Singing voice detection based on convolutional neural networks
Hong-Ming Huang, Woei-Kae Chen, Chien‐Hung Liu, Shingchern D. You · 2018
This paper investigates various structures of convolutional neural networks (CNN) for singing voice detection. The input features are MFCC (mel-frequency cepstrum coefficients), DFT (discrete Fourier transform) coefficients, and raw PCM samples. The simulation results show that DFT coefficients yields higher detection accuracy, up to 92%.