Spoken Language Identification System for Kashmiri and Related Languages Using Mel-Spectrograms and Deep Learning Approach

Language identification, being the front-end for various natural language processing tasks, plays an important role in language translation. Owing to this, the focus has been on the field of speech recognition involving the identification & recognition of languages by a machine. Spoken language identification is the identification of language present in a speech segment despite its size (duration & speed), ambiance (topic & emotion), and moderator (gender, age, demographic region). In this model, six languages have been identified that include five officially spoken languages of Jammu and Kashmir (Kashmiri, Urdu, Dogri, Hindi, and English) along with Ladakhi language by applying Convolutional Neural Network (CNN) approach. The data is recorded from TV channels of languages (Kashmiri, Ladakhi, Urdu, and Dogri) and standard corpora IIIT-H and VoxForge for Hindi and English respectively. Data is pre-processed by removing background noise and segmented into segments of 5 seconds each. Each segment is converted into Mel-spectrograms by applying Fast Fourier Transformations and at last, CNN serves as a classifier. Experimental results showed that an average training and testing accuracies of 100% have been achieved by running the model at 100 epochs.