AUTOMATIC SPEECH TO TEXT USING ACOUSTIC MODELLING AND DEEP LEARNING
Author(s):
Dr.R.Jayalakshmi, Sagar D, Somalaraju Yashwanth Varma, Syed Mohammad Musharraf, TEJASWIN R M, N Venkata Vamshi, Yelugubanti Manikanta
Keywords:
Speech recognition system, speech processing, Feature extraction techniques, modelling techniques, applications of SRS, NLP and ASR system, Word Error Rate(WER), CNN and RNN .
Abstract
Language is the most important means of communication and speech is its main medium. Many research activities are being conducted on Automatic Speech Recognition. But, ASR systems have a major drawback in their performance i.e., efficiency. Improving the efficiency in an ASR system is quite difficult. Currently, research is being carried out to the finding of the next state of the world using Hidden Markow Model (HMM). Our Study concludes that for ASR systems, Deep Learning techniques is a more suitable application, because it increases the efficiency of the whole process. We are going to represent our work on building a speaker independent, large vocabulary continuous speech
recognition system for English and Hindi. Our Study concludes that for ASR systems, Deep Learning techniques is a more suitable application, because it increases the efficiency of the whole process. Based on the conclusion, we will utilize readily available language models to build an offline live Speech recognition to Text System and Translation. Using offline speech recognition toolkits like VOSK and Kaldi an offline model is developed. Argos translate an open-source library is used to translate English speech to Hindi text. Raspberry Pi 4 is used to implement the offline speech recognition module
Article Details
Unique Paper ID: 159982
Publication Volume & Issue: Volume 9, Issue 12
Page(s): 966 - 970
Article Preview & Download
Share This Article
Join our RMS
Conference Alert
NCSEM 2024
National Conference on Sustainable Engineering and Management - 2024