[2210.14636] Fast Yet Effective Speech Emotion Recognition with Self-distillation