Audio recognition method and device

Abstract

The invention relates to an audio recognition method and device. The method comprises the following steps: obtaining an original audio, adding null data of a first duration before the head of the original audio, and adding null data of a second duration after the tail of the original audio to obtain an expanded audio; taking a third time length of the sum of the first time length and the second time length as a segmentation window, and sequentially performing windowing from the head part of the expanded audio by taking a first step length to obtain a plurality of sub-audios; conducting calculation to obtain time-frequency characteristic sequences of the sub-audios; a neural network obtaining the probability that the sub-audios belong to specific classifications according to the time-frequency characteristic sequence; and comparing the probabilities with a judgment threshold to judge whether the sub-audios are of specific classifications or not. According to the scheme provided by the invention, on the premise of ensuring the interactivity and interaction efficiency of a user, the statistical time is saved, the feedback efficiency is improved, the opening time node statistical function is added, and the precision is improved.

Description

technical field [0001] The present application relates to the technical field of data processing, in particular to an audio recognition method and device. Background technique [0002] With the development of Internet technology, online education and other similar industries are booming, and the number of online learners has increased dramatically. Teachers evaluate students' classroom participation and give feedback to improve teaching effectiveness through subjective feelings and statistics of students' speaking time. [0003] At present, there are several schemes for counting the duration of a user's opening in the prior art. [0004] Based on switch-type opening time statistics, that is, a recording button that can be turned on and off is set on the user client, and the user needs to press the button before speaking. The starting point of this program is to achieve simple and direct time statistics. However, when the audience is a group of young children, the characteri...

AI Technical Summary

Problems solved by technology

Based on the statistics of the user's speaking time, ideally this solution should have the best performance results, but this solution often has the following disadvantages: 1) The speech recognition model generally has high computational complexity and is not suitable for high-concurrency online environments use

Under this scheme, the classification quality depends on manual review, otherwise there will be more errors in the category standard; The model performance is poor

Images