WAKE-UP MODEL GENERATION METHOD, SMART TERMINAL WAKE-UP METHOD, AND DEVICES

摘要

A wake-up model generation method, a smart terminal wake-up method, and devices, which belong to the technical field of voice wake-up. The wake-up model generation method comprises: labeling the start time and end time of each wake-up word contained in wake-up word audios in a sample audio set, so as to obtain a labeled wake-up word audio, the time length of the wake-up word audio being not fixed (101); adding noise to the labeled wake-up word audio by using a negative sample audio containing background noise, so as to obtain a positive sample audio (102); extracting a plurality of audio frame features from the positive sample audio and the negative sample audio respectively, and labeling the frame labels of the positive sample audio and the negative sample audio, so as to obtain a plurality of audio training samples (103); and training a recurrent neural network by using the plurality of audio training samples, so as to generate a wake-up model (104). In the embodiments, a recurrent neural network of a variable-length input is used to perform model training, so that an operation of manually intercepting samples can be avoided, facilitating improvement of the wake-up effect of a smart terminal.