We are in the era of personal assistant devices. Over the last few years the number of users of Alexa, Siri, Cortana etc. has grown enormously. Along with the recent trends, the goal of this project is to localize a sound source inside a room with one microphone. Student can choose an arbitrary wake word that should be localized. Localization is done by compressed sensing in the frequency domain. This way we can play a hide and seek game with our personal assistant – after calling the assistant it should be able to localize us inside a known room. There are a few topics to be explored: what type of wake word gives good results, how to encode different positions in the room and can we build a system that is robust to noise?

Profile: Computer science and electrical engineering

Prerequisites: Signal processing and matlab/python programming

Context: Theory (40%), design (40%), simulations (20%)