Voice processing method and device based on cloud

Abstract

The present invention proposes a voice processing method and device based on the cloud, wherein the method includes: acquiring the voice to be recognized collected by the terminal; performing voice recognition on the voice to be recognized; during the voice recognition process, simultaneously detecting whether the voice to be recognized is end; when it is detected that the speech to be recognized ends, the recognition result of the speech to be recognized is fed back to the terminal. During the speech recognition process, it is detected whether the speech to be recognized ends at the same time, so that the accuracy of speech recognition can be guaranteed. Furthermore, the speech recognition performed by the cloud has nothing to do with the local terminal, so it has nothing to do with the hardware platform, there is no migration and development, and it is not limited by the local resources of the terminal. Therefore, the speech recognition model used in the cloud can be made very complex and fine, so that It can improve the anti-jamming performance of the system. In addition, the speech recognition model used in the cloud has been tailored for different applications, and there is no need to retrain the model, which simplifies the amount of development.

Description

technical field [0001] The invention relates to the technical field of speech processing, in particular to a cloud-based speech processing method and device. Background technique [0002] With the continuous development of terminal equipment, in order to simplify the user's operation, the user can input a piece of voice in the terminal, and then the terminal can recognize the voice and convert it into the corresponding text, so that the user can perform follow-up according to the converted text. Operations, such as surfing the Internet, shopping, etc. In the prior art, the terminal locally recognizes the voice through a voice activity detection (Voice Activity Detection, VAD) model, specifically, the time domain energy signal of the audio is used to determine the start point and the end point of the voice. [0003] In this way, limited by the terminal's local memory, storage space, and computing power, the terminal's local VAD model can only use a simple model structure, re...

AI Technical Summary

Problems solved by technology

In addition, the speech recognition model used in the cloud has been tailored for different applications, and there is no need to retrain the model, which simplifies the amount of development and is used to solve the existing limitations of the terminal's local memory, storage space, and computing power. The model can only use a simple model structure, resulting in poor anti-interference performance against noise, resulting in low recognition accuracy

Images