Voice wake-up method based on decoder and related equipment thereof

Abstract

The embodiment of the invention belongs to the field of artificial intelligence, is applied to the field of smart communities, and relates to a voice wake-up method based on a decoder and related equipment thereof.The voice wake-up method comprises the steps that an initial voice wake-up model is received, an output layer of the initial voice wake-up model comprises a mute pdf node, a useless pdf node and a wake-up word pdf node, the mute pdf node corresponds to a pdf corresponding to the mute, and the useless pdf node corresponds to a pdf corresponding to the useless pdf node; the wakeup word pdf node corresponds to the pdf corresponding to the wakeup word, and the useless pdf node corresponds to the pdf except the pdf corresponding to the wakeup word and the pdf corresponding to the mute; performing alignment operation on the voice training data set to obtain a voice frame training data set, and training the initial voice wake-up model to obtain a target voice wake-up model; inputting the to-be-recognized voice into the target voice wake-up model to obtain a target pdf probability; inputting the target pdf probability into a decoder to obtain a decoding result; and if the decoding result is a wake-up word, executing a wake-up operation. The target voice wake-up model can be stored in the block chain. According to the invention, the accuracy of voice wake-up is improved.

Description

technical field [0001] The present application relates to the technical field of artificial intelligence, and in particular, to a decoder-based voice wake-up method and related devices. Background technique [0002] Voice wake-up systems have been widely used in many smart devices, such as smart speakers. It is the forefront of voice interaction. Unlike other more complex voice interaction functions, voice wakeup will always be performed in the background, detecting in real time whether the user has said the wakeup word. Only when it is detected that the user has spoken the wake-up word will the subsequent more complex voice interaction logic be triggered. [0003] In the existing voice wake-up system, after the wake-up word is determined, a large amount of wake-up word speech is generally required for model training. This brings great inconvenience, because the recording data of this specific word is hard to come by, and in the ordinary voice data set, the proportion of ...

AI Technical Summary

Problems solved by technology

This method is also flawed, and it is difficult to distinguish the situation where the wake-up words are reversed. For example, the wake-up word is "Hello Xiaohai", but the user will wake up when the user says "Hello Xiaohai". Based on the above two situations, the existing voice The accuracy of waking up the system is not high

Images