Automatic data collecting and labeling method

Abstract

The invention discloses an automatic data collecting and labeling method. The method comprises the steps of setting a collection scene; adding a preset chirp signal before ripping of an original voicesignal; performing sound recording at a first preset number of preset collecting angles to obtain ripped signals at different angles; according to the preset chirp signal, processing the ripped signals to obtain processed ripped signals, wherein the processed ripped signals are a first preset number of ripped signals corresponding to a preset angle range; and labeling the ripped signal at any angle, and copying the labeling to serve as labeling of the ripped signals at other angles. According to the method, the voice data collecting and labeling efficiency can be improved and the error causedby manual operation is reduced.

Description

technical field [0001] The invention relates to the technical field of voice recognition, in particular to a method for automatic data collection and labeling. Background technique [0002] In the field of speech recognition, whether it is traditional speech recognition based on template matching methods (such as based on hidden Markov model) or speech recognition based on neural networks, labeled data are required for training or verification. [0003] At present, data recording and labeling are usually performed manually. First, set the scene, then record one piece for each angle, and label one piece. The collection and labeling efficiency of voice data is low. Contents of the invention [0004] In order to solve the above problems, the present invention provides a method for automatic data collection and labeling, including: [0005] Set the collection scene; [0006] Add a preset chirp signal before ripping the original voice signal; [0007] Recording is performed...

AI Technical Summary

Problems solved by technology

Leading to low efficiency of voice data collection and labeling

Images