||Signal Processing of Nano Sensor Data
||Larsen, Jan (Intelligent Signal Processing, Department of Informatics and Mathematical Modeling, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark)
Duun, Sune (Department of Micro and Nanotechnology, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark)
||Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark
||Pulse oximetry is a non-invasive method for measuring the arterial oxygen saturation. The measured signals suffer from noise and motion artifacts which make
the measurement results very unstable. Signal processing is therefore needed to estimate the oxygen saturation from the pulse oximetry recordings.
It is found that preprocessing the signals by well designed bandpass filters is
important. Simple methods previously used for pulse oximetry are implemented
as well as the more advanced Discrete Saturation Transform (DST) developed
by Masimo Coorporation and results are compared to solving the problem by
Independent Component Analysis (ICA).
The methods are tested on to data sets; one with oxygen saturation levels from
75% to 100% recorded on a resting subject and one recorded at normal oxygen
saturation level but with the subject doing motions. Good results are obtained
using ICA with constraints from the first data set, and the other methods perform well on this data set as well. When using the data set with motion artifacts,
all methods produce unreliable results.
Beside estimation of the oxygen saturation, the heart rate can be detected from
pulse oximetry. Heart rate detection is carried out by using the Fast Fourier
Transform (FFT) and is compared to a Bayesian periodic component detector
and both methods perform equally well.
Creation date: 2009-03-11
Update date: 2009-08-06