Method and device for preventing bullying on campus

Abstract

The present invention discloses a method and device for campus anti-bullying alarm, which collects image information and voice information in a certain period of time in the campus environment, compares and assigns behavior features, facial expression features and voice features extracted from the image information and voice information with information in a database, and calculates whether the current state is bullying through a specific algorithm. The beneficial effect of the present invention is that it can accurately identify whether the current state is bullying, avoid false alarms, and alarm in time.

Description

Technical Field

[0001] The present invention belongs to the technical field of campus prevention alarm, and in particular relates to a campus anti-bullying alarm method and device. Background Art

[0002] With the development of society and the popularization of education, campus safety issues have received increasing attention. Bullying has become a serious social problem on campus, causing serious harm to students' physical and mental health. Traditional campus safety monitoring systems usually rely on equipment such as cameras and alarms, but these devices can only record the occurrence of events, lack real-time monitoring and early warning functions, and cannot effectively prevent bullying. Therefore, there is an urgent need for an innovative campus anti-bullying system that can monitor the campus environment in real time and accurately identify bullying behavior, so as to improve campus safety and protect the physical and mental health of teachers and students.

[0003] Chinese patent application number 2023110344048 discloses an anti-bullying identification system and method based on semantic analysis and machine learning, which relates to the technical field of smart wearable devices, including obtaining voice information in a campus environment, extracting Mel-frequency cepstral coefficient characteristic voice in the voice information; matching the characteristic voice with the data in the set initial characteristic voice library, and performing bullying voice recognition based on the similarity; obtaining the weight value of the characteristic voice with high similarity, and triggering an alarm when the weight value reaches a preset bullying threshold. If the weight value is less than the preset bullying threshold, the characteristic voice is matched with the initial characteristic voice library again within a preset time and the weight value is calculated. When the weight value reaches the preset bullying threshold, an alarm is triggered; if the weight value is less than the preset bullying threshold and no new characteristic voice is matched within the preset time, the bullying identification process is terminated.

[0004] The above scheme can improve the recognition accuracy and speed by extracting voice features. However, at present, since students like to joke with each other and are prone to bickering after a dispute, it is difficult to ensure the accuracy of recognition by voice recognition alone. If multiple false alarms occur, when bullying actually occurs, it is easy for managers to be careless and fail to achieve the effect of bullying warning. Summary of the invention

[0005] In order to solve the above problems, the present invention provides a method for preventing bullying and alarming on campus, comprising the following steps: S1, data collection, using image sensors to collect image information of students in a certain period of time in the campus environment, and using sound sensors to collect voice information of students in the same period of time in the campus environment, and passing all the collected data to the analysis module; S2, data analysis, using the artificial intelligence module to read the behavior characteristics and facial expression characteristics of the students in the image information, and the artificial intelligence module to read the voice characteristics in the voice information; matching the above behavior characteristics, facial expression characteristics and voice characteristics with the feature data in the database, assigning values ​​to each behavior characteristic, facial expression characteristic and voice characteristic, and obtaining the behavior characteristic value Z1, the facial expression characteristic value Z2 and the voice characteristic value Z3; S3, obtaining preset parameters, reading the comprehensive threshold value T1 of each behavior feature, the comprehensive threshold value T2 of each facial expression feature, and the comprehensive threshold value T3 of each voice feature in the database, wherein the comprehensive threshold value T1 of the behavior feature is obtained based on the amplitude, speed, and duration of the behavior, the comprehensive threshold value T2 of the facial expression feature is obtained based on the intensity and duration of the expression, and the comprehensive threshold value T3 of the voice feature is obtained based on the frequency of occurrence and the derogatory degree of the sentence, and the weight of the behavior feature is set to W1, the weight of the facial expression feature is set to W2, and the weight of the voice feature is set to W3, wherein W1+W2+W3=1; S4. Determine the bullying status, set the bullying threshold to A, the value range of A is between 0.8-1.2, calculate the bullying value C, and substitute the behavior feature value Z1, the facial expression feature value Z2, and the voice feature value Z3 into the calculation formula; C=W1(Z1 / T1)+W2(Z2 / T2)+W3(Z3 / T3). If C>A, it is determined to be a bullying state, and step S5 is executed; if C<A, it is determined to be a non-bullying state, and step S1 is executed again; S5. Trigger an alarm, send an alarm message to the anti-bullying platform, and simultaneously send GPS coordinate information to report the location where the bullying behavior occurred.

[0006] Preferably, the speech features in the speech information read by the artificial intelligence module in step S2 include semantic features and volume features.

[0007] Preferably, the value of W1 is 50%, the value of W2 is 30%, and the value of W3 is 20%.

[0008] Preferably, in step S4, if C>A, the bullying value C1 for the next time period is calculated; if C1≥the average value of C and C1, step S5 is executed; if C1<the average value of C and C1, the bullying value C2 for the next time period is calculated; if C2≥the average value of C, C1 and C2, step S5 is executed; if C2<the average value of C, C1 and C2, step S1 is re-executed.

[0009] Preferably, after the alarm is triggered in step S5, the alarm information is sent to the anti-bullying platform and the student's facial information is added to the database. When the data is analyzed again later, the facial expression features of the corresponding students are assigned a value that is 0.1-0.3 higher than the original assignment.

[0010] Preferably, after the alarm is triggered in step S5, an alarm sounds on the spot to drive away the bullying students, and the entire behavior process on the spot is recorded.

[0011] A device for campus anti-bullying alarm method, comprising an image acquisition module, a voice acquisition module, an artificial intelligence module, a feature analysis module, a database module, a data calculation module, a communication module and an anti-bullying platform, wherein the image acquisition module is used to collect student behavior feature data and facial expression feature data, the voice acquisition module is used to collect student voice feature data, the artificial intelligence module assigns student behavior feature data, facial expression feature data and voice feature data to information stored in the database module through the feature analysis module, the data calculation module calculates the bullying value according to the above assignment and the parameters preset in the database, and determines whether to send an alarm message to the anti-bullying platform through the communication module according to the judgment result. Preferably, it also includes an audible and visual alarm module, which emits an audible and visual alarm on the spot when the alarm information is triggered.

[0012] Preferably, it also includes a GPS positioning module, and when the alarm information is triggered, the GPS positioning module is used to send the location information.

[0013] Preferably, it also includes a microphone and a speaker device, through which two-way remote communication can be achieved.

[0014] The advantages of the present invention are: This solution combines the behavioral features and facial expression features collected by the image sensor and the voice features collected by the sound sensor, and matches and assigns values ​​to the data in the database through the artificial intelligence module, so that bullying behavior can be more comprehensively identified. Compared with single voice recognition, this multi-dimensional recognition method can capture the characteristics of bullying behavior more comprehensively, thereby improving the accuracy of recognition.

[0015] This scheme can more comprehensively reflect the persistence and changing trend of bullying behavior by continuously calculating the bullying values ​​of multiple duration periods and comparing them with the average value.

[0016] This solution introduces a mechanism to dynamically adjust the voice feature value according to the volume feature size. This means that the system can respond more flexibly to students' emotional changes and can more sensitively identify potential bullying behaviors.

[0017] This program provides schools and educational institutions with a comprehensive and effective strategy for dealing with bullying through on-site alarms and video recording, immediate response and prevention, clear instructions and actions, behavior records and evidence preservation, improved deterrence effects, support for subsequent investigations and processing, and enhanced campus safety awareness.

[0018] The device used for campus anti-bullying alarm method in this scheme integrates multiple functional modules such as image acquisition, voice acquisition, artificial intelligence analysis, feature assignment, data calculation, communication transmission, etc., forming a complete and efficient anti-bullying alarm system. This comprehensiveness and integration ensures that the system can comprehensively monitor and analyze students' behavior from multiple dimensions and angles, improving the accuracy and efficiency of anti-bullying. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0021] Embodiment 1, a method for preventing bullying in schools, comprising the following steps: S1, data collection, using image sensors to collect image information of students in a certain period of time in the campus environment, and using sound sensors to collect voice information of students in the same period of time in the campus environment, and passing all the collected data to the analysis module; S2. Data analysis: Data analysis uses artificial intelligence modules to read the behavioral characteristics and facial expression characteristics of students in image information. The behavioral characteristics include twirling, pulling, pushing, beating, fighting, collision, trampling, etc. Facial expression characteristics include frowning eyebrows caused by anger, eyebrows down and gathered in the middle, eyes widening; mouth corners pulled down and sneered due to dissatisfaction; mouth corners raised due to provocation or ridicule of the victim; eyes slightly narrowed due to disdain and ridicule; bullies may also be nervous because of fear of being discovered or punished, which is manifested as slightly wrinkled eyebrows, flickering eyes, and tight mouth corners; At the same time, the artificial intelligence module reads the voice features in the voice information; such as derogatory words such as rookie and idiot; threatening words such as be careful, I'm welcome to you, and I dare not speak; forced words such as don't, I'm wrong. A large number of behavioral features, facial expression features, and voice features about bullying are stored in the database. The artificial intelligence module matches the extracted behavioral features, facial expression features, and voice features with the data in the database, and assigns values ​​to the corresponding behavioral features, facial expression features, and voice features. In the database, each behavior will have a corresponding score according to the amplitude, speed, and duration. The larger the amplitude, the faster the speed, and the longer the duration, the higher the score. Each behavioral feature has a comprehensive threshold T1, which refers to the average of all scores of each behavioral feature. Each facial expression feature also has a corresponding score according to the intensity and duration of the expression. The greater the intensity and the longer the duration, the higher the score. Each facial expression feature has a comprehensive threshold T2, which refers to the average of all scores of each facial expression feature. Each voice feature also has a corresponding score according to the frequency of occurrence and the degree of derogatory meaning. The higher the frequency of occurrence and the greater the degree of derogatory meaning, the higher the score. Each voice feature has a comprehensive threshold T3, which refers to the average of all scores of each voice feature. The artificial intelligence module matches and assigns the extracted behavioral features, facial expression features, and voice features according to the scores in the database, and then brings the corresponding feature scores into the calculation formula for calculation. The calculation formula of the bullying value C is: C=W1(Z1 / T1)+W2(Z2 / T2)+W3(Z3 / T3). Where W1 is the weight of the behavioral feature, W2 is the weight of the facial expression feature, W3 is the weight of the voice feature, and W1+W2+W3=1. The weight of each behavioral feature is different. In this embodiment, the value of W1 is 50%, the value of W2 is 30%, and the value of W3 is 20%. The adjustment of the weight is different according to the change of the scene. In this embodiment, the weight is set according to the scene on the playground. The probability of behavioral bullying on the playground is high. Therefore, the value of W1 is high. In the classroom, it is necessary to reduce the weight of the behavioral feature, increase the weight of the facial expression feature and the weight of the voice feature, because compared with the playground, the classroom is crowded and concentrated, the probability of behavioral bullying is low, and the probability of verbal bullying is high. The weights can be adjusted according to the actual situation to adapt to the identification of bullying behavior in different environments and scenarios. This dynamic weight adjustment method makes the system more flexible and intelligent.

[0022] The obtained bullying value result is compared with the set bullying threshold A. If C>A, it is determined to be a bullying state. If C<A, it is determined to be a non-bullying state. The value range of the bullying threshold A is 0.8-1.2. If it is lower than 0.8, the system will more sensitively identify bullying behavior, but it may also cause more false alarms. If it is higher than 1.2, it will more accurately identify non-bullying behavior, but it may also miss some actual bullying incidents. For this reason, in this embodiment, the bullying threshold A can be preferably set to 1, which can reduce false alarms and avoid missing bullying incidents.

[0023] The selected technicians simulated bullying behaviors of bullies and victims, in which the bullies pushed and insulted the victims, and then collected behavioral features, facial expression features, and voice features. The artificial intelligence module assigned values ​​to the behavioral features, facial expression features, and voice features, and the values ​​of Z1, Z2, and Z3 were 0.9, 0.7, and 0.5, respectively. The values ​​of the comprehensive thresholds T1, T2, and T3 were set to 0.8, 0.6, and 0.4, respectively. According to the above values ​​and parameters, they were substituted into the calculation formula, S = 0.5 * (0.9 / 0.8) + 0.3 * (0.7 / 0.6) + 0.2 * (0.5 / 0.4) = 0.5625 + 0.35 +0.25 = 1.1625. The bullying value C was 1.1625. The bullying value C is greater than the bullying threshold A, so the system identifies this behavior as bullying, executes step S5, triggers an alarm, sends an alarm message to the anti-bullying platform, and simultaneously sends GPS coordinate information to report the location where the bullying behavior occurred. Otherwise, re-execute step S1.

[0024] In order to further improve the accuracy of judgment, in step S4, if C>A, the bullying value C1 of the next time period is calculated. If C1≥the average value of C and C1, step S5 is executed; if C1<the average value of C and C1, the bullying value C2 of the next time period is calculated. If C2≥the average value of C, C1 and C2, step S5 is executed; if C2<the average value of C, C1 and C2, step S1 is re-executed. If the bullying behavior increases significantly in a short period of time (such as C1≥C and the average value of C1), the subsequent alarm steps are executed immediately; if the bullying behavior shows a fluctuating or downward trend (such as C1<C and the average value of C1), more observation time is given to more accurately judge the true situation of the bullying behavior. By continuously calculating the bullying values ​​of multiple durations and comparing them with the average value, the persistence and changing trend of bullying behavior can be more comprehensively reflected.

[0025] Embodiment 2, this embodiment has the same parts as embodiment 1. In this embodiment, the voice features read by the artificial intelligence module in the voice information include semantic features and volume features. When the artificial intelligence module assigns values ​​to the voice features, the assigned values ​​will be increased according to the size of the recognized volume features. For example, if the original voice feature value Z3 is 0.5, if the volume of the voice is large, it is determined that the speaking student is in an angry state, and is more likely to be in a bullying state. Therefore, the voice feature value Z3 is increased by 0.1 on the original basis. If the volume of the voice is small, it is determined that the student is not in an angry state, and the voice feature assignment is not increased. By introducing a mechanism for dynamically adjusting the voice feature value according to the size of the volume feature, it can respond more flexibly to the emotional changes of students, and the system can more keenly identify potential bullying behaviors.

[0026] After the alarm is triggered in step S5, the facial information of the student is sent to the anti-bullying platform and added to the database at the same time as the alarm information is sent. When the data is analyzed again later, the facial expression features of the corresponding students are assigned a value, which is increased by 0.1-0.3 compared to the original assignment. Since the bully or the bullied is likely to be in a bullying state again, when the facial expression features of the corresponding students are assigned a value, if it is detected that the student has been in a bullying state, the size of the facial expression feature assignment is increased. For students who have been in a bullying state, the system increases the attention by increasing the assignment of their facial expression features. This helps schools or relevant institutions to promptly detect and intervene in potential bullying behaviors and protect the physical and mental health of students.

[0027] After the alarm is triggered in step S5, an alarm is sounded on the spot to drive away the bullying students, and the entire behavior process on the scene is recorded. The immediate issuance of the alarm can quickly attract the attention of the on-site personnel and effectively stop the ongoing bullying behavior. This immediate response helps to reduce the harm and psychological shadow caused by bullying to students. The alarm is not only a warning to the bullying behavior, but also a clear instruction to the surrounding students to guide them to take action, such as seeking help or staying away from the bullying scene. This helps to form a positive campus atmosphere and encourage students to support and protect each other. Recording the entire behavior process on the scene can ensure that the authenticity and accuracy of the bullying behavior are recorded. These video materials can be used as important evidence for the subsequent handling of bullying incidents, which helps schools and educational institutions to deal with bullying seriously and maintain campus order and fairness. At the same time, the issuance of the alarm and the existence of video recordings can have a strong deterrent effect on potential bullies. They may restrain their behavior for fear of being recorded and punished, thereby effectively reducing the occurrence of campus bullying incidents.

[0028] Embodiment 3, a device for campus anti-bullying alarm method, including an image acquisition module, a voice acquisition module, an artificial intelligence module, a feature analysis module, a database module, a data calculation module, a communication module and an anti-bullying platform, the image acquisition module is used to collect student behavior feature data and facial expression feature data, the voice acquisition module is used to collect student voice feature data, the artificial intelligence module assigns student behavior feature data and facial expression feature data as well as voice feature data to the information stored in the database module through the feature analysis module, the data calculation module calculates the bullying value according to the above assignment and the parameters preset in the database, and determines whether to send an alarm message to the anti-bullying platform through the communication module according to the judgment result, and also includes an audio-visual alarm module, when the alarm message is triggered, the audio-visual alarm module sends an audio-visual alarm on the spot, and also includes a GPS positioning module, when the alarm message is triggered, the GPS positioning module is used to send location information, and also includes a microphone and a speaker device, through which two-way remote communication can be achieved.

[0029] The device integrates multiple functional modules such as image acquisition, voice acquisition, artificial intelligence analysis, feature assignment, data calculation, communication transmission, etc., forming a complete and efficient anti-bullying alarm system. This comprehensiveness and integration ensures that the system can comprehensively monitor and analyze students' behaviors from multiple dimensions and angles, improving the accuracy and efficiency of anti-bullying. Through microphones, speaker equipment and communication modules, the device can achieve two-way remote communication, allowing relevant personnel to monitor and guide the on-site situation remotely. This remote communication and monitoring function not only improves the flexibility of the system, but also enhances the efficiency and coverage of campus safety management.

[0030] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for preventing bullying on campus, characterized in that: The steps include: S1, data collection, using image sensors to collect image information of students in a certain period of time in the campus environment, and using sound sensors to collect voice information of students in the same period of time in the campus environment, and passing all the collected data to the analysis module; S2, data analysis, using the artificial intelligence module to read the behavior characteristics and facial expression characteristics of the students in the image information, and the artificial intelligence module to read the voice characteristics in the voice information; matching the above behavior characteristics, facial expression characteristics and voice characteristics with the feature data in the database, assigning values ​​to each behavior characteristic, facial expression characteristic and voice characteristic, and obtaining the behavior characteristic value Z1, the facial expression characteristic value Z2 and the voice characteristic value Z3; S3, obtaining preset parameters, reading the comprehensive threshold value T1 of each behavior feature, the comprehensive threshold value T2 of each facial expression feature, and the comprehensive threshold value T3 of each voice feature in the database, wherein the comprehensive threshold value T1 of the behavior feature is obtained based on the amplitude, speed, and duration of the behavior, the comprehensive threshold value T2 of the facial expression feature is obtained based on the intensity and duration of the expression, and the comprehensive threshold value T3 of the voice feature is obtained based on the frequency of occurrence and the derogatory degree of the sentence, and the weight of the behavior feature is set to W1, the weight of the facial expression feature is set to W2, and the weight of the voice feature is set to W3, wherein W1+W2+W3=1; S4. Determine the bullying status, set the bullying threshold to A, the value range of A is between 0.8-1.2, calculate the bullying value C, and substitute the behavior feature value Z1, the facial expression feature value Z2, and the voice feature value Z3 into the calculation formula; C=W1(Z1 / T1)+W2(Z2 / T2)+W3(Z3 / T3). If C>A, it is determined to be a bullying state, and step S5 is executed; if C<A, it is determined to be a non-bullying state, and step S1 is executed again; S5. Trigger an alarm, send an alarm message to the anti-bullying platform, and simultaneously send GPS coordinate information to report the location where the bullying behavior occurred.

2. The campus anti-bullying alarm method according to claim 1 is characterized in that: In step S2, the artificial intelligence module reads the voice features in the voice information including semantic features and volume features.

3. The campus anti-bullying alarm method according to claim 2 is characterized in that: The value of W1 is 50%, the value of W2 is 30%, and the value of W3 is 20%.

4. The campus anti-bullying alarm method according to claim 3 is characterized in that: In step S4, if C>A, the bullying value C1 for the next time period is calculated; if C1≥the average value of C and C1, step S5 is executed; if C1<the average value of C and C1, the bullying value C2 for the next time period is calculated; if C2≥the average value of C, C1 and C2, step S5 is executed; if C2<the average value of C, C1 and C2, step S1 is re-executed.

5. The campus anti-bullying alarm method according to claim 4 is characterized in that: After the alarm is triggered in step S5, the alarm information is sent to the anti-bullying platform and the student's facial information is added to the database. When the data is analyzed again later, the facial expression features of the corresponding students are assigned a value that is 0.1-0.3 higher than the original assignment.

6. The campus anti-bullying alarm method according to claim 5 is characterized in that: After the alarm is triggered in step S5, an alarm sounds on the spot, the bullying students are driven away, and the entire behavior process on the spot is recorded.

7. A device for campus anti-bullying alarm method according to claim 6, characterized in that: It includes an image acquisition module, a voice acquisition module, an artificial intelligence module, a feature analysis module, a database module, a data calculation module, a communication module and an anti-bullying platform. The image acquisition module is used to collect student behavior feature data and facial expression feature data, the voice acquisition module is used to collect student voice feature data, the artificial intelligence module assigns student behavior feature data, facial expression feature data and voice feature data to the information stored in the database module through the feature analysis module, the data calculation module calculates the bullying value according to the above assignment and the parameters preset in the database, and determines whether to send an alarm message to the anti-bullying platform through the communication module according to the judgment result.

8. The device for campus anti-bullying alarm method according to claim 7 is characterized in that: It also includes an audible and visual alarm module, which emits an audible and visual alarm on the spot when the alarm information is triggered.

9. The device for campus anti-bullying alarm method according to claim 8 is characterized in that: It also includes a GPS positioning module, which is used to send location information when the alarm information is triggered.

10. The device for campus anti-bullying alarm method according to claim 9 is characterized in that: It also includes microphone and speaker devices, through which two-way remote communication can be achieved.

Images