An Efficient System for Audio-Based Sign Language Translator Through MFCC Feature Extraction

This research study discusses about converting speech to text and text to visual Indian Sign Language (ISL) with MFCC feature extraction that uses Mel Frequency Cepstral Coefficients (MFCC), Gaussian Hidden Markov Models (GHMM) to convert speech to text and Natural Language Toolkit (NLTK). The proposed system follows a traditional speech recognition pipeline, including preprocessing, feature extraction, model training, and recognition. The recognized speech in the text format is processed through the NLTK framework to preprocess the text. The preprocessed text is considered as keywords, which are further used to display the output as a video containing the sign language for the spoken keywords. The system achieves high accuracy in recognizing spoken words and translating them into their corresponding words using the GHMM model. The system's real-time implementation and high accuracy make it a promising tool for facilitating communication and enhancing accessibility. The proposed system's advantages over existing translator systems include its reliance on widely-used MFCC and GHMM algorithms, real-time performance, NLTK speech processing, and high recognition accuracy.