Language identification is a procedure in speech processing that aims to automatically identify the language or dialect spoken in an audio recording. Language identification is used daily in a variety of crucial tasks including multilingual translation and emergency call routing. Over the past decade, deep neural networks have redefined a variety of state-of-the-art results in speech processing applications. Recently, there has been a growing interest in using particular deep neural networks, recurrent neural networks, because of their advantages in predicting unsegmented time series and sequences with changing contextual windows. The objective of this project is to investigate the application of Long Short-Term Memory (LSTM) recurrent neural network architectures to language identification.

I previously worked on a machine learning project in the CSAIL Medical Vision Group with Professor Polina Golland and Dr. Christian Wachinger on the feature-based segmentation of parotid glands in CT scans of the head and neck. I also developed autonomous navigation and vision systems as a contestant in the MASLAB IAP Robotics Competition.