CNN-RNN Hybrid Model to Classify a Local Language Slangs using Spectral features

Voice is significant in the environment of Artificial Intelligence. Many applications, such as speech processing, voice recognition, emotion recognition, and natural language processing, use voice categorization. Tamil is regarded as one of the most classical languages in the world. It has a number of colloquial and regional slangs, including Chennai, Kovai, Madurai, and Nellai. Classifying slang is a difficult task. The dataset includes sample audio speech from several Tamil slangs such as Chennai, Kovai, Madurai, and Nellai, as well as a variety of age and gender groups. Till now there was no approach to classify Tamil slangs. The proposed work aims to create Novel Hybrid Model to classify Tamil slangs by extracting Spectral features. The collected audio files are converted to spectrogram images. CNN are well-suited for extracting features from spectrograms, which are commonly, used representations of audio data. RNNs are useful for capturing temporal dependencies in sequential data, especially in forms such as GRU and LSTM. The CNN’s derived features are fed into a recurrent neural network, such as an LSTM or GRU, to extract temporal dependencies and identify Tamil slang. Thus a powerful Novel Hybrid approach was developed and implemented for Tamil slang. The blended combination of CNN with GRU gives 86% of accuracy whereas the blended combination of CNN with LSTM produces 91% of accuracy in classification of Tamil slangs.