(PI - Ng Boon
Poh),
in collaboration with Institute for Infocomm
Research
Space-time processing can be view as an
evolution of the traditional array signal
processing techniques such as antenna array and
beamforming. Operating simultaneously on multiple
sensors, space-time receivers process signal
samples both in time and space, thereby improving
resolution, interference suppression, and service
quality. Existing systems use temporal signal
processing alone and, thus, cannot effectively
cope with co-channel interference (CCI).
Space-time processing exploits the spatial
dimension to mitigate CCI in a way that is
impossible with a signal antenna. By incorporating
the rich structure of space and time signals,
e.g., constant modulus, finite alphabet and
cyclostationarity, space-time processing has
reduced CCI significantly. The focus of the
previous work was on the receiver side. Here, we
will address space-time processing at both source
and receiver ends. Test-beds will be setup to
study the performance of the various algorithms
developed. We will also look into the possible
applications of the algorithms and systems
developed in the areas of computer networking and
multimedia processing.
Advanced
Signal Processing for Active Sonar Systems
(PI - Wan
Chunru), in collaboration with DSO National
Laboratories, Singapore
In this project, advanced signal
processing techniques to improve the
detection and localization performance of
active sonar system performance are
investigated. The following are the major
areas of study.
1. Optimal Detection in Presence of
Reverberation: In conventional active
sonar systems, time-varied gain is used to
counter the large variation of the
reverberation, where the reverberation
variance is modeled as the function of time.
This method usually fails to adequately
compensate for the non-stationarity of the
reverberation process. As a result, the
sonar false alarm rate is not constant.
Since the control of detector false alarm
rates is very important both for reducing
operator fatigue and for increasing the
effectiveness of computer aided detection
algorithms. We have conducted research in
designing the constant false alarm rate (CFAR)
detector for active sonar systems. The
approaches of general maximum likelihood
testing (GLRT) have been studied.
Preliminary simulation studies and real data
experiments have been conducted in
comparison with the existing active sonar
detection algorithms and the promising
results are obtained.
2. Reverberation Modeling: In
supporting the optimal detection mentioned
above, a realistic reverberation model is
necessary. The reverberation modeling
includes (1) statistical modeling which
allow us to design the optimal detectors,
(2) reverberation level prediction which
enable sonar performance prediction, and (3)
waveform generation which facilitates active
sonar signal processing algorithms
testing.
Technology Description
Microphone array system comprises multiple
microphones and digital signal processing
part. There are many applications of
microphone array. One of them is microphone
array for videoconferencing system.
In videoconferencing, it is always expected
that not only the desired voice from a desired
direction can be enhanced and the
interferences/noise from other directions can
be suppressed, but also the video camera can
be focused on the desired talking speaker
automatically even when he/her is moving. With
the proprietary microphone array technology
developed at CSP, such natural interface for
videoconferencing and other applications can
be provided even under adverse acoustic
conditions.
The performance measured in a real conference
room of the developed system is described as
follows:
Localization with high accuracy:
(a) around 2° azimuth error in average,
(b) around 4° elevation error in average, and
(c) around 0.2m distance error in average at
3m distance in a real conference room;
Short response time: less than 0.5 seconds;
Narrow beam with high gain:
(a) 10° beam range (adjustable) and
(b) around 30dB main lobe to side lobe ratio;
Voice detection with high efficiency: more
than 95% successful detection in the presence
of hand clipping, cough, object dropping,
knocking door etc.
Application
Videoconferencing, E-classroom for
E-education, home
& office automation,
voice oriented & voice activated
systems,
surveillance & security, broadcasting, etc.
Technology Description
The adaptive noise cancellation microphone system
comprise of two closely-coupled microphones and
associated digital signal processing part. It can
enhance the signal from a predefined direction and
cancel signals from other directions. It is very
useful to receive desired signal in noisy
environment. CSP has developed a novel concept and
proprietary algorithm, as well as a real-time DSP-based
demonstrator. The performance is described as
follows:
Very small size: almost same as that of single
microphone;
High directivity: around 30° beam range;
3D directivity: same beam range for both azimuth and
elevation;
High interference/noise suppression: up to 30dB.
Application
Mobile phone, PDA, handheld device, voice input
device, PC, voice oriented/ activated device etc.
