An automatic correction system of singing intonation based on deep learning

Abstract

In order to solve the problems of low accuracy and slow correction speed in traditional singing intonation correction system, an automatic singing intonation correction system based on deep learning is proposed. In the hardware, floating-point DSP and TDSP-TF984 chip are selected as the core chips of automatic correction processor of singing intonation. The data input module and parameter calculation module of singing intonation are designed to improve the singing intonation data collector. In the software, the group delay estimation method is used to collect the singing intonation signal, and the deep learning algorithm is used to decompose the false component of the singing intonation signal. The autocorrelation function and characteristic distribution operator of the singing intonation signal are obtained to realise the singing intonation signal correction. The experimental results show that the highest accuracy of the proposed system is about 97.8%, and the shortest correction time is about 1 s.