Fig. 1From: Emotional voice conversion using neural networks with arbitrary scales F0 based on wavelet transformLog-normalized F0 (a) and interpolated log-normalized F0 (b). The red curve target F0; the blue curve source F0Back to article page