Our sound localization framework should be adapted to speech signals to enhance human-machine interaction. In order to do this, the system should be able to distinguish between speech signals and ambient noise. Currently, the whole sound mixture that is captured by the microphones is taken into account to establish the sound localization map.
Implementation of a signal detector that detects and isolates speech signals from a sound mixture. Other ambient sounds should be considered if they surpass a certain threshold level as they may also contain valuable localization information about the environment. The results of the filtering should enable a service robot to focus on a person that tries to interact with it.
To measure the performance of the signal detector, the filter should be implemented on a service robot and testted with various sound mixtures in an office environment.
This thesis is a part of the HeRBiE project.
- Working knowledge of C++
- Interest in digital signal processing
- Interest in robotics