Wake-up model generation method and intelligent terminal wake-up method and device

Abstract

The invention discloses a wake-up model generation method, an intelligent terminal wake-up method and device. The invention discloses a wake-up model generation method, belongs to the technical fieldof voice wake-up, and the wake-up model generation method comprises the following steps: marking the start and stop time of each wake-up word contained in wake-up word audios in a sample audio set, and obtaining marked wake-up word audios, and the time length of the wake-up word audios is not fixed; adding noise to the marked wake-up word audio by using a negative sample audio containing background noise to obtain a positive sample audio; respectively extracting a plurality of audio frame features from the positive sample audio and the negative sample audio, and performing frame label labelingon the positive sample audio and the negative sample audio to obtain a plurality of audio training samples; and training the recurrent neural network by using the plurality of audio training samplesto generate a wake-up model. According to the embodiment of the invention, the model training is carried out by using the variable-length input recurrent neural network, so that the operation of manually intercepting the sample can be avoided, and the wake-up effect of the intelligent terminal is improved.

Description

technical field [0001] The present invention relates to the technical field of data security, in particular to a wake-up model generation method, a smart terminal wake-up method and a device. Background technique [0002] At present, voice wake-up has a wide range of applications, such as robots, mobile phones, wearable devices, smart homes, and vehicles. Different smart terminals will have different wake-up words. When the user speaks a specific wake-up word, the smart terminal can switch from the standby state to the working state. Only when the state switching is completed quickly and accurately can the user use it almost unconsciously. Other functions of the smart terminal, therefore, it is very important to improve the wake-up effect. [0003] In the prior art, a wake-up technology based on a neural network is mainly used for waking up an intelligent terminal. In the data preparation stage, it is necessary to manually intercept the positive sample data to a fixed time...

AI Technical Summary

Problems solved by technology

In the data preparation stage, it is necessary to manually intercept the positive sample data to a fixed length of time t, and the duration of recording the wake-up words cannot exceed this length of time t, which will greatly increase labor costs, and cannot be used for slow-speed wake-up voices. In addition, due to the short time of the wake-up word, the training of the neural network is insufficient, which will eventually affect the wake-up effect of the smart terminal; in addition, in the terminal wake-up phase, because the neural network needs to process the time in the terminal memory each time Audio with a length of t, so that there will be a large amount of repeated data to be processed between two adjacent time lengths t, thus increasing the calculation time and power consumption of the terminal

Images