Continuous telugu speech recognition on T-LPC and DNN techniques

Abstract

This paper deals with the basic application speech recognition. There are many languages in the world but one of the regional language is Telugu. Recognition of this language helps in many applications for 8 crores of people stay in AP and Telangana states. Recognition is done by recording the speech signals and database creation. Pre-processing is done by 2 stage DNN (seep neural networks) where denoising, framing is done. The preprocessed signal features are extracted using TLPC(teager energy operator linear prediction filter). The features extracted are classified using DNN which generates adequate results. The results are obtained for continuous speech of Telugu language.