Mr Brian Lithgow currently represents the interests of the
DSP group within the ECSE Department at Monash
University. Some signal processing
topics currently researched by him and colleagues are:
- Cochlear implant signal
processing.
- Neurophysiological signal
processing in the auditory system.
- Nerve modelling -- this
incorporates the VHDL hardware implementation [co-researcher John Zakis]
of multiple nerve fibres/neurones -- this "Digital Brain"
project if sucessful has the potential to produce a new form of hardware
intellegence with an impact comparable with the microprocessor.
- Tinnitus Suppresion and
Modelling
- Vestibular Organ
Diagnostics and Modelling.
- Mammographic Image
processing using wavelets for microcalcification detection.
- Digital hearing aid using
wavelet feature extraction and neural network classification.
Implementation of features on TI TMS320C54.
- Advanced optimisation
algorithms and applications to digital signal processing algorithms for
hearing aids—a wavelet approach.
- Diagnostic pattern
detection for aid in pathogology classification eg. A wavelet based
Auditory Brainstem Response (ABR) analysis technique appears to be the
only ABR technique able to detect and localise Superior Olive Complex
abnormalities. Implementation on the TI TMS320C6201.
An example of undergraduate project success includes:
- National Institution of
Engineers (Aust) Electrical College
student award 1998 for "Ambient noise removal in cochlear implants
using wavelets: An application of the TMS320C50".
Dr David Morgan represents the Monash Centre for Biomedical
Engineering:
- Many enquiries into the
action of muscle length and tension sensors are limited by the number of
afferent nerve axons that can be identified in a signal recorded from a
bundle. As the impulses are so short and the recording time is so long,
identification needs to be done in real time to avoid excessive data
files. Such identification has traditionally relied on window
discriminators, and is limited to two or three axons. DSPs have the
potential to do this much better.
- Distributed stimulation of
muscle promises reduced fatigue in functional neuromuscular stimulation.
However, this technique requires a tension signal from the muscle, which
will not be available directly in the FNS situation. Likely indirect
sources include accelerometers and miniture gyroscopes. Extracting the
necessary information from these is likely to be an appropriate use for a DSP.
Dr. Robert Mahony (ANU):
- Subspace algorithm
developments for digital hearing aids.
- Mathematical modelling and
algorithm development.
Mr. John Zakis:
- Cochlear modelling and
imaging.
- VHDL hardware
implementation of multiple nerve fibres/neurones
Assoc. Prof David Suter:
- Image processing:
restoration of film
- Detection of
microcalcifications in mammograms.