Mapping the Speech Signal onto Electromagnetic Articulography Trajectories Using Support Vector Regression