Voice wakeup method and device

Abstract

The invention belongs to the field of voice wakeup, and discloses a voice wakeup method and device. The voice wakeup method comprises the steps that obtained currently input voice is subjected to voice characteristic extraction; according to the extracted voice characteristics, whether or not wakeup words exist in the currently input voice is determined according to a keyword detection model builtin advance, wherein keywords in the keyword detection model at least comprises preset wakeup words; when it is determined that the wakeup words exit in the currently input voice, whether or not instruction words exist in a preset time threshold value is further judged; the keywords in the keyword detection module at least comprise preset instruction words; the current voice wakeup mode is judgedaccording to whether or not the instruction words exit within the preset time threshold value; corresponding feedback is given according to the voice wakeup mode, and a user interacts with a voice system more naturally.

Description

technical field [0001] The invention belongs to the field of voice wake-up, in particular to a voice wake-up method and device. Background technique [0002] With the development of voice technology, many smart devices can interact with users through voice. The voice interaction system of the smart device completes the user's instructions by recognizing the user's voice. In traditional voice interaction, the user usually manually activates the voice, such as pressing the record button, to perform voice interaction. In order to enable users to cut into the voice more smoothly and simulate the behavior of calling each other at the beginning of human-to-human interaction, a voice wake-up function is designed. [0003] At present, the existing voice wake-up methods mainly include: before performing voice interaction with the smart device, the user first needs to say a wake-up word, which can be preset for the smart device. The wake-up module of the voice interaction system de...

AI Technical Summary

Problems solved by technology

[0004] In the above-mentioned existing voice wake-up methods, since the user needs to perform voice wake-up before each command is issued, that is, it is necessary to say the wake-up word first, and then issue the voice of the command, so that after the voice interaction system completes an command operation, it needs to perform a new command. Keyword detection wastes system resources: and for users, they need to say a wake-up word every time before issuing an instruction. The voice wake-up method is cumbersome and the user experience is poor.

[0005] In response to this defect, Google launched the One-shot (one-language joint control) method, but it also has defects in wake-up. The shortcoming of the One-shot wake-up method is that the voice system does not give timely feedback. When the user finishes speaking a sentence, it may not be recognized. Intention, you need to say the whole sentence again. The disadvantage of the ordinary wake-up method is that you have to say a wake-up word every time you say an instruction, but the advantage is that there is timely feedback

Images