Fig. 4From: Small-parallel exemplar-based voice conversion in noisy environments using affine non-negative matrix factorizationEstimation of parallel dictionary using a speaker transformation matrixBack to article page