Wake-up voice determination method, device, equipment and medium

Abstract

The invention provides a wake-up voice determination method and device, equipment and a medium, and belongs to the technical field of audio processing. According to the technical scheme provided by the embodiment of the invention, after whether a to-be-processed voice is a wake-up voice or not is preliminarily determined through primary classification, secondary classification verification is carried out under the condition that the to-be-processed voice is determined to be the wake-up voice, false wake-up can be avoided through two times of classification, and the wake-up accuracy is improved. And during secondary verification, feature extraction is performed based on the classification result of the primary classification, and the target length is obtained in an interpolation mode, so that the length of the voice features of feature extraction is not limited, the situation that the speech features include mute segments or wake-up words are incomplete can be effectively avoided, and the wake-up accuracy and success rate are further improved.

Description

technical field [0001] The present disclosure relates to the technical field of audio processing, and in particular to a method, device, device and medium for determining a wake-up voice. Background technique [0002] In recent years, with the continuous development of audio processing technology, intelligent voice interaction systems such as smart speakers and vehicle-mounted voice interaction systems have become popular. In order to reduce user operations, a voice wake-up function is provided. By recognizing the collected voice, it is determined Whether it is a wake-up voice, and then the voice wake-up of the device can be realized. [0003] In related technologies, the wake-up voice determination method usually includes: extracting features of the voice to be processed, obtaining fixed-length voice features, and inputting them into a convolutional neural network (Convolutional Neural Networks, CNN) network for classification. [0004] The length of the wake-up word read ...

AI Technical Summary

Problems solved by technology

[0004] The length of the wake-up word read by the user is generally not fixed. The above method limits the length of the speech features input into the CNN network. The input of fixed-length speech features into the network for discrimination may result in some silent segments (that is, parts that are not the user's voice) being sent. , or part of the wake-up word is truncated, resulting in inaccurate determination of the wake-up voice, resulting in a low wake-up success rate, and prone to false wake-ups, with low accuracy

Images