Audio Segmentation Method based on Attention Mechanism

摘要

Provided is an audio split method based on attention mechanism. According to one embodiment of the present invention, the audio split method obtains a mapping relationship between an inputted text and an audio spectrum feature vector for generating an audio signal, which is automatically synthesized using the inputted text, and splits the inputted audio signal using the mapping relationship. Accordingly, through the audio splitting using the attention mechanism, the audio split method may significantly reduce effort, time, and cost while having high quality.