CN-121983055-A - Method and equipment for fusing intelligent voice recognition and dispatching desk for trunking communication

Abstract

The invention discloses a method and equipment for fusing intelligent voice recognition and dispatching desk for trunking communication, belonging to the crossing field of professional wireless communication and intelligent voice recognition. Aiming at the problem that the traditional trunking communication scheduling relies on manual work and a general ASR technology to be inadequately adapted in a professional scheduling scene, the scheme obtains a user voice command through a voice acquisition module, the user voice command is converted into a text through an ASR voice recognition module, the FAQ mapping module matches the corresponding FAQ problem, the scheduling command is generated and issued and executed through a scheduling command generation module in combination with a preset rule, and meanwhile feedback is collected and executed, and the FAQ model and the command mapping rule are finely adjusted through an adaptive model training module. According to the scheme, manual operation is reduced, scheduling efficiency and accuracy are improved, and self-adaptive optimization is supported.

Inventors

• LI JUNYU
• RUAN JUN
• LI SHAOSEN
• ZHANG QIANG
• Qu Fafu
• YANG ZHEN
• WU XINWEN
• JIANG LIANGYONG
• BIN BIN
• QIU GUIYAO
• ZHANG ZHE
• HU MENGLIN
• YANG QINGSHI
• ZHU ZHIJUN
• DU HAOTAO
• HUANG CHANGYU
• SUN HAO

Assignees

• 中国南方电网有限责任公司超高压输电公司昆明局

Dates

Publication Date

20260505

Application Date

20251215

Claims (10)

1. 1. The intelligent voice recognition and dispatching desk fusion method for the trunking communication is characterized by comprising the following steps of: Acquiring voice instruction data input by a user at a trunking terminal, and converting the voice instruction data into corresponding text data; matching the text data, and outputting FAQ problems corresponding to the text data; matching the FAQ problem with a preset scheduling instruction mapping rule to obtain a corresponding scheduling instruction; The scheduling instruction is issued to a cluster terminal, and the cluster terminal executes the scheduling instruction; And collecting execution feedback data generated after the scheduling instruction is completed, constructing a self-adaptive training model, transmitting the execution feedback data into the self-adaptive training model module, and performing incremental fine adjustment updating on the preset scheduling instruction mapping rule based on the execution feedback data.
2. 2. The method for merging intelligent speech recognition and dispatch table for trunking communication according to claim 1, wherein the step of matching the text data and outputting FAQ problems corresponding to the text data further comprises constructing a FAQ mapping model, wherein the FAQ mapping model is a model formed by combining a pre-training language model and a fine tuning model, and the pre-training language model is a BERT fine tuning model.
3. 3. The method for intelligent speech recognition and dispatching desk fusion for trunking communication according to claim 1, wherein when the adaptive training model is updated incrementally, the method is used for optimizing the preset dispatching instruction mapping rule by combining ASR text data and FAQ label data collected from actual speech logs and manual annotation data of trunking communication.
4. 4. The method for intelligent voice recognition and dispatching desk fusion for trunking communication according to claim 1, wherein when the corresponding dispatching instructions are matched based on the FAQ problem and a preset dispatching instruction mapping rule, the FAQ problem is matched with a preset dispatching instruction template, the corresponding type is determined, and the dispatching instructions meeting trunking communication requirements are generated by combining a predefined trunking dispatching strategy.
5. 5. The method for intelligent speech recognition and dispatching desk fusion for trunking communication according to claim 1, wherein the execution feedback data comprises execution success information and execution failure information generated in the process of executing the dispatching instruction by the trunking terminal or dispatching desk, and manual correction information generated by manually correcting the dispatching instruction or the execution result.
6. 6. The method for intelligent voice recognition and dispatching desk fusion for trunking communication according to claim 1, wherein voice instruction data input by a user at a trunking terminal is collected through a microphone array, and preliminary noise reduction processing is performed on the voice instruction data through noise reduction processing in the collecting process.
7. 7. The method for intelligent speech recognition and dispatch platform fusion for trunking communication according to claim 1, wherein the matching processing of the text data further comprises semantic understanding of the text data through a FAQ mapping model, extracting demand information in the text data, comparing the demand information with a preset FAQ problem library, and outputting a corresponding FAQ problem after comparison.
8. 8. The method for intelligent speech recognition and dispatch platform integration for trunking communication according to claim 1, wherein when the adaptive training model performs incremental fine tuning update on the preset dispatch command mapping rule, the incremental fine tuning update operation is repeatedly performed according to a preset time period to continuously adapt to the change of trunking communication scene.
9. 9. The intelligent voice recognition and dispatching desk fusion equipment for cluster communication is characterized by comprising a voice acquisition module, an ASR voice recognition module, a FAQ mapping module, a dispatching instruction generation module, a dispatching execution module and an adaptive model training module; the output end of the voice acquisition module is connected with the input end of the ASR voice recognition module and is used for transmitting the acquired voice instruction data input by the user at the trunking terminal to the ASR voice recognition module; The output end of the ASR speech recognition module is connected with the input end of the FAQ mapping module and is used for transmitting the converted text data corresponding to the voice instruction data to the FAQ mapping module; the output end of the FAQ mapping module is connected with the input end of the scheduling instruction generating module and is used for transmitting the FAQ problem corresponding to the text data obtained by matching to the scheduling instruction generating module; the output end of the scheduling instruction generating module is connected with the input end of the scheduling execution module and is used for transmitting the generated scheduling instruction to the scheduling execution module; The feedback output end of the dispatching execution module is connected with the input end of the self-adaptive model training module and is used for transmitting the collected execution feedback data after the cluster terminal or the dispatching desk executes the dispatching instruction to the self-adaptive model training module; the output end of the self-adaptive model training module is respectively connected with the FAQ mapping module and the scheduling instruction generating module and is used for updating the model in the FAQ mapping module and a preset scheduling instruction mapping rule based on the execution feedback data.
10. 10. A computer readable storage medium, wherein computer executable instructions are stored on the computer readable storage medium, and when the computer executable instructions are executed by a processor, the method for merging intelligent voice recognition and dispatching desk for cluster communication according to claim 1 can be realized.

Description

Method and equipment for fusing intelligent voice recognition and dispatching desk for trunking communication Technical Field The application relates to the crossing field of professional wireless communication and intelligent voice recognition technology, in particular to a method and equipment for fusing intelligent voice recognition and dispatching desk for trunking communication. Background The technical scheme is in the cross field of professional wireless communication and intelligent voice recognition technology, and focuses on integrating artificial intelligent voice processing capability into a professional cluster communication system. From the technical background, the scheduling operation of the traditional trunking communication system is highly dependent on manpower, and has the problems of response delay and large operation load, and although the general speech recognition (ASR) technology is mature, the application of the general speech recognition (ASR) technology in a professional scheduling scene is insufficient, and particularly, the problems of recognition accuracy, professional term suitability and real-time performance in a complex environment are faced. The related technology development shows the trend of wideband, intellectualization and fusion, a trunked communication system is evolving from narrowband voice to wideband multimedia, a 5G technology provides a more reliable transmission basis for the trunked communication system, and meanwhile, an ASR technology is optimized through a deep learning model and combines the technologies of noise reduction, hot word enhancement and the like, so that the practicability of the trunked communication system in a professional scene is gradually improved. In the aspect of the current state of the art, in the existing trunking communication system, a dispatching desk mainly performs terminal dispatching through manual operation or simple rule instructions, and in the field of intelligent voice recognition, an existing ASR (AutomaticSpeechRecognition) system can convert voice into text, but is generally applied to a question-answering system, a voice assistant or a general voice recognition scene. Disclosure of Invention In order to solve the technical problems that ASR and scheduling instructions are directly associated and the capability of FAQ model driving scheduling decision and self-adaptive updating is lacking in the prior art, the invention provides a method and equipment for fusing intelligent voice recognition and scheduling stations for trunking communication. The application provides a cluster communication-oriented intelligent voice recognition and dispatching desk fusion method, which comprises the following steps of obtaining voice instruction data input by a user at a cluster terminal, converting the voice instruction data into corresponding text data, carrying out matching processing on the text data, outputting FAQ problems corresponding to the text data, generating corresponding dispatching instructions by the FAQ problems and preset dispatching instruction mapping rules, sending the dispatching instructions to the cluster terminal, executing the dispatching instructions by the cluster terminal, collecting execution feedback data generated after the dispatching instructions, constructing an adaptive training model, transmitting the execution feedback data into the adaptive training model module, and carrying out incremental fine adjustment updating on the preset dispatching instruction mapping rules based on the execution feedback data. Further, the application also provides that the text data is matched, and the FAQ problem step corresponding to the text data is output, further comprises the step of constructing a FAQ mapping model, wherein the model in the FAQ mapping module is a model formed by combining a pre-training language model and a fine tuning model, and the pre-training language model is a BERT fine tuning model. Furthermore, the application also provides that when the self-adaptive model training module performs incremental updating, ASR text data and FAQ label data collected from the actual voice log and the manual annotation data of the trunking communication are combined together to be used for optimizing a preset scheduling instruction mapping rule. Further, the application also provides that when a corresponding scheduling instruction is generated based on the FAQ problem and a preset scheduling instruction mapping rule, the FAQ problem is matched with a preset scheduling instruction template, the corresponding type is determined, and the scheduling instruction meeting the requirement of trunking communication is generated by combining a predefined trunking scheduling strategy. Furthermore, the application also provides that the execution feedback data comprises execution success information and execution failure information generated in the process of executing the scheduling instruction by the cluster term