Phoneme detection method and device based on multiple tasks

Abstract

The invention discloses a phoneme detection method based on multiple tasks. The phoneme detection method comprises the following steps: step A1) training a phoneme detection model; a2) obtaining a voice sequence to be detected; a3) segmenting the voice sequence into a plurality of basic sub-sequences; a4) moving the endpoints of the basic sub-sequences to obtain a group of transformation sub-sequence sets; a5) inputting all the transformation sub-sequences into a phoneme detection model to obtain predicted phonemes and corresponding confidence coefficients; a6) taking the transform sub-sequence with the highest confidence as a new basic sub-sequence; and a7) judging whether the basic sub-sequence meets a termination condition, if so, obtaining and outputting a phoneme detection result anda phoneme position, and if not, returning to the step a4). The technical problems that phoneme recognition and phoneme alignment tasks cannot be completed at the same time, the phoneme alignment accuracy is low, and the phoneme recognition task and the phoneme alignment task cannot share a learning result are solved.

Description

technical field [0001] The invention relates to the field of data intelligence, in particular to a multi-task-based phoneme detection method and device. Background technique [0002] With the development of deep learning technology, deep speech processing technologies such as speech recognition, voiceprint recognition, speech synthesis, and speech emotion analysis continue to break through. As the smallest speech unit divided by the natural properties of speech, phonemes play a very important role in deep speech processing and are the basis of most speech processing. At the same time, phonemes are of great significance to the rapid response of the deep speech processing system in actual scenarios. At the same time, there are very few speech databases containing phoneme alignment information in the existing data sets, and are limited by the phoneme definition specifications of the database itself. It is easy to encounter situations where the phoneme definition specifications...

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to provide a phoneme detection method and device based on multi-task, which is used to solve the problem that the phoneme recognition and phoneme alignment tasks cannot be completed at the same time, the accuracy of phoneme alignment is low, and the phoneme recognition task and phoneme alignment task cannot share learning results. question

Images