Joint source-target Gaussian Mixture Model based voice conversion algorithms.