Fig. 2From: Small-parallel exemplar-based voice conversion in noisy environments using affine non-negative matrix factorizationConstruction of parallel dictionariesBack to article page