Cross speaker forward mapping for accent conversion 

Sandesh Aryal and Ricardo Gutierrez-Osuna

o Gu

tierrez-Osuna

       Synthesis samples

    L1 L2 ANS CSFM L1n
S1
S2
S3
S4
S5

L1: original L1 utterances, resynthesized from MFCCs
L2: original L2 utterances, resynthesized from MFCCs
ANS: the articulatory normalization-synthesis method (baseline)
CSFM: the proposed cross-speaker forward mapping  
L1n: original L1 utterances, normalized to the vocal tract length and pitch range of L2