Correctly identifying the language spoken is essential for the success of speech recognition technologies. Although there have been major advancements in audio transcription systems for various Western linguistic systems, the progress in this area for Indian languages has been lagging. Several artificial intelligence methods have been explored for this task; however, there remain many areas that can be improved. In this study, we introduce a methodology to identify nine widely spoken Indian languages: Bengali, Hindi, Marathi, Gujarati, Tamil, Telugu, Malayalam, Kannada, and Urdu. We extracted Mel-Frequency Cepstral Coefficients (MFCCs) from the speech as features to experiment with four deep learning models: a Convolutional Neural Network (CNN), a Long Short-Term Memory (LSTM), a Bidirectional LSTM (BiLSTM), and a Multi-Layer Perceptron (MLP). The MLP model achieved the best performance with a 97% score in accuracy, precision, recall, and F1-score. The proposed method demonstrated both efficiency and speed, as evidenced by its outstanding ability to identify languages from very short audio clips.