Language Identification in Audio with Mel-Frequency Cepstral Coefficients and Multi-Head Long Short-Term Memory

Language barrier is one of the real-life problems faced by many individuals when they travel to different places of the world. To build a conversation between two individuals language plays a vital role in communicating with each other. Due to this, we are unable to express ourselves and end up in helpless situations. In order to solve this problem, we have language detection which comes in handy all over the world. Our main goal of this research is to create a strong and reliable model for identifying languages in audio signals. Precise language identification is necessary for a range of applications, including speech recognition systems, language translation services, and content filtering. We utilize Mel-Frequency Cepstral Coefficients (MFCC) as features in our model. For language classification, we employ a Multi-Head Long Short-Term Memory (LSTM) structure. The performance of the proposed model is evaluated using accuracy, F1-score, recall, and precision. Our model achieved an accuracy of 96.88%.