Text this: Reordering Features with Weights Fusion in Multiclass and Multiple-Kernel Speech Emotion Recognition