All-for-One and One-For-All: Deep learning-based feature fusion for Synthetic Speech Detection.