Identification of Weakly Pitch-Shifted Voice Based on Convolutional Neural Network

<div>Detection rates of strongly pitch-shifted voice. (a-<i>x</i>) Voice pitch-shifted by Audition. (b-<i>x</i>) Voice pitch-shifted by GoldWave. (c-<i>x</i>) Voice pitch-shifted by Audacity; <i>x represents subfigures in same column</i>. (<i>y</i>-1) TIMIT for training and TIMIT for testing. (<i>y</i>-2) TIMIT for training and UME for testing. (<i>y</i>-3) UME for training and TIMIT for testing. (<i>y</i>-4) UME for training and UME for testing; <i>y represents subfigures in same row.</i></div>

International Journal of Digital Multimedia Broadcasting

fig6

Figure 6

Figure 6: Identification of Weakly Pitch-Shifted Voice Based on Convolutional Neural Network 