Spoken language identification systems have applications in verbal translation, international security, user interfacing, big data and a multitude of other automatic speech recognition applications. However, determining which features of speech are the most effective means of identification is the subject of ongoing research. A relatively unexplored area in spoken language identification is the glottal flow signal. Glottal flow parameters have already been used to classify other paralinguistic features of speech, and thus there is evidence that different languages have different glottal flow forms. The purpose of this project is to extend research on glottal flow parameterisation by conducting a pilot investigation on glottal flow differences between languages with a focus on its application in spoken language identification. This will be done by developing and reviewing an automatic spoken language identification system using glottal flow parameters of voiced speech.
Undergraduate
A real-time language detection system that works on a PC, and if time permits a raspberry pi.
None
Lab allocations have not been finalised