Supervision: Benjamin Ricaud

Project type: Semester project (master) Master thesis

Finished

Rooms with wide dimensions present a modal behaviour due to stationnary waves at low frequencies. This behaviour may have prejudicial effect on the music rendering quality in this frequency range. The room modal behaviour can be characterized by:

• a set of resonances at given discrete frequencies, with a given quality factor and amplitudes, observable on room frequency responses at a given source/microphone position,

• a given time-response resulting from the time constant of each resonance, observable on sprectograms (time-frequency-amplitude representation) at a given source/microphone position,

• a specific spatial distribution of sound pressure, characterized by nodes and antinodes within the geometry of the room.

Although it is possible to assess the frequency- and time-domain responses of a room with a limited number of microphones, the full characterization of the spatial distribution is accessible only at the cost of a fine microphone sensing, which may last hours in a room with standard dimensions (say 5 m x 4 m x 3m).

The requested work will investigate the possibility to employ compressive sensing techniques in a view to allowing an accurate imaging of the room modes distribution with a limited number of microphone, thus sparing hours of measurement work (and processing) to the practitioners. It will mainly consist in applying such techniques to a given room mode scenario, of simple geometry (shoebox semi-reverberant room) to allow direct comparison with analytical models.

Context: theory (60%), numerical model (40%)

Prerequisites: Signal processing, Matlab programming