Segmental and Supra Segmental Feature Based Speech Recognition System for Under Resourced Languages