Distributed voice monitoring method, device and system, storage medium and equipment

Abstract

The invention provides a distributed voice monitoring method. The distributed voice monitoring method comprises the following steps: acquiring audio stream data belonging to the same machine room; according to a preset review strategy, collecting to-be-reviewed audio data from the audio stream data; inputting the to-be-audited audio data into a pre-trained audio recognition model to obtain a predicted value corresponding to the audio recognition model; and generating an audio machine review result according to the prediction value. According to the method, high-input-output-ratio, high-coverage, low-delay, high-recognition-rate and efficient voice monitoring and auditing can be realized, and the audio content monitoring requirement of high-activity voice social application in a multi-operator hybrid networking deployment network environment can be met.

Description

technical field [0001] The present invention relates to the technical field of voice content monitoring, in particular, the present invention relates to a distributed voice monitoring method, device, system, computer-readable storage medium and computer equipment. Background technique [0002] With the rapid popularization of the Internet, social applications that use voice communication as the main means of communication, making friends, chatting and live broadcasting are very popular among people. However, the huge user base is likely to cause great uncertainty in the voice broadcast content and voice chat content in the form of voice rooms. There are criminals who spread illegal and bad information through voice social applications, which affects the normal operation of the application platform. Therefore, it is necessary to Real-time review and identification of chat and live broadcast content in audio format, so as to promptly crack down on violations and bad behaviors ...

AI Technical Summary

Problems solved by technology

However, the huge user base is likely to cause great uncertainty in the voice broadcast content and voice chat content in the form of voice rooms. There are criminals who spread illegal and bad information through voice social applications, which affects the normal operation of the application platform. Therefore, it is necessary to Real-time review and identification of chat and live broadcast content in audio format, so as to promptly crack down on violations and bad behaviors in voice social applications

[0003] At present, monitoring and auditing can be realized through user reports, housing management inspections, or regular collection of audio data combined with machine recognition, but its limitations are that the monitoring and auditing methods of user reporting and housing management inspections have low coverage, large information lag, and low efficiency of monitoring and auditing , it is easy to cause a malicious event to occur and have a bad social impact; and the method of regularly collecting audio data combined with machine recognition usually adopts a centralized method to provide services. In a multi-operator network deployment environment, all audio data must be sent to The central machine identification system, the central machine identification system is huge and complex, with high construction costs, and a large amount of cost is required for cross-operator traffic charges, and the input-output ratio is low

[0004] Therefore, the existing monitoring and review methods are difficult to meet the voice content review requirements of highly active voice social applications. Low latency, high recognition rate and efficient voice monitoring are very big challenges

Images