Automatic Spoken Language/Dialect Recognition

With the trend of globalizing the communication technology and providing services to a wide, multilingual public, the ability of machines to distinguish between languages has become increasingly important. Applications such as multilingual spoken dialog systems, database‐ or archive‐search and retrieval systems, as well as human communication systems (call‐routing, automatic
translation) face the problem of potential presence of multiple (and possibly unknown) languages/dialects at the input. Automatic spoken language/dialect recognition is the process by which language of dialect of a digitized speech utterance is recognized by a computer. The project is going to develop innovative approaches to spoken language/dialect recognition with the state‐of‐the‐art
machine learning techniques. The main issues to be investigated include available cues for spoken language/dialect identification, feature extraction via learning, and appropriate classification and language/dialect modelling techniques. In particular, the aforementioned research issues would be investigated by taking real environmental factors, e.g., noise and mismatch conditions, into account for robustness. Based on the proposed approaches, a spoken language/dialect identification prototype of high performance would be established for a real world application, e.g., real‐time language/dialect detection from an audio stream
associated with films or video clips for indexing/retrieval.

In order to take this project, it is essential to have good machine learning and speech signal processing background knowledge as well as excellent programming skills.