Project Description
The I-Vector is a state of the art classifier for speaker and language recognition designed by Dr. Najim Nehak. During the summer of 2013, I developed an envelope for this technology to make it more available to the naïve user. That is, parties interested in its end-user functionality, namely Speaker and Language recognition. The desktop application was eventually shipped to interested sponsors and the LLTalk project was born. The aim of the project was for a team of two students in the University of Puerto Rico to research the viability of I-Vector Speaker Recognition on mobile platforms, such as the Android operating system. Several of the project’s challenges included the porting of legacy code to the target platform and its optimization for a mobile ARM architecture. The year-long project culminated with a fully functional prototype that allowed researchers to perform speaker, language, and gender recognition experiments, winning best presentation for the Industrial Affiliates Program of 2014.