Abstract:
Speech recognition on various applications is widely used but none of them support Indian regional dialects. American or British dialect of English language is supported by speech engine for voice recognition .In this research paper, an algorithm has been proposed which improves accuracy of voice recognition over Indian regional dialects. It is based on enhanced learning mechanism in which all possible translated texts are stored. Stored data is then used for prediction. For experimentation purpose a voice command interface based Map application (VIMA) using Bing map platform is created. It can be used to find location of places, routes between two places and directions to reach from one point to other. A basic VIMA uses standardized Microsoft desktop speech recognition engine. Proposed algorithm is implemented with basic VIMA and referred as extended VIMA. Accuracy of voice recognition in extended VIMA has been calculated and compared to that of basic VIMA. A path breaking results have been achieved. This research paper studies the variation in accuracy of voice recognition for Indian dialects from all geographical regions of India like East, West, South, North and Centre as mentioned by Indian Railways.
