A police dispatching method, device and storage medium

Abstract

This invention relates to a police dispatching method, device, and storage medium. The method includes: responding to an incoming alarm call, recognizing voice information based on a voice intelligent recognition model to obtain a recognition result; the recognition result includes semantic information and caller association information; determining case information based on the semantic information and the caller association information; wherein the case information includes a case type; obtaining police personnel information for the corresponding jurisdiction based on the case information and matching it with the case type to generate pre-dispatch information; the police personnel information includes expertise in handling legal provisions; obtaining manual dispatch information corresponding to the alarm call and performing similarity detection on the pre-dispatch information, and determining a dispatch order based on the similarity detection result. The technical solution of this invention can send pre-dispatch information to alarms, allowing jurisdictions that may need to respond to prepare, allocate police personnel, and accurately dispatch officers based on the case type.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a police dispatching method, device and storage medium. Background Technology

[0002] The current police dispatch method is as follows: the receiving officer receives the alarm call from the party concerned, obtains the case information, and then selects available police officers from the vicinity of the crime scene. The officer then sends a message to the officer through the police terminal, informing him of the location of the crime and the party's contact information, so that the officer can go to the crime scene to handle the alarm.

[0003] The personnel receiving the caller need to listen to the caller's voice message and make a police situation assessment based on the sound, then assign the police task to the jurisdiction. This process, where the personnel receiving the caller handle the alarm information, involves a time lag between the alarm being raised and the police task being assigned, resulting in a delay in response time. The jurisdiction can only receive police information sent or relayed by the alarm center. This information generally includes the location and the event, but cannot determine the type of case, making it difficult to dispatch police officers in a targeted manner. Summary of the Invention

[0004] This invention provides a police dispatching method, device, and storage medium, with the aim of improving the speed of dispatching orders and dispatching police officers based on cases during the dispatching process.

[0005] In a first aspect, embodiments of the present invention provide a police dispatching method, including:

[0006] In response to an incoming alarm call, the system identifies the voice information based on a voice intelligent recognition model to obtain a recognition result; wherein, the recognition result includes semantic information and alarm caller identity association information;

[0007] Based on the semantic information and the identity association information of the person who made the report, the case information is determined; wherein, the case information includes the case type;

[0008] Based on the case information, obtain the information of police officers in the corresponding jurisdiction and match it with the case type to generate pre-dispatch information; the police officer information includes their expertise in handling legal provisions;

[0009] The manual dispatch information corresponding to the alarm call and the pre-dispatch information are obtained and similarity detection is performed. The dispatch order is determined based on the similarity detection result.

[0010] In a second aspect, embodiments of the present invention provide an electronic device, comprising:

[0011] One or more processors;

[0012] Memory, used to store one or more programs;

[0013] When the one or more programs are executed by the one or more processors, the one or more processors implement the police dispatching method provided in any embodiment of the present invention.

[0014] Thirdly, embodiments of the present invention provide a storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform a police dispatching method as provided in any embodiment of the present invention.

[0015] The present invention provides a police dispatching method, device and storage medium that uses a voice intelligent recognition model to identify alarm voices and determine the case type, pre-dispatch orders and determine dispatch instructions, which solves the problems of dispatch delay and poor targeting, realizes that pre-dispatch information allows jurisdictions that may need to be dispatched to prepare, improves dispatch speed and enables precise dispatch of orders based on case type. Attached Figure Description

[0016] Figure 1 A flowchart of a police dispatching method provided in Embodiment 1 of the present invention;

[0017] Figure 2 This is a schematic diagram of the structure of a police system provided in Embodiment 1 of the present invention;

[0018] Figure 3 This is a flowchart of the voice intelligent recognition model registration stage provided in Embodiment 1 of the present invention;

[0019] Figure 4 This is a flowchart of the recognition stage of the speech intelligent recognition model provided in Embodiment 1 of the present invention;

[0020] Figure 5 This is a schematic diagram of the similarity detection model provided in Embodiment 1 of the present invention;

[0021] Figure 6 This is a schematic diagram of the structure of an electronic device provided in Embodiment 2 of the present invention. Detailed Implementation

[0022] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures. Example 1

[0023] Figure 1 This is a flowchart of a police dispatching method provided in Embodiment 1 of the present invention. This embodiment is applicable to situations where police dispatching is performed through a police system. The method can be executed by computer equipment, such as a server. Figure 2 As shown, the server includes a cloud server and multiple regional servers. The cloud server can receive emergency calls, while the regional servers communicate with police terminals. The layout of the police system can also differ; for example, instead of a cloud server, voice information can be synchronized to each regional server, which then performs voice recognition to determine if the case falls within its jurisdiction. This method specifically includes:

[0024] Step 110: In response to the incoming alarm call, the voice information is recognized based on the intelligent voice recognition model to obtain the recognition result;

[0025] The recognition results include semantic information and information related to the caller's identity. Through the recognition of voice information, semantic information includes the event content, location, and people involved. Voice information can reveal information associated with the caller, including their identity, gender, and mood. Furthermore, voice signals often contain rich environmental background sounds, such as various noises and the voices of non-callers. The intelligent voice recognition model can remove noise from emergency calls and identify semantic information and information related to the caller's identity.

[0026] Step 120: Determine case information based on the semantic information and the association information of the person who made the report;

[0027] The case information includes case types. The jurisdiction server may contain a classification database for corresponding case types. Different case types are set according to preset regulations.

[0028] Step 130: Based on the case information, obtain the information of police officers in the corresponding jurisdiction and match it with the case type to generate pre-dispatch information;

[0029] The information on police officers includes their expertise in handling legal provisions. The jurisdiction server can set case tags for police officers, and jurisdictional staff can modify these tags to prioritize assigning them cases based on their expertise. A mapping between expertise tags and a case type classification database is established. This mapping allows for the selection of overlapping officers, identifying them as potential dispatch personnel for pre-dispatch orders. It should be noted that a pre-dispatch order notifies officers meeting the dispatch criteria to prepare for a response or provides them with directions to the dispatch location. The pre-dispatch order information is sent to the corresponding jurisdiction server, which filters the dispatch personnel's task information and sends the pre-dispatch order information to the mobile terminals associated with the officers meeting the dispatch criteria. These mobile terminals can be phones with a police app installed or police terminals corresponding to the police system.

[0030] The recognition results include semantic information and alarm-related information. Semantic information includes time content, time and location, and event task. Alarm-related information includes (alarm-reporter's emotion and gender). There may be some unclear aspects in some recognition items. For example, the location identified by the voice recognition model may be two: jurisdiction A and jurisdiction B. In this case, when sending the pre-dispatch order, the pre-dispatch order will be sent to jurisdiction A and jurisdiction B at the same time.

[0031] The following scenarios may occur when a pre-dispatch message is sent: If the caller is female and is experiencing any of the following emotions: anger, fear, sadness, or disgust, the pre-dispatch message requires at least one female police officer to respond to the call to calm the caller down. If the location of the incident is less than a preset distance (e.g., 3km) from multiple jurisdictions, and the incident is classified as a major case or a case involving a large number of people, the pre-dispatch message will be sent to all jurisdictions within a preset distance from the incident location. If the caller is experiencing happiness or boredom, the pre-dispatch message will not be sent to avoid wasting police resources on false alarms; the message will be sent manually.

[0032] Step 140: Obtain the manual dispatch information and the pre-dispatch information corresponding to the alarm call, perform similarity detection, and determine the dispatch instruction based on the similarity detection result.

[0033] The process involves the following steps: After receiving an emergency call, the receiving officer generates a manual dispatch order based on the voice information they hear. This manual dispatch order is then sent to the corresponding cloud server in the police system. The cloud server compares the pre-dispatch information with the manual dispatch order and issues an emergency dispatch command to the jurisdiction server. The jurisdiction server then forwards the command to the officer's terminal.

[0034] The technical solution in this embodiment recognizes the voice information while the operator answers the emergency call and generates pre-dispatch information, which is then sent to the corresponding jurisdiction so that local police officers can prepare for dispatch. The system uses a similarity check between manual dispatch and pre-dispatch information to determine the final dispatch, guiding dispatch and avoiding oversights due to reliance solely on intelligent dispatch. The server leverages its computational advantages to recognize the voice information, shortening recognition time and improving both speed and accuracy.

[0035] Optionally, the process of recognizing speech information based on the intelligent speech recognition model to obtain recognition results includes:

[0036] The voice information is input into a semantic recognition model to obtain the semantic information, wherein the semantic information includes event content, event location, and event participants;

[0037] The voice information is input into an identity association recognition model to obtain the identity association information of the person making the alarm, wherein...

[0038] The voice information is input into the first SVM classifier for gender matching to obtain the corresponding gender label;

[0039] The voice information is input into the second SVM classifier and the third SVM classifier corresponding to the gender label, respectively, for emotion matching and speaker matching. The classification result is determined by voting based on the minimum distance criterion, and the corresponding emotion label and identity label are obtained.

[0040] In one implementation of this embodiment, the speech intelligent recognition model can be a baseline system, trained a preset number of times using a speech training set. The speech training set includes neutral, angry, fearful, happy, bored, sad, and disgusted emotions. During training, acoustic features categorized into gender, emotion, and identity are concatenated and combined into a fusion feature vector for multi-dimensional information recognition. Figure 3 and Figure 4 The baseline system used in the speech intelligent recognition model takes a fused feature vector as input. This system is a novel system constructed by connecting multiple SVM classifiers in series and parallel, and includes recognition of three types of speakers: gender, emotion, and identity. Figure 3 During the registration phase, for SVM training, gender classification is first performed using the combined features of multidimensional speaker information recognition, and the classification model is saved. Then, based on the classification results, all speech samples of the same gender are combined to form two speech library subsets labeled male and female. Then, speaker identity and emotion are classified separately in the redefined male and female speech data. Finally, the gender-based model training results are saved as Model 2 and Model 3 in sequence. Figure 4 In the sample recognition stage, the test speech is first matched for gender according to Model 1 to obtain label 1 corresponding to gender. Then, it is matched for emotion and speaker according to the two models saved during the training stage. The classification result is determined by voting based on the minimum distance criterion, and finally, the corresponding emotion label 2 and identity label 3 are obtained. In addition, multi-dimensional decision-making is different from traditional single-dimensional decision-making. It needs to consider the results of three labels at the same time and calculate the average value to obtain the final decision result.

[0041] The baseline system is essentially an algorithm for solving multi-label classification, employing a "problem transformation approach." First, multi-dimensional speaker information recognition is viewed as a classification problem involving three categories of labels. Then, through appropriate feature extraction and other speech waveform data processing, the original multi-dimensional information classification problem is transformed into a problem utilizing several existing, more mature multi-classification algorithms. The trained intelligent recognition model can accurately identify the gender, emotion, and identity of the person making the alarm.

[0042] Optionally, determining the case information based on the semantic information and the association information of the person who reported the incident includes:

[0043] Based on the event content, event location, people involved in the event, and the information related to the person who reported the incident, the event content and the case type are determined; wherein, the case type is set according to preset legal provisions.

[0044] Optionally, the step of obtaining police officer information for the corresponding jurisdiction based on the case information and matching it with the case type to generate pre-dispatch information includes:

[0045] The corresponding jurisdiction is determined based on the location of the incident;

[0046] Retrieve the police officer information for the corresponding jurisdiction from the police officer database;

[0047] The officers were identified by comparing the legal provisions they were good at handling with the types of cases described.

[0048] Among these measures, police officers who are skilled at handling cases where legal provisions overlap with the aforementioned case types will be designated as dispatch personnel. This approach ensures that incidents are handled by officers who are proficient in that type of case, making the response more targeted and improving the efficiency and quality of case handling.

[0049] Optionally, the police officer information may also include work status, working hours, and length of service.

[0050] After comparing the police officers' expertise in handling legal provisions with the type of case to determine the dispatching officers, the process also includes:

[0051] If all the dispatched officers are in the dispatched state, the case matching degree is calculated among the non-dispatched officers in the corresponding jurisdiction, and the officer with the highest case matching degree is selected as the dispatched officer. The case matching degree is obtained by multiplying the case type matching degree, working hours matching degree, and police service length matching degree by their respective matching weights and summing the results.

[0052] If overlapping police officers are all on duty and there is no officer to handle the corresponding case type, the case matching degree is calculated among the remaining non-dispatch personnel, and the officer with the highest case matching degree is selected for dispatch. The case matching degree includes case type matching degree, working hours matching degree, and police service length matching degree. Each matching degree has a matching weight, i.e., case matching degree = case type matching degree * matching weight 1 + working hours matching degree * matching weight 2 + police service length matching degree * matching weight 3.

[0053] Optionally, based on the preset legal provisions corresponding to the case type, the numerical difference between the case type and the legal provision number that the police officers are good at handling is determined to determine the case type matching degree;

[0054] The working time matching degree is determined based on the degree of matching between the average processing time of the case type and the remaining working time of the police officers.

[0055] The higher the required length of police service based on the importance of the case, the better the match between the required length of police service and the case.

[0056] The case type matching degree is determined based on the voice recognition results to identify the relevant article of the Criminal Law or the Public Security Administration Punishment Law that may be violated, and then the numerical difference between this article and the specific legal provisions the officer is skilled at handling is assessed. A smaller numerical difference indicates a higher case type matching degree. The working time matching degree is the degree of matching between the average processing time for this type of case based on big data statistics and the officer's remaining working time. If the remaining working time is greater than the average processing time, and the difference is smaller, the working time matching degree is higher. This avoids wasting police resources and prevents officers from working excessive overtime, affecting their normal rest. The officer's remaining working time can be determined based on the current time and the officer's shift. For example, if the officer's shift is from 8:00 AM to 5:00 PM, and the current time is 4:00 PM, then the remaining working time is 1 hour. The seniority matching degree is related to the importance of the case; cases involving emergencies or large numbers of people require officers with longer service records. It's important to note that longer service time doesn't necessarily equate to a higher match between service time and target duration. For example, in pursuit-type cases, shorter service time often indicates better physical stamina among officers, resulting in a higher match between service time and target duration. Each jurisdiction can adjust matching weights 1, 2, and 3 according to its specific characteristics to suit the nature of the cases within its jurisdiction. For instance, if a jurisdiction primarily handles economic crime cases, matching weight 3 can be appropriately increased to allow more experienced officers to respond quickly. Similarly, if a jurisdiction has experienced a high caseload and significant stamina depletion among officers recently, matching weight 2 can be increased to allow for rest and recovery.

[0057] Optionally, the step of obtaining the manual dispatch information corresponding to the alarm call and the pre-dispatch information, performing similarity detection, and determining the dispatch instruction based on the similarity detection result includes:

[0058] If the similarity between the manual dispatch information and the pre-dispatch information is greater than a first preset threshold, the pre-dispatch information is identified as the final dispatch, and the dispatch order is issued to the jurisdiction corresponding to the final dispatch.

[0059] If the similarity between the manual dispatch information and the pre-dispatch information is less than the first preset threshold and greater than the second preset threshold, compare the similarity information in the event location information set, and send the dispatch order to the jurisdiction where the event location in the manual dispatch information and the event location in the pre-dispatch information overlap.

[0060] If the similarity between the manually dispatched order information and the pre-dispatch order information is less than the second preset threshold, the manually dispatched order information is identified as the final dispatch order, and the dispatch instruction is issued to the jurisdiction corresponding to the final dispatch order.

[0061] Optionally, the step of obtaining the manual dispatch information corresponding to the alarm call and the pre-dispatch information and performing similarity detection includes:

[0062] Based on the word vector model, a first text vector corresponding to the pre-dispatch information and a second text vector corresponding to the manual dispatch information are generated.

[0063] The first local feature information of the first text vector and the second local feature information of the second text vector are extracted using a preset convolutional neural network.

[0064] A self-attention mechanism is used to extract the first self-attention information of the first text vector and the second self-attention information of the second text vector;

[0065] The first semantic matrix is ​​obtained by concatenating the first text vector, the first local feature information, and the first self-attention information; the second semantic matrix is ​​obtained by concatenating the second text vector, the second local feature information, and the second self-attention information.

[0066] The first semantic matrix and the second semantic matrix are used to extract the first feature information and the second feature information using a bidirectional long short-term memory network;

[0067] First semantic features and second semantic features are extracted from the first feature information and the second feature information. The first semantic features and the second semantic features are concatenated to obtain a feature vector. The feature vector is then fed into a classifier for classification, and the cosine phase velocity between texts is calculated.

[0068] Among them, information similarity detection adopts Figure 5 The model shown uses Word2Vec to generate word vectors from pre-assigned and manually assigned order information. These vectors are then fed into the information interaction model for training. The information interaction layer is optimized using PSO-CNN, and attention and self-attention mechanisms are added to increase the information between sentences. After the information is combined, BiLSTM is used to extract features. Finally, after pooling, cosine similarity is used to calculate and predict the matching degree between the two texts.

[0069] The model's vector generation layer represents text as word vectors, facilitating subsequent semantic representation of the vectors. Converting text to vectors can be achieved using methods such as Word2Vec or GloVe pre-training. This embodiment primarily employs the Word2Vec method because of its simplicity, efficiency, and ability to obtain relatively accurate word vector representations from the corpus. This method uses a sliding window approach for feature extraction and has two training modes: one predicts the context words based on the target word, and the other predicts the target word based on the context words.

[0070] In the information interaction layer, convolutional neural networks (CNNs) are used to extract local features of the text. For short texts, sometimes the features of local information can represent the meaning of the entire sentence. Therefore, to address the semantic diversity problem in text matching, the model also extracts local features of the text for information interaction. Since the structure of CNNs is relatively complex and determined by hyperparameters, which cannot be obtained through training but are generally set based on existing experience, the hyperparameters will vary depending on the dataset and the task. Therefore, before conducting the experiments, this paper first uses a CNN to match the text matching dataset used in this paper. A particle swarm optimization algorithm combined with the CNN model is used to optimize the learning rate. After several iterations, a learning rate suitable for text matching on this dataset is obtained. Then, a CNN (PSO-CNN) is used to extract local feature information 1 and local feature 2 of text vector 1 and text vector 2. A self-attention mechanism is used to calculate deeper semantic information features and extract self-attention information 1 and self-attention information 2 of text vector 1 and text vector 2. Semantic matrix 1 is obtained by concatenating text vector 1, local feature information 1, and self-attention information. Semantic matrix 2 is obtained by concatenating text vector 2, local feature information 2, and self-attention information. Feature information 1 and feature information 2 are extracted from semantic matrix 1 and semantic matrix 2 using BiLSTM.

[0071] The interaction and prediction layer mainly involves concatenating the extracted semantic features and then calculating the text matching results. This paper uses the average pooling method to extract semantic features 1 and 2 from feature information 1 and feature information 2, concatenates the two semantic features to obtain a feature vector, and then feeds the feature vector into a classifier for classification. By calculating the cosine phase velocity between texts, the information similarity of the manually dispatched order information can be obtained.

[0072] In this embodiment, the similarity detection model vectorizes dispatch information to match two dispatch messages to each other's low-level semantic information. Then, PSO-CNN is used to extract the structural features of the text. Furthermore, to capture long-distance dependencies, word order and contextual information are considered during the matching process. To obtain deeper semantic information, interactions are performed within the text. Bi LSTM is used for feature extraction to address text structural issues and resolve long-distance dependencies in some sentences. This model can make more accurate judgments on the similarity of dispatch information based on phrases and context, resulting in more accurate results. Example 2

[0073] Figure 6 This is a schematic diagram of the structure of an electronic device provided in Embodiment 2 of the present invention, as shown below. Figure 6 As shown, the electronic device includes a processor 610, a memory 620, an input device 630, and an output device 640; the number of processors 610 in the electronic device can be one or more. Figure 6 Taking a processor 610 as an example; the processor 610, memory 620, input device 630, and output device 640 in the electronic device can be connected via a bus or other means. Figure 6 Taking the example of a connection between China and Israel via a bus.

[0074] The memory 620, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the police dispatching method in this embodiment of the invention. The processor 610 executes various functional applications and data processing of the electronic device by running the software programs, instructions, and modules stored in the memory 620, thereby implementing the aforementioned police dispatching method.

[0075] The memory 620 may primarily include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a given function; the data storage area may store data created based on terminal usage. Furthermore, the memory 620 may include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 620 may further include memory remotely located relative to the processor 610, which can be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0076] Input device 630 can be used to receive input digital or character information, and to generate key signal inputs related to user settings and function control of the electronic device. Output device 640 may include display devices such as a display screen. Example 3

[0077] Embodiment 3 of the present invention also provides a storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform a police dispatching method, including:

[0078] In response to an incoming alarm call, the system identifies the voice information based on a voice intelligent recognition model to obtain a recognition result; wherein, the recognition result includes semantic information and alarm caller identity association information;

[0079] Based on the semantic information and the identity association information of the person who made the report, the case information is determined; wherein, the case information includes the case type;

[0080] Based on the case information, obtain the information of police officers in the corresponding jurisdiction and match it with the case type to generate pre-dispatch information; the police officer information includes their expertise in handling legal provisions;

[0081] The manual dispatch information corresponding to the alarm call and the pre-dispatch information are obtained and similarity detection is performed. The dispatch order is determined based on the similarity detection result.

[0082] Of course, the computer-executable instructions provided in the embodiments of the present invention are not limited to the method operations described above, but can also perform related operations in the police dispatching method provided in any embodiment of the present invention.

[0083] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0084] Although the present invention has been described in detail above with general descriptions, specific embodiments, and experiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A police dispatching method characterized by comprising: The method comprises the following steps: In response to an accessed alarm call, voice information is recognized based on a voice intelligent recognition model to obtain a recognition result; wherein the recognition result comprises semantic information and alarm person association information; According to the semantic information and the alarm person association information, case information is determined; wherein the case information comprises a case type; According to the case information, police personnel information corresponding to a jurisdiction is obtained and matched with the case type to generate pre-assignment information; the police personnel information comprises a law provision that the police personnel is good at handling; wherein the pre-assignment information is used to send to one or more jurisdictions in the recognition result; The pre-assignment information and artificial assignment information corresponding to the alarm call are subjected to similarity detection, and a police instruction is determined according to a similarity detection result; The method of obtaining the pre-assignment information and the artificial assignment information corresponding to the alarm call and determining the police instruction according to the similarity detection result comprises the following steps: If the similarity of the artificial assignment information and the pre-assignment information is greater than a first preset threshold, the pre-assignment information is determined as final assignment, and the police instruction is issued to a jurisdiction corresponding to the final assignment; If the similarity of the artificial assignment information and the pre-assignment information is less than the first preset threshold and greater than a second preset threshold, similarity information in an event location information set is compared, and the police instruction is sent to a jurisdiction in which an event location in the artificial assignment information coincides with an event location in the pre-assignment information; If the similarity of the artificial assignment information and the pre-assignment information is less than the second preset threshold, the artificial assignment information is determined as final assignment, and the police instruction is issued to a jurisdiction corresponding to the final assignment.

2. The method of claim 1, wherein, The method of recognizing the voice information based on the voice intelligent recognition model to obtain the recognition result comprises the following steps: The voice information is input into a semantic recognition model to obtain the semantic information, wherein the semantic information comprises event content, an event location, and an event person; The voice information is input into an identity association recognition model to obtain the alarm person identity association information, wherein, The voice information is input into a first SVM classifier for gender matching to obtain a corresponding gender label; The voice information is respectively input into a second SVM classifier and a third SVM classifier corresponding to the gender label for emotion matching and speaker matching, and a classification result is determined by voting in a minimum distance criterion to obtain a corresponding emotion label and an identity label.

3. The method of claim 2, wherein, The method of determining the case information according to the semantic information and the alarm person association information comprises the following steps: According to the event content, the event location, the event person, and the alarm person association information, the event content and the case type are determined; wherein the case type is set according to a preset law provision.

4. The method according to claim 2 or 3, characterized in that, The method of obtaining the police personnel information corresponding to the jurisdiction and matching the police personnel information with the case type to generate the pre-assignment information comprises the following steps: The corresponding jurisdiction is determined according to the event location; The police personnel information corresponding to the jurisdiction is obtained from a police personnel database; The law provision that the police personnel is good at handling is compared with the case type to determine a police officer.

5. The method of claim 4, wherein, The police information further includes working state, working time length and police age time length; After the police officer who is good at handling the regulations of the case type is determined, the method further includes: If the working state of the police officer is the police state, the case matching degree is calculated among the police officers who are not on duty in the corresponding jurisdiction, and the police officer with the highest case matching degree is selected as the police officer, wherein the case type matching degree, the working time length matching degree and the police age time length matching degree are multiplied by the corresponding matching weight respectively to obtain the case matching degree.

6. The method of claim 5, wherein, According to the preset regulations corresponding to the case type, the digital difference between the case type and the police officer's good at handling regulations is determined, and the case type matching degree is determined. According to the matching degree between the average processing time length of the case type and the remaining working time length of the police officer, the working time length matching degree is determined. According to the demand of the case importance degree on the police age time length, the higher the police age time length matching degree is determined.

7. The method of claim 1, wherein, The similarity detection of the artificial dispatch information corresponding to the alarm telephone and the pre-dispatch information includes: Based on the word vector model, the first text vector corresponding to the pre-dispatch information and the second text vector corresponding to the artificial dispatch information are generated; Using a preset convolutional neural network, the first local feature information of the first text vector and the second local feature information of the second text vector are extracted; Using a self-attention mechanism, the first self-attention information of the first text vector and the second self-attention information of the second text vector are extracted; The first text vector, the first local feature information and the first self-attention information are spliced to obtain a first semantic matrix, and the second text vector, the second local feature information and the second self-attention information are spliced to obtain a second semantic matrix; The first semantic matrix and the second semantic matrix are extracted by using a bidirectional long short-term memory network to obtain first feature information and second feature information; First semantic features and second semantic features are extracted from the first feature information and the second feature information, the first semantic features and the second semantic features are connected to obtain a feature vector, and the feature vector is input into a classifier for classification to calculate the cosine similarity between the texts.

8. An electronic device, comprising: It includes: One or more processors; Memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the police dispatching method as claimed in any one of claims 1-7.

9. A storage medium containing computer-executable instructions, wherein: The computer executable instructions, when executed by a computer processor, are used to perform the police dispatching method as claimed in any one of claims 1-7. The computer executable instructions, when executed by a computer processor, are used to perform the police dispatching method as claimed in any one of claims 1-7.

Images