Summary of Spoken Indian Languages Classification Using ML and DL

Abstract

Unlike in some other parts of the world, speech recognition technology is legal in the West. It's not to the same degree that this happens in East Asian countries. It's possible that linguistic barriers are a major cause of this chasm. In addition, countries with many languages, such as India, must be taken into account if voice-based language identification is ever going to be practical. The challenge is in finding a technique to clearly and effectively identify the features that may differentiate across languages. The model processes audio data, creating spectrogram images from them before extracting features. Then, the Deep Learning (DL) is employed to streamline the output identification process by emphasizing the most crucial characteristics and attributes. Realizing that a person's vocal signal may be understood or observed was a major inspiration for the concept. This research work employ spectrograms (for visual data) and deep learning techniques to categorize Indic languages inside the IIITH Indic voice database. Finally, a model-based comparative analysis has been conducted by analyzing the accuracy, precision, recall, and f1-score to show that the proposed approach is more robust than existing models.