![](bg2.png)
between RIIV and the changes in tidal volume and
respiratory rate (Johhansson et al., 1999a and 1999b;
Nilsson et al., 2003). As some commercial pulse
oximeters also provide PPG signal, these results
imply that the pulse oximeters can be a potential tool
for the acquisition of arterial oxygen saturation,
heart rate and respiration at the same time.
The relationship between RIIV and respiratory
signal has been examined extensively in the past
decades (Johhansson et al., 1999a and 1999b;
Nilsson et al., 2003). However, little research has
been done concerning the coherence between the
raw PPG signal acquired from pulse oximeter and
respiratory signal. As the RIIV may deviate with the
varying respiratory rate, the fixed bandpass filter for
PPG signal filtration may limit the accuracy of
analysis in practical conditions, especially in slow
and fast breathing cases. The objective of the present
study was to investigate whether such coherence
exists between raw PPG signal and respiratory signal.
The multi-channel autoregressive (AR) spectral
estimation method proposed by Morf et al. (1978),
was utilized for the coherence analysis under
different breathing rates and the breath-holding state
for five healthy male subjects. The two-channel AR-
based cross-spectral analysis demonstrated that raw
PPG signal and respiration were coherent
(magnitude-squared coherence greater than 0.5) at
the respiratory frequency in the subjects studied,
with changes in respiration leading to changes in
PPG. No coherence was found in breath-holding
cases for the subjects participated. The results of this
research verify that there exists the corresponding
respiratory component in spectrum of raw PPG
signal. The results may provide another attractive
approach to acquire the respiratory information from
PPG without the need of filtering. The results also
imply the possibility to acquire the physiological
parameters other than arterial oxygen saturation
form pulse oximeters.
2 METHODS AND MATERIALS
2.1 Subjects and Experiments
Five healthy male subjects (non-smoker and with no
prior history of cardiovascular disease) aged
between 22 and 24 took part in the experiments after
giving the informed consent. All subjects were asked
to refrain from caffeine and alcoholic drink at least 4
hours before the experiments. All of the experiments
were performed at the same university laboratory
with the room temperature being maintained at about
25 degrees centigrade during the night time (from 9
to 11 pm). The subjects were required of having a
resting period of at least 5 minutes under relaxation
status before the experiment.
0 5 10 15 20 25 30
-0.15
-0.1
-0.05
0
0.05
0.1
Respiration
Time (second)
0 5 10 15 20 25 30
0
0.5
1
1.5
2
2.5
3
3.5
Time (second)
PPG from Pulse Oximeter
Figure 1: Typical signals acquired. Upper: respiratory
signal (natural respiration); lower: PPG signal derived
form pulse oximeter.
Each experiment included two stages classified
by different respiratory rate (natural respiration and
holding the breath in order). Each stage was
maintained at least one minute, and the intervals
between stages were three minutes. Throughout the
experiment, the subjects were seated in a
comfortable chair with their right upper arm kept at
the height of heart level.
2.2 Signal Measurement
The physiological data acquisition system MP150
®
(Biopac Inc.) was utilized for signal measurement.
Pulse oximetry signal (by pulse oximeter module
OXY100C) and respiratory signal (by temperature
amplifier module SKT100C with fast response
thermistor sensor TSD202A) were collected
simultaneously during each experimental stage. The
Pulse oximetry probe (TSD123A, with infrared
wavelength 910 nm) was attached to the right index
finger, whereas the respiratory signal was acquired
at the nostril during the measurement. The analysis
package Biopac AcqKnowledge
®
(version 3.9.1)
was used for signal management, including the
signal quality pre-screening, data storage and
retrieval. The signals were verified visually by a
well-trained technician. A typical respiratory signal
and PPG signal acquired form pulse oximeter are
shown in Figure 1. If the signal quality was poor, the
signal would be excluded from further analysis and
the subject was asked to repeat the experiment once
again.
As the dominant components of the processed
signals primarily locate around the frequencies
below 6 Hz, a sampling frequency of 60 Hz is
RESPIRATORY INFORMATION IN ARTERIAL OXYGEN SATURATION MEASUREMENT
169