Voice-triggered device, method, and computer readable storage medium

Abstract

The invention belongs to the voice trigging field, and discloses a voice-triggered device, method, and a computer readable storage medium; the voice-triggered device comprises a first level waking unit and a main control unit; the first level waking unit comprises the following parts: a voice obtaining module used for obtaining inputted voice information; a first level waking chip used for using afirst level waking voice identification algorithm to identify the voice information, and sending the voice information to the main control unit if the identification result contains a preset waking word; a first connecting interface connected with the main control unit. The main control unit comprises a second connecting interface connected with the first level waking unit, and a second level waking chip used for receiving the voice information, using a second level waking voice identification algorithm to identify the voice information, and determining the waking is finished when the identification result contains the preset waking word; the main control unit then enters a state that waits a voice operation order. The method and device can solve the waking power consumption problems, canreduce the wrong waking rate, and can improve the user experiences.

Description

technical field [0001] The invention belongs to the technical field of voice recognition, and in particular relates to a voice wake-up device, a method and a computer-readable storage medium. Background technique [0002] At present, it is a hot spot in the industry to realize product interaction through voice, such as smart speakers, mobile phones, and tablets. In the smart speaker industry, basically the power supply is directly provided by the adapter, so the design requirements for power consumption are very low, which leads to high power consumption when the smart speaker is waiting for voice wake-up (compared to portable products) Said), about 1W, but this also has its advantages: the back-end wake-up keyword speech model runs on the CPU, and there is no limit to the resource requirements, so it can run a large keyword speech model, which also This makes the wake-up rate and false wake-up rate better. [0003] However, on portable products such as mobile phones and t...

AI Technical Summary

Problems solved by technology

But this also has a disadvantage: due to the limited resources of the DSP chip, the algorithm running in the chip and the wake-up keyword voice model will be greatly restricted, making the wake-up rate and false wake-up rate indicators not very good.

Images