A campus anti-bullying method, system, device and medium

Abstract

The present application discloses a method, system, device and medium for preventing bullying in schools, which relates to the technical field of behavior analysis. The method comprises: collecting first scanning data of a target area, and judging whether the number of people gathered in the target area is greater than a preset value based on the first scanning data. If so, collecting voice data of the target area, and synchronously collecting second scanning data of the target area; judging whether there are preset bullying words in the voice data, and if so, performing voice analysis on the voice data to obtain corresponding voice analysis results; the voice analysis results include bullying text, semantic information and voice category; performing behavior recognition on the second scanning data to obtain corresponding abnormal behavior; judging whether bullying occurs based on the voice analysis results and the abnormal behavior, and if so, generating early warning information based on the bullying text and the abnormal behavior. The present application has the effect of solving the problem of privacy infringement.

Description

Technical Field

[0001] The present application relates to the technical field of behavior analysis, and in particular to a method, system, device and medium for preventing bullying in schools. Background Art

[0002] School bullying mostly occurs among primary and secondary school students and has various manifestations, such as physical violence, verbal bullying, etc. In order to curb school violence and prevent students from being physically and psychologically harmed, researchers have developed wearable smart watches and embedded mobile phone programs to provide students with convenient information access, health monitoring, personal navigation, communication and other functions. However, when using the above-mentioned smart watches and mobile phone programs, the access and acquisition of relevant information may infringe on students' privacy.

[0003] In addition, some schools deploy high-definition cameras in key areas of the campus (such as teaching buildings, office complexes, laboratory buildings, playgrounds, canteens, dormitory corridors, etc.), and conduct behavioral analysis on the collected videos to determine whether there is campus violence, but this method may still infringe on students' privacy. The above campus anti-bullying methods cannot be used in private places on campus, such as toilets, dormitories and bathhouses, and the places of use are limited.

[0004] In addition, there is a bullying monitoring method that combines action and voice. This method detects campus bullying incidents by simultaneously extracting action features and voice features and performing fusion and dimensionality reduction, and then training a BPNN model. This method is real-time and diverse, and supports multiple dialects and languages. It is suitable for students from different cultural backgrounds, but it has a high algorithm complexity, which may lead to slower detection speed.

[0005] With regard to the current methods of preventing bullying on campus, how to solve the problem of infringement of students' privacy is a major problem that needs to be solved. Summary of the invention

[0006] In order to solve the problem of privacy infringement, the present application provides a campus anti-bullying method, system, device and medium.

[0007] In the first aspect, the present application provides a method for preventing bullying in schools, which adopts the following technical solutions:

[0008] Collecting first scanning data of a target area, and judging whether the number of people gathered in the target area is greater than a preset value according to the first scanning data, and if so, collecting voice data of the target area, and synchronously collecting second scanning data of the target area;

[0009] Determine whether there is a preset bullying word in the voice data, and if so, perform voice analysis on the voice data to obtain a corresponding bullying text;

[0010] Performing behavior recognition on the second scanning data to obtain corresponding abnormal behavior;

[0011] Generate early warning information based on the bullying text and the abnormal behavior.

[0012] By adopting the above technical solution, the first scanning data of the target area is collected, and it is judged whether the number of people gathered in the target area is greater than the preset value based on the first scanning data. If the number of people gathered in the target area is the preset value, the voice data of the target area is collected, and the second scanning data of the target area is collected synchronously, and then it is judged whether there are preset bullying words in the voice data. If there are preset bullying words in the voice data, the voice data is subjected to voice analysis to obtain the corresponding voice analysis results, which include bullying text, semantic information and voice category. Then, the second scanning data is subjected to behavior recognition to obtain the corresponding abnormal behavior. Finally, it is judged whether bullying occurs based on the voice analysis results and abnormal behavior. If bullying occurs, an early warning message is generated based on the bullying text and abnormal behavior. In the above method, radar scanning technology and voice analysis technology are applied to the field of anti-bullying. Compared with the method in the background technology, the problem of infringing on students' privacy is solved, and the technical complexity is low, making this method easier to promote and apply.

[0013] Optionally, the step of determining whether there is a preset bullying word in the voice data, and if so, performing voice analysis on the voice data to obtain a corresponding voice analysis result includes:

[0014] Performing text recognition on the voice data to obtain corresponding voice text;

[0015] Perform semantic recognition according to the speech text and a preset BERT model to obtain corresponding semantic information;

[0016] Performing voice category recognition on the voice data to obtain a corresponding voice category;

[0017] Determine whether there is a preset bullying word in the voice text, and if so, extract the bullying text from the voice text according to the voice text and the voice category;

[0018] A speech analysis result is generated based on the bullying text, semantic information, and speech category.

[0019] By adopting the above technical solution, in order to generate speech analysis results, text recognition is performed on the speech data to obtain the corresponding speech text, and then semantic recognition is performed based on the speech text and the preset BERT model to obtain the corresponding semantic information, and then speech category recognition is performed on the speech data to obtain the corresponding speech category, and then it is determined whether there are preset bullying words in the speech text. If there are preset bullying words in the speech text, the bullying text is extracted from the speech text according to the speech text and the speech category, and finally the speech analysis results are generated based on the bullying text, semantic information and speech category.

[0020] Optionally, the step of performing text recognition on the voice data to obtain corresponding voice text includes:

[0021] Performing a first preprocessing operation on the voice data to obtain first voice information;

[0022] Performing a first feature extraction on the first voice information to obtain first feature data;

[0023] The speech text is generated according to the first feature data, a preset acoustic model, a preset language model and a preset dictionary.

[0024] By adopting the above technical solution, in order to obtain the speech text corresponding to the speech data, the speech data is first preprocessed to obtain the first speech information, and then the first feature extraction is performed on the first speech information to obtain the first feature data, and finally the speech text is generated according to the first feature data, the preset acoustic model, the preset language model and the preset dictionary.

[0025] Optionally, the step of performing voice category recognition on the voice data to obtain a corresponding voice category includes:

[0026] Performing a second preprocessing operation on the voice data to obtain second voice information;

[0027] performing a second feature extraction on the second voice information to obtain second feature data;

[0028] Training a preset first neural network model according to a preset audio data set to obtain a trained neural network model;

[0029] The second feature data is input into the trained neural network model to obtain a corresponding speech category.

[0030] By adopting the above technical solution, in order to obtain the corresponding speech category in the speech data, the speech data is first subjected to a second preprocessing operation to obtain second speech information, and then a second feature extraction is performed based on the second speech information to obtain second feature data, and then the preset first neural network model is trained according to the preset audio data set to obtain a trained neural network model, and finally the second feature data is input into the trained neural network model to obtain the corresponding speech category.

[0031] Optionally, the step of performing behavior recognition on the second scanning data to obtain corresponding abnormal behavior includes: the second scanning data includes an electromagnetic wave signal emitted to the target area by a radar and a received echo signal;

[0032] Acquiring the electromagnetic wave signal and the echo signal, and performing mixing processing on the electromagnetic wave signal and the echo signal to obtain a discrete intermediate frequency signal;

[0033] Performing Fourier transform on the discrete intermediate frequency signal to obtain a Doppler heat map;

[0034] Performing feature extraction on the Doppler heat map through a preset second neural network model to obtain a corresponding feature vector;

[0035] The feature vector is input into a preset Transformer model to obtain the corresponding abnormal behavior.

[0036] By adopting the above technical solution, in order to obtain the corresponding abnormal behavior, the electromagnetic wave signal and the echo signal are first obtained, and the electromagnetic wave signal and the echo signal are mixed to obtain a discrete intermediate frequency signal, and then the discrete intermediate frequency signal is Fourier transformed to obtain a Doppler heat map, and then the Doppler heat map is feature extracted through a preset second neural network model to obtain a corresponding feature vector, and finally the feature vector is input into a preset Transformer model to obtain the corresponding abnormal behavior.

[0037] Optionally, the step of generating warning information according to the bullying text and the abnormal behavior includes:

[0038] Determining a comprehensive bullying degree based on the bullying texts and the abnormal behaviors;

[0039] Acquire first time information corresponding to the bullying text and second time information corresponding to the abnormal behavior, and determine the bullying frequency according to the first time information and the second time information;

[0040] A warning category is determined based on the comprehensive bullying level and the bullying frequency, and warning information is generated based on the warning category.

[0041] By adopting the above technical solution, in order to generate targeted warning information based on specific bullying texts and abnormal behaviors, the comprehensive bullying level is first determined based on the bullying texts and abnormal behaviors, and then the first time information corresponding to the bullying texts and the second time information corresponding to the abnormal behaviors are obtained, and the bullying frequency is determined based on the first time information and the second time information. Finally, the warning category is determined based on the comprehensive bullying level and the bullying frequency, and warning information is generated based on the warning category.

[0042] Optionally, the step of determining the comprehensive bullying degree according to the bullying text and the abnormal behavior includes:

[0043] Grading the bullying degree of the bullying text to obtain a voice bullying degree;

[0044] Grading the bullying degree of the abnormal behavior to obtain the behavior bullying degree;

[0045] A comprehensive bullying level is determined according to the verbal bullying level and the behavioral bullying level.

[0046] By adopting the above technical solution, in order to determine the comprehensive bullying level, the bullying level of the bullying text is first graded to obtain the voice bullying level, and then the bullying level of the abnormal behavior is graded to obtain the behavioral bullying level. Finally, the comprehensive bullying level is determined based on the voice bullying level and the behavioral bullying level.

[0047] In the second aspect, the present application also provides a school anti-bullying system, which adopts the following technical solution:

[0048] A school anti-bullying system, comprising:

[0049] A gathering number judgment module is used to collect first scanning data of a target area, and judge whether the gathering number of the target area is greater than a preset value according to the first scanning data. If so, collect voice data of the target area, and simultaneously collect second scanning data of the target area;

[0050] A speech analysis module, used to determine whether there are preset bullying words in the speech data, and if so, perform speech analysis on the speech data to obtain corresponding speech analysis results; the speech analysis results include bullying text, semantic information and speech category;

[0051] A behavior recognition module, used for performing behavior recognition on the second scanning data to obtain corresponding abnormal behavior;

[0052] An early warning module is used to determine whether bullying occurs based on the voice analysis result and the abnormal behavior, and if so, generate early warning information based on the bullying text and the abnormal behavior.

[0053] In a third aspect, the present application also provides a computer device, which adopts the following technical solution:

[0054] A computer device comprises a memory and a processor, wherein the memory stores a computer program executable on the processor, and the processor implements the method described in the first aspect when executing the computer program.

[0055] In a fourth aspect, the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0056] A computer-readable storage medium stores a computer program that can be loaded by a processor and execute the method described in the first aspect.

[0057] In summary, the present application includes at least the following beneficial technical effects: collecting the first scanning data of the target area, and judging whether the number of people gathered in the target area is greater than the preset value based on the first scanning data; if the number of people gathered in the target area is the preset value, then collecting the voice data of the target area, and synchronously collecting the second scanning data of the target area, and then judging whether there are preset bullying words in the voice data; if there are preset bullying words in the voice data, then performing voice analysis on the voice data to obtain corresponding voice analysis results, which include bullying text, semantic information and voice categories; then performing behavior recognition on the second scanning data to obtain corresponding abnormal behaviors; finally, judging whether bullying occurs based on the voice analysis results and abnormal behaviors; if bullying occurs, generating early warning information based on the bullying text and abnormal behaviors. In the above method, radar scanning technology and voice analysis technology are applied to the field of anti-bullying, which solves the problem of infringing students' privacy compared to the method in the background technology, and the technical complexity is relatively low, making the present method easier to promote and apply. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the specific implementation of the present application or the technical solution in the prior art, the following is a brief introduction to the drawings required for the specific implementation or the prior art description. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0059] Figure 1 It is a schematic diagram of the overall process of an embodiment of the present application.

[0060] Figure 2 It is a schematic diagram of the structure of the system of this application.

[0061] Figure 3 It is a structural block diagram of the computer device of the present application. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solutions and advantages of this application more clear, the following Figure 1-3 It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0063] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0064] The embodiment of the present application discloses a method for preventing bullying in schools.

[0065] Reference Figure 1 , a method to prevent bullying in schools, including:

[0066] Step S11, collecting the first scanning data of the target area, and judging whether the number of people gathered in the target area is greater than a preset value according to the first scanning data, if so, collecting the voice data of the target area, and synchronously collecting the second scanning data of the target area.

[0067] Specifically, the first scanning data of the target area is collected, and it is determined whether the number of people gathered in the target area is greater than a preset value based on the first scanning data. If the number of people gathered in the target area is greater than the preset value, the voice data of the target area is collected, and the second scanning data of the target area is collected synchronously; if the number of people gathered in the target area is not greater than the preset value, a keep quiet message is generated.

[0068] It should be noted that the first scanning data and the second scanning data are both data obtained by scanning the target area by the radar, and there is a time sequence relationship between the first scanning data and the second scanning data, that is, the acquisition time of the first scanning data is earlier than the acquisition time of the second scanning data. In addition, in this embodiment, the preset value is 1.

[0069] Step S12, determining whether there are preset bullying words in the voice data, and if so, performing voice analysis on the voice data to obtain corresponding voice analysis results.

[0070] Among them, the speech analysis results include bullying text, semantic information and speech categories.

[0071] Specifically, it is determined whether there are preset bullying words in the voice data. If there are preset bullying words in the voice data, a voice analysis is performed on the voice data to obtain a corresponding voice analysis result.

[0072] It is understandable that the preset bullying words are pre-set sensitive words or sensitive sentences related to bullying, such as "help", "beat you to death", "what's wrong with beating you", etc.

[0073] It should be noted that step S12 performs speech recognition on the speech data through the speech recognition (ASR) technology and the speech classification (CLS) technology. The speech recognition (ASR) technology can perform real-time recognition during the speech input process without waiting for the complete speech input, thereby improving the real-time and interactivity of the event. The speech classification (CLS) technology uses a neural network to perform emotion classification and outputs the probability corresponding to the emotion polarity in the speech, that is, outputs the emotion in the speech.

[0074] Step S13, performing behavior recognition on the second scanning data to obtain corresponding abnormal behavior.

[0075] Step S14, judging whether bullying occurs based on the speech analysis results and abnormal behavior, and if so, generating warning information based on the bullying text and abnormal behavior.

[0076] Specifically, whether bullying has occurred is determined based on the voice analysis results and abnormal behavior. If bullying has occurred, a warning message is generated based on the bullying text and abnormal behavior.

[0077] In the above implementation, the first scanning data of the target area is collected, and it is determined whether the number of people gathered in the target area is greater than a preset value based on the first scanning data. If the number of people gathered in the target area is the preset value, the voice data of the target area is collected, and the second scanning data of the target area is collected synchronously, and then it is determined whether there are preset bullying words in the voice data. If there are preset bullying words in the voice data, the voice data is subjected to voice analysis to obtain corresponding voice analysis results, which include bullying text, semantic information and voice category. The second scanning data is then subjected to behavior recognition to obtain corresponding abnormal behavior. Finally, it is determined whether bullying occurs based on the voice analysis results and abnormal behavior. If bullying occurs, a warning message is generated based on the bullying text and abnormal behavior. In the above method, radar scanning technology and voice analysis technology are applied to the field of anti-bullying. Compared with the method in the background technology, the problem of infringing on students' privacy is solved, and the technical complexity is relatively low, making this method easier to promote and apply.

[0078] As a further implementation method of the campus anti-bullying method, it is determined whether there are preset bullying words in the voice data. If so, a voice analysis is performed on the voice data to obtain a corresponding voice analysis result. The steps of performing a voice analysis on the voice data to obtain a corresponding voice analysis result include:

[0079] Step S21, performing text recognition on the voice data to obtain corresponding voice text.

[0080] Step S22, performing semantic recognition based on the speech text and the preset BERT model to obtain corresponding semantic information.

[0081] Step S23, performing voice category recognition on the voice data to obtain a corresponding voice category.

[0082] Step S24, determining whether there is a preset bullying word in the voice text, and if so, extracting the bullying text from the voice text according to the voice text and the voice category.

[0083] It should be noted that, in this embodiment, the execution order of step S23 and the execution order of step S24 are in parallel.

[0084] Step S24, generating a speech analysis result according to the bullying text, semantic information and speech category.

[0085] In the above implementation, in order to generate a speech analysis result, text recognition is performed on the speech data to obtain the corresponding speech text, and then semantic recognition is performed based on the speech text and the preset BERT model to obtain the corresponding semantic information. Then, speech category recognition is performed on the speech data to obtain the corresponding speech category. Then, it is determined whether there are preset bullying words in the speech text. If there are preset bullying words in the speech text, the bullying text is extracted from the speech text based on the speech text and the speech category. Finally, a speech analysis result is generated based on the bullying text, semantic information and speech category.

[0086] As a further implementation method of the campus anti-bullying method, the step of performing text recognition on the voice data to obtain the corresponding voice text includes:

[0087] Step S31, performing a first preprocessing operation on the voice data to obtain first voice information.

[0088] It should be noted that, in this embodiment, the first preprocessing operation includes noise reduction, and the noise reduction operation can reduce interference with subsequent operations.

[0089] Step S32: extract the first feature of the first voice information to obtain the first feature data.

[0090] It should be noted that, in this embodiment, the first feature data is the Mel frequency cepstral coefficient, that is, the MFCC feature. The first speech data is sampled in the form of a data stream to obtain the sampled data, and then the sampled data is subjected to fast Fourier transform FFT, Melton band transform, discrete cosine transform DCT and regularization to obtain the first feature data.

[0091] Step S33, generating speech text according to the first feature data, a preset acoustic model, a preset language model and a preset dictionary.

[0092] In the above embodiment, in order to obtain the speech text corresponding to the speech data, the speech data is first preprocessed to obtain the first speech information, and then the first feature extraction is performed on the first speech information to obtain the first feature data, and finally the speech text is generated according to the first feature data, the preset acoustic model, the preset language model and the preset dictionary.

[0093] As a further implementation method of the campus anti-bullying method, the step of performing voice category recognition on the voice data to obtain the corresponding voice category includes:

[0094] Step S41, performing a second preprocessing operation on the voice data to obtain second voice information.

[0095] It should be noted that, in step S41, the first preprocessing operation includes sampling rate unification and normalization processing. Specifically, the sampling rate of the voice data is adjusted to the same standard, and then the adjusted voice data is normalized to obtain the second voice data.

[0096] Step S42: extract the second feature based on the second voice information to obtain second feature data.

[0097] It should be noted that, in this embodiment, the second feature data is Mel-frequency cepstral coefficients, that is, MFCC features.

[0098] Step S43: training the preset first neural network model according to the preset audio data set to obtain a trained neural network model.

[0099] It should be noted that the preset first neural network model can be a recurrent neural network RNN ​​or a convolutional neural network CNN.

[0100] Step S44: input the second feature data into the trained neural network model to obtain the corresponding speech category.

[0101] It should be noted that the speech categories include crying, laughing, clapping, dog barking, cat meowing and other sound categories; the trained neural network model performs emotion classification and timbre classification on the speech according to the second feature data to obtain the speech category.

[0102] In the above implementation, in order to obtain the corresponding speech category in the speech data, a second preprocessing operation is first performed on the speech data to obtain second speech information, and then a second feature extraction is performed based on the second speech information to obtain second feature data. Then, the preset first neural network model is trained according to the preset audio data set to obtain a trained neural network model, and finally, the second feature data is input into the trained neural network model to obtain the corresponding speech category.

[0103] As a further implementation method of the campus anti-bullying method, the step of performing behavior recognition on the second scanning data to obtain the corresponding abnormal behavior includes: the second scanning data includes an electromagnetic wave signal emitted to the target area by a radar and a received echo signal;

[0104] Step S51, acquiring an electromagnetic wave signal and an echo signal, and performing mixing processing on the electromagnetic wave signal and the echo signal to obtain a discrete intermediate frequency signal.

[0105] Step S52, performing Fourier transform on the discrete intermediate frequency signal to obtain a Doppler heat map.

[0106] It should be noted that the Doppler heat map is a method for graphically representing the distance characteristics and speed characteristics of a target (a student in this application). The Doppler heat map can intuitively reflect the information of the target's spatial state and motion state by displaying the target's distance and Doppler frequency shift on a two-dimensional plane, thereby intuitively reflecting the target's spatial position and motion state.

[0107] Step S53, extracting features from the Doppler heat map using a preset second neural network model to obtain a corresponding feature vector.

[0108] It should be noted that, in this embodiment, the second neural network model is a convolutional neural network CNN, which extracts spatial features from the Doppler heat map through the convolutional layer of the convolutional neural network CNN, retains important features through the pooling layer of the convolutional neural network CNN, and extracts structural features from the Doppler heat map through the fully connected layer of the convolutional neural network CNN, and integrates the spatial features, retained important features and structural features to obtain a feature vector.

[0109] Step S54, inputting the feature vector into a preset Transformer model to obtain the corresponding abnormal behavior.

[0110] It can be understood that the correlation of each position of the feature sequence in the input feature vector is calculated through the self-attention mechanism based on the Transformer model, thereby capturing the global dependency; the entire Transformer model is composed of an encoder and a decoder. The encoder is responsible for encoding the feature sequence and generating a high-level feature representation, while the decoder processes the feature representation to output the result, that is, the abnormal behavior.

[0111] In the above embodiment, in order to obtain the corresponding abnormal behavior, the electromagnetic wave signal and the echo signal are first obtained, and the electromagnetic wave signal and the echo signal are mixed to obtain a discrete intermediate frequency signal, and then the discrete intermediate frequency signal is Fourier transformed to obtain a Doppler heat map, and then the Doppler heat map is feature extracted through a preset second neural network model to obtain a corresponding feature vector, and finally the feature vector is input into a preset Transformer model to obtain the corresponding abnormal behavior.

[0112] As a further implementation method of the campus anti-bullying method, the step of generating early warning information according to bullying texts and abnormal behaviors includes:

[0113] Step S61, determining the comprehensive bullying degree based on the bullying text and abnormal behavior.

[0114] Step S62, obtaining first time information corresponding to the bullying text and second time information corresponding to the abnormal behavior, and determining the bullying frequency according to the first time information and the second time information.

[0115] Step S63, determining a warning category according to the comprehensive bullying degree and bullying frequency, and generating warning information according to the warning category.

[0116] In the above implementation, in order to generate targeted warning information based on specific bullying texts and abnormal behaviors, the comprehensive bullying level is first determined based on the bullying texts and abnormal behaviors, and then the first time information corresponding to the bullying texts and the second time information corresponding to the abnormal behaviors are obtained, and the bullying frequency is determined based on the first time information and the second time information. Finally, the warning category is determined based on the comprehensive bullying level and the bullying frequency, and warning information is generated based on the warning category.

[0117] As a further implementation method of the school bullying prevention method, the steps of determining the comprehensive bullying level based on the bullying texts and abnormal behaviors include:

[0118] Step S71, grading the bullying degree of the bullying text to obtain the voice bullying degree.

[0119] It is understandable that when bullying texts are graded for bullying severity, they are mainly graded based on the insult, threat, and sensitivity of the words or sentences in the bullying texts. For example, if the bullying text is "Help me", the voice bullying severity can be level one; if the bullying text is "Fight", "What if I beat you?" or "Let me go", the voice bullying severity can be level two; if the bullying text is "Don't tell the teacher", the voice bullying severity can be level three; if the bullying text is "Borrow some money", "Robbery" or "Take out the money", the voice bullying severity can be level three.

[0120] Step S72, grading the bullying degree of the abnormal behavior to obtain the bullying degree of the behavior.

[0121] Specifically, the bullying degree of abnormal behaviors is graded according to a preset behavior data set to obtain the behavior bullying degree.

[0122] It can be understood that in step S72, if the abnormal behavior is "beating", "kicking" or "slapping", the level of behavioral bullying can be level one; if the abnormal behavior is "pushing", the level of behavioral bullying can be level two; if the abnormal behavior is "pulling" or "snatching", the abnormal behavior can be level three.

[0123] Step S73, determining the comprehensive bullying degree according to the verbal bullying degree and the behavioral bullying degree.

[0124] In the above implementation, in order to determine the comprehensive bullying level, the bullying text is first graded to obtain the voice bullying level, and then the abnormal behavior is graded to obtain the behavioral bullying level. Finally, the comprehensive bullying level is determined based on the voice bullying level and the behavioral bullying level.

[0125] The embodiment of the present application also discloses a campus anti-bullying system.

[0126] refer to Figure 2 , a school anti-bullying system, comprising:

[0127] A gathering number judgment module is used to collect the first scanning data of the target area, and judge whether the gathering number of the target area is greater than a preset value according to the first scanning data. If so, the voice data of the target area is collected, and the second scanning data of the target area is collected synchronously;

[0128] The speech analysis module is used to determine whether there are preset bullying words in the speech data. If so, the speech data is subjected to speech analysis to obtain corresponding speech analysis results; the speech analysis results include bullying text, semantic information and speech category;

[0129] A behavior recognition module, used to perform behavior recognition on the second scanning data to obtain corresponding abnormal behavior;

[0130] The early warning module is used to determine whether bullying has occurred based on the voice analysis results and abnormal behavior. If so, it generates early warning information based on the bullying text and abnormal behavior.

[0131] The campus anti-bullying method of the present invention can implement any method in the campus anti-bullying system, and the specific working process of the campus anti-bullying method of the present invention can refer to the corresponding process in the above-mentioned campus anti-bullying system.

[0132] The embodiment of the present application also discloses a computer device.

[0133] refer to Figure 3 A computer device includes a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, any of the above-mentioned campus anti-bullying methods is implemented.

[0134] The embodiment of the present application also discloses a computer-readable storage medium.

[0135] A computer-readable storage medium stores a computer program that can be loaded by a processor and execute any of the above-mentioned campus anti-bullying methods.

[0136] Among them, computer-readable storage media can be any tangible medium that contains or stores a program, which can be used by or in conjunction with an instruction execution system, apparatus or device; the program code contained on the computer-readable medium can be transmitted using any appropriate medium, including but not limited to wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0137] The above are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Any feature disclosed in this specification (including the abstract and drawings), unless otherwise stated, can be replaced by other equivalent or alternative features with similar purposes. That is, unless otherwise stated, each feature is only an example of a series of equivalent or similar features.

Claims

1. A method for preventing bullying in schools, characterized in that: include: Collecting first scanning data of a target area, and judging whether the number of people gathered in the target area is greater than a preset value according to the first scanning data, and if so, collecting voice data of the target area, and synchronously collecting second scanning data of the target area; the first scanning data and the second scanning data are both data obtained by scanning the target area by radar, and the collection time of the first scanning data is earlier than the collection time of the second scanning data; Determine whether there is a preset bullying word in the voice data, and if so, perform voice analysis on the voice data to obtain a corresponding voice analysis result; the voice analysis result includes bullying text, semantic information and voice category; the voice category includes an emotion category; Performing behavior recognition on the second scanning data to obtain corresponding abnormal behavior; Determining whether bullying occurs according to the voice analysis result and the abnormal behavior, and if so, generating warning information according to the bullying text and the abnormal behavior; The step of performing behavior recognition on the second scanning data to obtain the corresponding abnormal behavior includes: the second scanning data includes an electromagnetic wave signal emitted by the radar to the target area and a received echo signal; Acquiring the electromagnetic wave signal and the echo signal, and performing mixing processing on the electromagnetic wave signal and the echo signal to obtain a discrete intermediate frequency signal; Performing Fourier transform on the discrete intermediate frequency signal to obtain a Doppler heat map; Performing feature extraction on the Doppler heat map by using a preset second neural network model to obtain a corresponding feature vector; The feature vector is input into a preset Transformer model to obtain the corresponding abnormal behavior.

2. A campus anti-bullying method according to claim 1, characterized in that: The step of determining whether there is a preset bullying word in the voice data, and if so, performing voice analysis on the voice data to obtain a corresponding voice analysis result, comprises: Performing text recognition on the voice data to obtain corresponding voice text; Perform semantic recognition according to the speech text and a preset BERT model to obtain corresponding semantic information; Performing voice category recognition on the voice data to obtain a corresponding voice category; Determine whether there is a preset bullying word in the voice text, and if so, extract the bullying text from the voice text according to the voice text and the voice category; A speech analysis result is generated based on the bullying text, semantic information, and speech category.

3. A campus anti-bullying method according to claim 2, characterized in that: The step of performing text recognition on the voice data to obtain corresponding voice text includes: Performing a first preprocessing operation on the voice data to obtain first voice information; Performing a first feature extraction on the first voice information to obtain first feature data; The speech text is generated according to the first feature data, a preset acoustic model, a preset language model and a preset dictionary.

4. A campus anti-bullying method according to claim 3, characterized in that: The step of performing voice category recognition on the voice data to obtain the corresponding voice category includes: Performing a second preprocessing operation on the voice data to obtain second voice information; performing a second feature extraction on the second voice information to obtain second feature data; Training a preset first neural network model according to a preset audio data set to obtain a trained neural network model; The second feature data is input into the trained neural network model to obtain a corresponding speech category.

5. A campus anti-bullying method according to claim 1, characterized in that: The step of generating early warning information according to the bullying text and the abnormal behavior includes: Determining a comprehensive bullying degree based on the bullying texts and the abnormal behaviors; Acquire first time information corresponding to the bullying text and second time information corresponding to the abnormal behavior, and determine the bullying frequency according to the first time information and the second time information; A warning category is determined based on the comprehensive bullying level and the bullying frequency, and warning information is generated based on the warning category.

6. A campus anti-bullying method according to claim 1, characterized in that: The step of determining the comprehensive bullying degree according to the bullying text and the abnormal behavior comprises: Grading the bullying degree of the bullying text to obtain a voice bullying degree; Grading the bullying degree of the abnormal behavior to obtain the behavior bullying degree; A comprehensive bullying level is determined according to the verbal bullying level and the behavioral bullying level.

7. A campus anti-bullying system, characterized in that: include: A gathering number judgment module is used to collect first scanning data of a target area, and judge whether the gathering number of the target area is greater than a preset value according to the first scanning data. If so, collect voice data of the target area, and simultaneously collect second scanning data of the target area; the first scanning data and the second scanning data are both data obtained by scanning the target area by radar, and the collection time of the first scanning data is earlier than the collection time of the second scanning data; A speech analysis module, used to determine whether there are preset bullying words in the speech data, and if so, perform speech analysis on the speech data to obtain corresponding speech analysis results; the speech analysis results include bullying text, semantic information and speech categories; the speech categories include emotional categories; A behavior recognition module, used for performing behavior recognition on the second scanning data to obtain corresponding abnormal behavior; An early warning module, configured to determine whether bullying occurs based on the speech analysis result and the abnormal behavior, and if so, generate early warning information based on the bullying text and the abnormal behavior; The step of performing behavior recognition on the second scanning data to obtain the corresponding abnormal behavior includes: the second scanning data includes an electromagnetic wave signal emitted by the radar to the target area and a received echo signal; Acquiring the electromagnetic wave signal and the echo signal, and performing mixing processing on the electromagnetic wave signal and the echo signal to obtain a discrete intermediate frequency signal; Performing Fourier transform on the discrete intermediate frequency signal to obtain a Doppler heat map; Performing feature extraction on the Doppler heat map by using a preset second neural network model to obtain a corresponding feature vector; The feature vector is input into a preset Transformer model to obtain the corresponding abnormal behavior.

8. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program executable on the processor, and the processor implements the method according to any one of claims 1 to 6 when executing the computer program.

9. A computer-readable storage medium, characterized in that: A computer program is stored which can be loaded by a processor and execute the method according to any one of claims 1 to 6.

Images