Knowledge Distilled Ensemble Model for sEMG-based Silent Speech Interface