Survey And Anlysis On Automated Speech Reading Techniques On Various Languages Using Deep Learning

Abstract

Speech reading is a process of understanding the language with the help of lip movements with or without sound especially useful for hearing impaired, or aged people. In the presented paper, we explore variation of lip texture features for recognition of visemes.  Aiding the cause, Deep learning, and Machine learning sign language recognition approach is used for Regional Indian Language word recognition and achieved high accuracy at an estimate of computation. A study of comparisons on automated Speech reading approaches is presented in this paper with the main focus being with regard to deep learning and related methodologies with promising result for both feature extraction and classification schema for lip-reading sentences. It might investigate the efficiency of different neural network topologies such as (CNNs), (RNNs), in lip reading exercises, or attentional strategies. RNN is a recurrent  kind of artificial neural network that utilizes time series data or sequential data. A neural network type called a convolutional network architecture, or CNN or ConvNet, is particularly adept at processing input with a grid-like architecture, like an image. Work focuses on benefits of Attention-Transformers and Temporal Convolutional Networks for classification compared to Recurrent Neural Networks. Paper includes analysis of well-organized lip reading techniques for varied linguistic systems. Comparison of various Algorithms, discussion of the trade-offs between many algorithms, adopted methodologies performance and limitations have been formulated. The study focuses more on Machine Learning & Deep Learning technologies regardless of particular application areas for speech reading domain.