Sign language is a complete natural language used by deaf and dumb people. It has its own grammar and it differs with spoken language to a great extent. Since people without hearing and speech impairment lack the knowledge of the sign language, the deaf and dumb people find it difficult to communicate with them. The conception of system that would be able to translate the sign language into text would facilitate understanding of sign language without human interpreter. This paper describes a systematic approach that takes Indian Sign Language (ISL) video as input and converts it into text using frame sequence generator and image augmentation techniques. By incorporating these two concepts, we have increased dataset size and reduced overfitting. It is demonstrated that using simple image manipulation techniques and batch of shifted frames of videos, performance of sign language recognition can be significantly improved. Approach described in this paper achieves 99.57% accuracy on the dynamic gesture dataset of ISL.

We present a novel approach for improving the performance of an End-to-End speech recognition system for the Gujarati language. We follow a deep learning based approach which includes Convolutional Neural Network (CNN), Bi-directional Long Short Term Memory (BiLSTM) layers, Dense layers, and Connectionist Temporal Classification (CTC) as a loss function. In order to improve the performance of the system with the limited size of the dataset, we present a combined language model (WLM and CLM) based prefix decoding technique and Bidirectional Encoder Representations from Transformers (BERT) based post-processing technique. To gain key insights from our Automatic Speech Recognition (ASR) system, we proposed different analysis methods. These insights help to understand our ASR system based on a particular language (Gujarati) as well as can govern ASR systems’ to improve the performance for low resource languages. We have trained the model on the Microsoft Speech Corpus, and we observe a 5.11% decrease in Word Error Rate (WER) with respect to base-model WER.