|
Research: Signal processing algorithms for acoustic sensor networks for improvement of sound quality and speech intelligibility
The quality of sound recordings can be significantly enhanced by using
multiple microphones. In practice, the performance is often limited because traditional microphone
arrays have a static local configuration with a limited amount of wired
microphones. Therefore, the distance between
the sound sources and the microphone array can be large, yielding low SNR signals. Current
technology makes it possible to produce small microphones, each having
an individual processing unit with a wireless communication link. When
spread out over an area, these small microphones form a wireless acoustic
sensor network. Signals of all the sensors can then be combined to generate
a signal with very high sound quality.
The project concerns three sub problems:
1)
Multi-microphone signal processing algorithms must be redesigned for
random (unknown) microphone positions. Positions can even change during
operation. At each point in time, the subset of sensors that capture
the most useful information must be defined.
2) Because of bandwidth
and computational constraints, the signal processing algorithms must be redesigned for
distributed processing and compression. Furthermore, the effect of lossy audio-compression on the
performance of the algorithms stated above will be investigated.
3)
The effect of the lack of synchronization between sampling times of the
sensors on the performance of the developed algorithms will be
investigated.
In charge of: Teaching Assistent for H01Q7 P&O elektrotechniek ('Acoustic modem') Promotor(s):
Marc Moonen
Active Projects: |
