can be automatically estimated by analyzing the EEG signal of the operating object, and then potential operational risks can be identified and warned, realizing more intelligent operational risk assessment management, and providing more accurate and effective support for operational risk management in high-risk scenarios.

[0080] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A risk assessment management system based on neural recognition technology, characterized in that: include: An electroencephalogram (EEG) signal acquisition module is used to acquire the EEG signal of the subject through an EEG device; An EEG signal preprocessing module, used for obtaining the EEG signal of the operation object and preprocessing the EEG signal to obtain a grayscale image of the EEG signal in the frequency domain; An EEG frequency domain feature extraction module is used to extract EEG frequency domain features from the EEG signal frequency domain grayscale image to obtain EEG signal frequency domain image semantic coding features; An EEG signal frequency domain feature enhancement module is used to perform spatial modulation feature enhancement processing on the EEG signal frequency domain image semantic coding features to obtain EEG signal frequency domain image semantic enhancement coding features, wherein the EEG signal frequency domain feature enhancement module includes: an image pixel granularity decoupling unit, used to perform feature fine-grained decoupling on the EEG signal frequency domain image semantic features to obtain a set of EEG signal frequency domain image semantic pixel granularity features; a pixel granularity semantic association calculation unit, used to calculate the EEG signal frequency domain image semantic spatial attenuation pixel granularity semantic association score between every two EEG signal frequency domain image semantic pixel granularity features in the set of EEG signal frequency domain image semantic pixel granularity features to obtain a set of EEG signal frequency domain image semantic spatial attenuation pixel granularity semantic association score representations; a spatial modulation enhancement unit, used to perform spatial modulation enhancement on the EEG signal frequency domain image semantic features based on the set of EEG signal frequency domain image semantic spatial attenuation pixel granularity semantic association score representations to obtain the EEG signal frequency domain image semantic enhancement coding features; An operation risk assessment management module, used to perform emotion recognition based on the semantic enhancement coding features of the EEG signal frequency domain image to determine whether the emotion of the operation object is abnormal, and determine whether to generate an operation risk warning prompt; Wherein, the pixel granularity semantic association calculation unit includes: A spatial position labeling subunit is used to label the spatial position of each EEG signal frequency domain image semantic pixel granularity feature vector in the set of EEG signal frequency domain image semantic pixel granularity feature vectors to obtain a set of EEG signal frequency domain image semantic spatial position data; A semantic association score measurement subunit, used for inputting any two semantic pixel granularity feature vectors of the EEG signal frequency domain image from the set of semantic pixel granularity feature vectors of the EEG signal frequency domain image into a semantic association score measurement network to obtain a set of semantic association score vectors between semantic pixel granularities of the EEG signal frequency domain image; A spatial span value determination subunit, configured to determine the spatial span value of any two pixel granularity feature vectors in the set of EEG signal frequency domain image semantic spatial position data based on the EEG signal frequency domain image semantic pixel granularity feature vectors in the set of EEG signal frequency domain image semantic spatial position data to obtain a set of EEG signal frequency domain image semantic spatial span values; A semantic association attenuation adaptive attenuation modulation subunit is used to perform semantic association attenuation adaptive attenuation modulation on each semantic association score vector between semantic pixel granularities in the set of semantic association score vectors between semantic pixel granularities of the EEG signal frequency domain image based on each semantic spatial span value of the EEG signal frequency domain image in the set of semantic spatial span values ​​of the EEG signal frequency domain image, so as to obtain a set of semantic association score vectors between semantic spatial attenuation pixel granularities of the EEG signal frequency domain image as a set representing the semantic association score between semantic spatial attenuation pixel granularities of the EEG signal frequency domain image; Wherein, the semantic association score measurement subunit is used to: Cascading any two EEG signal frequency domain image semantic pixel granularity feature vectors in the set of EEG signal frequency domain image semantic pixel granularity feature vectors to obtain an EEG signal frequency domain image semantic pixel granularity cascade feature vector; After multiplying the cascade feature vector of semantic pixel granularity of the EEG signal frequency domain image by the weight matrix, the cascade feature vector is added to the bias vector by position to obtain a semantic association score vector between semantic pixel granularity of the EEG signal frequency domain image; Wherein, the semantic association attenuation adaptive attenuation modulation subunit is used to: Dividing the semantic space span value of the EEG signal frequency domain image by a doubled semantic association attenuation hyperparameter to obtain a semantic association attenuation modulation factor of the semantic space span of the EEG signal frequency domain image; Calculate the natural exponential function value with the natural constant e as the base and the negative number of the semantic association attenuation modulation factor of the semantic space span of the EEG signal frequency domain image as the exponential power to obtain the class support modulation factor of the semantic space span of the EEG signal frequency domain image; Calculate the positional division between the semantic association score vector between the semantic pixel granularity of the EEG signal frequency domain image and the semantic space span class support modulation factor of each group to obtain the semantic association score vector between the attenuation pixel granularity of the EEG signal frequency domain image semantic space; Wherein, the spatial modulation enhancement unit is used to: After arranging the set of semantic association score vectors between pixel granularities of semantic space attenuation of EEG signal frequency domain image into a semantic association score feature map between pixel granularities of semantic space attenuation of EEG signal frequency domain image, the feature map of semantic association score between pixel granularities of semantic space attenuation of EEG signal frequency domain image is input into a channel dimension modulation layer based on a point convolution layer and a weight allocation network based on a sigmoid function to obtain a semantic association weight map between pixel granularities of semantic space attenuation of EEG signal frequency domain image; The semantic association weight map between the attenuated pixel granularity of the semantic space of the EEG signal frequency domain image and the semantic coding feature map of the EEG signal frequency domain image are calculated by multiplying the map by position points to obtain the semantic enhancement coding feature map of the EEG signal frequency domain image as the semantic enhancement coding feature of the EEG signal frequency domain image.

2. The risk assessment management system based on neural recognition technology according to claim 1 is characterized in that: The EEG signal preprocessing module comprises: A signal decomposition and filtering processing unit, configured to obtain an EEG signal of the operation object, perform signal decomposition and filtering processing on the EEG signal to obtain a set of filtered EEG signal components; The signal frequency domain conversion and grayscale processing module is used to convert the set of filtered EEG signal components into a two-dimensional time-frequency spectrum diagram, and grayscale the two-dimensional time-frequency spectrum diagram to obtain the EEG signal frequency domain grayscale diagram.

3. The risk assessment management system based on neural recognition technology according to claim 2 is characterized in that: The EEG frequency domain feature extraction module is used to: input the EEG signal frequency domain grayscale image into an EEG frequency domain feature extractor based on the GoogLeNet network to obtain an EEG signal frequency domain image semantic coding feature image as the EEG signal frequency domain image semantic coding feature.

4. The risk assessment management system based on neural recognition technology according to claim 3 is characterized in that: The image pixel granularity decoupling unit is used to perform feature fine-grained spatial decoupling of the EEG signal frequency domain image semantic coding feature map along the channel dimension to obtain a set of EEG signal frequency domain image semantic pixel granularity feature vectors as a set of EEG signal frequency domain image semantic pixel granularity features.

5. The risk assessment management system based on neural recognition technology according to claim 4 is characterized in that: The operational risk assessment management module is used to: Inputting the semantic enhancement coding feature map of the EEG signal frequency domain image into a classifier-based emotion recognizer to obtain a recognition result, wherein the recognition result is used to indicate whether the emotion of the operating object is abnormal; In response to the recognition result that the emotion of the operation object is abnormal, an operation risk warning prompt is generated.

6. A risk assessment management method based on neural recognition technology based on the risk assessment management system based on neural recognition technology according to any one of claims 1 to 5, characterized in that: include: Collecting the EEG signal of the subject through an EEG device; After obtaining the EEG signal of the operation object, preprocessing the EEG signal to obtain a grayscale image of the EEG signal in the frequency domain; Extracting EEG frequency domain features from the EEG signal frequency domain grayscale image to obtain EEG signal frequency domain image semantic coding features; Performing spatial modulation feature enhancement processing on the semantic coding feature of the EEG signal frequency domain image to obtain semantic enhancement coding feature of the EEG signal frequency domain image; Emotion recognition is performed based on the semantic enhancement coding features of the EEG signal frequency domain image to determine whether the emotions of the operating object are abnormal, and to determine whether to generate an operation risk warning prompt.

Images