ANALYSIS AND SPEECH RECOGNITION OF UNDER-RESOURCED KANGRI DIALECTWITH ENSEMBLE MODEL

Abstract Speech recognition architectures have been developed for mainstream languages considering the advantages of large amount of resources and limited phonological diversity. These techniques are not effective for dialects with less or no resources available like the Kangri (क ागडी) dialect of Himachal Pradesh, India. A work in progress is presented where an ensemble model is proposed for such dialects. The results for some of the common existing techniques are shown to demonstrate the need for a combination of models