J. Biomedical Science and Engineering, 2008, 1, 79-84
Published Online August 2008 in SciRes. http://www.srpublishing.org/journal/jbise JBiSE
Noninvasive method for determining blood pre-
ssure and contours of arterial and volume pulses
K. Razi Naqvi1, Luca Parigi 1, Camer W. Vellani2 & Santosh Kumar2
1Department of Physics, Norwegian University of Science and Technology, N-7491 Trondheim, Norway. 2The Aga Khan University, Stadium Road, P.O. Box
3500, Karachi 74800, Pakistan. Correspondence should be addressed to K. Razi Naqvi (razi.naqvi@ntnu.no).
A noninvasive method for monitoring blood
pressure, based on the principles established
by Riva-Rocci and Korotkoff (K), is described; it
furnishes, after a single compression-deflation
cycle of the arm-encircling cuff, values of sys-
tolic and diastolic blood pressures as well as
the contours of the brachial arterial pulse and
the corresponding volume pulse. K-sounds are
detected by a single microphone situated in the
cubital fossa, and the time-varying cuff pres-
sure P(t) is read by a piezoresistive pressure
sensor. The behavior of P(t) during deflation is
resolved into two parts, P(t)=p(t)+b(t); p is a train
of positive going pulses (arising from arterial
pulsations), whereas b is a slowly changing
baseline. Noise pulses in the microphone out-
put are rejected by using the observation that
the first few K-sounds are emitted when p is
close to a maximum, and the last few when dp/dt
is close to a maximum. The performance of the
instrument is illustrated by showing how it
copes with ambient noise and involuntary man-
ual perturbations of P, and by presenting con-
tours of various pulses.
Keywords: Blood pressure, oscillometry, bra-
chial pulse wave analysis
Construction of a completely electronic version of the
auscultatory method of blood pressure measurement has
occupied many inventors and investigators over the last
fifty years [120]; the task requires the replacement of
the manometer and the stethoscope by appropriate trans-
ducers. Continuous measurement of the cuff pressure is
no longer regarded as a difficult task, but obtaining a
reliable record of the Korotkoff sounds (hereafter called
K-sounds), free from contamination by ambient noise, is
far from straightforward. Gilford [1] appears to have
been the first to devise an automatic sphygmomanometer
in which genuine K-sounds could be distinguished from
extraneous noise. Using a microphone in place of a
stethoscope, and a pressure transducer (of his own de-
sign) for detecting the oscillations in the cuff pressure
(hereafter called p-pulses), he fed the resulting pulses to
a coincidence circuit, and emphasized that “the pulses
received from the pressure channel and applied to the
coincidence circuit begin before sounds are heard and
die after the sound pulses have disappeared” (emphasis
added). Without making any provisions for recording the
time-dependence of the cuff pressure itself, he was able
to infer the values of the systolic blood pressure (sbp)
and the diastolic blood pressure (dbp) by registering the
cuff pressures corresponding to the first and the last co-
incidence event, respectively. At about the same time,
Geddes, Spenser and Hoff (hereafter GSH) described an
instrument in which the amplified outputs of a small mi-
crophone (embedded in the cuff so as to suppress acous-
tic artefacts) and a pressure sensor were summed and
plotted, after further amplification, on a single-channel
strip chart recorder; with this design, the K-sounds ap-
peared as spikes on the recorded trace [2].
Our main purpose here is to demonstrate that it is now
possible to incorporate into a single instrument the desir-
able features of many previous, special-purpose devices
(based on electronic filtering and/or graphic displays).
Using contemporary equipment and methods of data
analysis, we have explored the feasibility of exploiting
the temporal relationship between p-pulses and K-sounds,
noted first by Korns [21], as a means of distinguishing
between K-sounds and extraneous noise signals, and of
recording the contours of arterial and volume pulses. Our
study has revealed that both these aims can be realized
without much difficulty. We have developed a procedure
for data analysis (in the time domain) that is rapid as
well as transparent. The device presented here will
henceforth be called a sphygmopiezophonometer; it can
always be used, if one chooses to ignore the K-sounds or
if the microphone happens to malfunction, as a sphyg-
mopiezometer, a variant of the modern electronic oscil-
lometer—an instrument that makes no use of the K-
sounds, and relies instead on the fact that the cuff pres-
sure undergoes small oscillations superimposed on a
slow decline [22].
The experimental arrangement used by the authors is a
straightforward adaptation of traditional sphygmoma-
SciRes Copyright © 2008
80 K. R. Naqvi et al. / J. Biomedical Science and Engineering 1 (2008) 79-84
SciRes Copyright © 2008 JBiSE
nometry. The cuff, wrapped snugly around the upper arm
of the subject, was inflated by compressing a rubber bulb.
The rubber tube connecting the cuff to the mercury ma-
nometer was cut at a point near the manometer, and the
two severed parts were reconnected through two ends of
a T-piece, whose third end was connected through a stiff
polythene tube to the input port of a piezoresistive pres-
sure sensor, covering 050 kPa range (Motorola MPX
5050DP, Fujikura XFPM-050KP) or 025 kPa range
(Fujikura XFPM-025KP). K-sounds were detected by a
microphone, affixed to the skin overlying the brachial
artery in the cubital fossa by means of a double adhesive
collar, connected to a cardiophonograph (Nihon Kohden
Cardiofax model 6453, phono attachment AK 631D).
The signals from the two transducers were digitized by a
16-bit ADC and subsequently read by a personal com-
puter. A sampling frequency of 500 Hz per channel was
employed for routine measurements of sbp and dbp,
since the microphone output provides only event mark-
ers (in the form of spikes, whose positions suffer negli-
gible changes when the sampling frequency is lowered
from 5000 Hz to 500 Hz). When a faithful reproduction
of the waveforms was also needed, the sampling fre-
quency was raised to 5000 Hz for each channel. The
standard adult cuff of the sphygmomanometer was in-
flated by means of the rubber bulb and deflated manually
by its screw-adjustable valve at an average rate of ap-
proximately 12 mm Hg per second. The operator
touched only the release valve in the bulb, which was
held in a retort stand. The static response of each pres-
sure sensor was calibrated by using a mercury manome-
ter, and was found to be linear within the limits specified
by the manufacturer.
The ultimate goal of our data handling procedure is to
resolve P(t), the measured cuff pressure, into two com-
ponents: P=b+p, where b is a slowly changing, smooth
baseline, and p a train of pulses (hereafter called p-pulses)
that would be non-negative in the absence of noise. The
required resolution was achieved in three stages, which
are described in separate subsections.
3.1. Noise reduction and differentiation
Signal averaging, an established technique for enhancing
the signal to noise ratio, is applicable when the following
conditions are satisfied: (i) the timing of the signal is
known; (ii) an invariant component (the signal) is pre-
sent when measurements are repeated; (iii) the noise is
truly random (and thus uncorrelated with the signal) with
zero mean. By averaging repeated measurements, the
signal component adds coherently, while the noise tends
to cancel out. In our case, replicate measurements will
not yield the same waveform, but the technique can still
be applied if the data set is sufficiently dense. If the data
string consists of N=rM points recorded with a sampling
Figure 1. Plots showing the output of the pressure sensor and its
smoother representations obtained by two different methods.
rate R, where r and M are integers and r<<N, one can
regard it as the superposition, with a nonzero but negli-
gible jitter in the timing of the signal, of r data strings of
M points each. On replacing successive sets of r points
by their arithmetic means, one will obtain a smoother,
less dense data string that is, to all intents and purposes,
equivalent to that obtained by true signal averaging of r
independent data streams collected with a sampling rate
of R/r. Even with a sampling frequency of 500 Hz, the
choice r=10 was found to provide an adequate compro-
mise between a sufficiently high effective sampling rate
and an acceptable signal-to-noise ratio. It is worth point-
ing out that, if one has access to MATLAB, one can also
use a function named “decimate”; the process involves
removal of high-frequency noise by means of a lowpass
filter, and subsequent resampling of the smoothed signal
at a lower rate. We have verified that decimation (with a
third-order Chebyshev Type I filter) provides an output
that is practically identical with that obtained by averag-
ing. Figure 1 illustrates the noise reduction obtained by
using the two methods and their equivalence for the
measurements described here.
The raw data (in volts) were first converted to pres-
sure readings (in mm Hg), and then replaced by a
smoother, less dense set. The time derivative, qdp/dt, of
the p-pulses was approximated as p/t, where is the
first forward difference operator. Since differentiation of
noisy data degrades the signal-to-noise ratio, the calcu-
lated derivatives were decimated (with a decimation ratio
of 5) before being plotted. It will be convenient to call
the derivative of each p-pulse as a q-wave.
3.2. Location of the pressure minima
In what follows, a point on a P-versus-t plot is consid-
ered to be a minimum if the pressure at the point is lower
than that at any of its m neighbors on either side, with
the understanding that the parameter m, which can be set
to unity for noise-free data, is to be assigned a value at
the discretion of the experimenter. Obviously, the value
of m must be smaller than the number of data points ly-
ing within the shortest beat-to-beat interval; at the same
time, it must be sufficiently large to reject false minima;
an inappropriate value can be easily recognized, and
immediately rectified. The decimated pressure data dur-
K. R. Naqvi et al. / J. Biomedical Science and Engineering 1 (2008) 79-84 81
SciRes Copyright © 2008 JBiSE
Figure 2. An overview plot showing the output of the pressure
sensor (in mm Hg) during the deflation phase and the spikes
(arbitrary units) representing the microphone signal.
ing the decompression phase were next subjected to a
routine that searches for the minima in a user-selected
segment of the data string; the length of the section was
such that the first minimum occurred at a pressure larger
than sbp and the last at a pressure smaller than dbp. Once
the minima had been located, cubic spline interpolation
was used to construct a smooth curve passing through all
the minima (with the same time difference between suc-
cessive points as in the decimated time-pressure data);
henceforth, we will refer to a plot of this curve as the
baseline. The difference between the output of the pres-
sure sensor and the baseline, p=Pb, represents the fluc-
tuating part of the cuff pressure; a plot of p(t) against
time consists of a series of pulses, which have been des-
ignated above as p-pulses.
The capabilities of sphygomopiezphonometry will be
illustrated by arranging the results under four heads, the
first three of which concern the measurement of blood
pressure. We will first examine a data set where the mi-
crophone output shows no signs of extraneous noise;
secondly, we will examine the temporal relationship be-
tween K-sounds and p-pulses, so that we may be able to
handle situations which call for the application of some
noise-rejection criterion for distinguishing K-sounds
from acoustic contamination of the microphone output;
thirdly, we will deal with a case where the deflation of
the cuff is purposely made to follow an uneven course,
and show that this creates no particular problem. Finally,
we will examine the contours of the p-pulses and com-
pare them with waveforms obtained by other methods.
4.1. An Overview Plot
We begin by exhibiting a plot that provides an overview
of a single measurement session, and is reminiscent of
the approach used by GSH [2]; a plot like this will usu-
ally be sufficient for routine measurements of sbp and
dbp. The vertical scale in Figure 2 is labeled as the cuff
Figure 3. Temporal relationship between p-pulses and K-sounds.
pressure (P), but the ordinate is the sum P+aM, where M
is the microphone output and a is a suitable scaling fac-
tor; for interpreting an overview plot such as that appear-
ing in the main panel, let us adopt a rough-and-ready
criterion and assume (tentatively) that an audible K-
sound corresponds to a spike that can be spotted on the
scale used for the overview plot. On this basis, the first
K-sound was detected when t was close to 8 s (see inset
a), corresponding to a value of nearly 130 mm for sbp,
and the last K-sound was detected when t was close 46 s
(see inset b), corresponding to a value of about 62 mm
for dbp. In most cases, a visual examination of the over-
view plot would be sufficient to confirm, or reveal addi-
tional pulses deviating from, these relationships.
4.2. Temporal Relationship between K-sounds
and p-pulses
When qdp/dt is plotted against t, one obtains the curve
appearing in Figure 3a. An expanded view of four early
q-waves, along with the corresponding acoustic signal
A(t), is given in Figure 3b, which shows that each q-
wave has a crest (a positive peak) and a trough (a nega-
tive peak) and that A becomes appreciable only in the
region where q is close to zero; similar plots for a set of
four late q-waves (42.5t/s46.5), displayed in Figure
3c, reveal that when the K-sounds are about to become
inaudible, they almost coincide with the maximum of the
corresponding q-wave. This temporal relationship has
been found to hold in all the traces recorded by us, ob-
tained by examining more than 150 subjects. We con-
clude, on the basis of these results, that the following
statement will provide a firm basis for the identification
of sbp and dbp: the first few K-sounds almost coincide
with a maximum of p and the last few with a maximum
of q.
We now raise the question of contamination of the mi-
crophone signal by extra-arterial noise. Instead of turn-
ing attention immediately to another experiment, where
such noise is manifest even in an overview plot, we first
take a closer look at the data that have already been
shown and focus attention on the time range t/s52.5
(when the K-sounds have presumably become inaudible).
82 K. R. Naqvi et al. / J. Biomedical Science and Engineering 1 (2008) 79-84
SciRes Copyright © 2008 JBiSE
Since both q and A are dominated, at such long times, by
noise centred around the base line, we have displaced
one of these curves, and plotted A and C
q, with C=25
mm Hg/s. An examination of the resulting curves, shown
in Figure 4, confirms that one can identify, close to the
peak of each q-wave, a corresponding spike in the acous-
tic signal. Although the noise in the microphone output
is electrical rather than acoustic, these plots justify the
conclusion that extraneous acoustic noise will not con-
found the observer, unless it happens to be particularly
severe near the systole and/or the diastole; interference
by noise at other points will not affect the determination
of sbp and dbp, however large its amplitude.
As a second example of interference from noise, we
examine data from a different experiment, where acous-
tic noise itself contaminates the output. Figure 5a shows
an overview plot where some noise pulses can be seen
clearly, especially near the systole. An expanded view of
the signal during the early phase of deflation (Figure 5b)
shows four q-waves, but a larger number of acoustic
pulses; it is easy to pick out genuine K-sounds, since
these occur in the middle part of the corresponding q-
wave. Figure 5c, which has no noise pulses, has been
included to provide another illustration of the near coin-
cidence, when P(t) is close to dbp, of the tops of the q-
wave and the corresponding K-sound.
Figure 4. Contamination of the microphone output by a noise
Figure 5. Discrimation of acoustic noise pulses from genuine K-
Figure 6. Detection of motion artefact.
4.3. Insensitivity to arm movement
Involuntary movement of the arm can give rise to
changes in the cuff pressure that occur over a time scale
that is shorter than the deflation stage (ca. 50 s), but
longer than the beat rate. We would now like to show
that such medium-frequency changes in the cuff pressure
will not invalidate the analysis. Panel a in Figure 6 illus-
trates the occurrence of an additional hump in the cuff
pressure that lasts for about six beats; the upper inset
provides a magnified view of the cuff pressure, along
with the minima and the base line, and the lower inset
displays plots of the corresponding acoustic pulses and
the p-pulses. An examination of the p-pulses (panel b
and the lower inset in panel a) reveals that only three
pulses (45t/s47) suffer from noti-ceable distortion, but
this is not large enough to mask the aforementioned tem-
poral relationship between the p-pulses and K-sounds.
Jazbinsek, Luznik and Trontelj [20] have discussed this
aspect in greater detail, and have presented a different
approach for dealing with the problem; they have also
developed an alternative method for resolving P into a
base line and a succession of p-pulses [19].
4.4. Contours of brachial and volume pulses
In the past, contours of the brachial pulse have been ob-
tained either directly, through brachial arterial puncture
[23], or indirectly, by applying a piezoelectric sensor
directly over the artery [24]; the latter arrangement is
incapable of reading absolute pressure. It is a matter of
considerable interest to ask whether sphygmopiezo-
phonometry alone can provide the contour of the bra-
chial pulse. To answer this question, it will be necessary
to examine the contours of some brachial pulses derived
from our data (Figure 7) and compare them with the
traces obtained earlier by other methods.
As may be seen from the plots shown in Figures 2
and 3, the contours of the p-pulses depend on the value
of P itself. Previous workers have identified several fac-
tors which contribute to this variability [13,25]; these
include the changing compliance of the cuff and the tis-
K. R. Naqvi et al. / J. Biomedical Science and Engineering 1 (2008) 79-84 83
SciRes Copyright © 2008 JBiSE
Figure 7. Contours of brachial pulses from three subjects. The
smooth curve (solid line) was obtained by decimating the noisy
data (dotted) according to the method described in the text and
illustrated in Figure 1.
sue between the artery and the cuff. The data for Figure
7a pertain to the same set as that used for plotting Fig-
ures 24, whereas those for Figure 7b come from the
set used for the plots in Figure 5. As Chio [13] has ex-
plained, if P>sbp, the arterial vessels in the arm are
blocked and function as a vibration-conducting medium;
the pulse contour found through suprasystolic cuff
sphygmography are therefore expected to represent the
arterial pulse itself; a comparison of some typical con-
tours, plotted in Figure 7, with the waveforms published
by previous investigators [23, 24] fully supports this
surmise. It should be noted that, since the values of sbp
and dbp have already been determined, the vertical scale
can be immediately converted to the blood pressure itself.
When P is slightly lower than dbp, arterial inflow
through the upper arm is restored, but venous outflow is
still blocked; under these conditions, the contour of a p-
pulse is modulated by the elastic properties of the tissue,
so that the apparatus becomes, as others have recognized
[25], a plethysmograph and a p-pulse now becomes a
volume pulse.
The availability, towards the late 1970’s, of low-cost,
high-fidelity silicon-based pressure sensors [26] and the
emergence, at the same time, of 8-bit microprocessors,
paved the way for a new generation of automated and
portable blood pressure monitors, particularly those
based on oscillometry, an old technique that utilizes only
pressure data [22]. We would like to sum up by mention-
ing the advantages of our approach over the electronic
instruments developed by previous workers for monitor-
ing blood pressure and recording the contours of the bra-
chial pulse.
Most previous practitioners of sphygmopiezophono-
metry chose to place the microphone inside the cuff [3–
6,17,20], since this affords better protection against am-
bient noise. However, a study by Weber and coworkers
[27] revealed that auscultation of the K-sounds inside the
cuff and in the cubital fossa yield significantly different
values of sbp and dbp; this is why we decided to affix
the microphone where the stethoscope is located during a
conventional measurement. We have shown above that,
armed with modern means of recording and analyzing
the basic data (sound and pressure) and the idea of coin-
cidence exploited so far only by Gilford [1], the ap-
proach pioneered by Riva-Rocci and Korotkoff can be
transformed into a totally objective, trustworthy proce-
dure in which the microphone is placed outside the cuff,
exactly where the stethoscope is positioned for ausculta-
A forthcoming article [28] will confirm and clarify a
recent claim by Amoore [29] that, due to the distortion
introduced by the filtering network, the pulsating output
of an electronic oscillometer cannot be identified with
the p-pulses; it will also provide a recipe for recovering
the p-pulses from the oscillating output.
Having shown above that one can obtain not only the
values of sbp and dbp, but also the waveform of the bra-
chial pulse, we conclude by suggesting that our method
may also be used for pulse wave analysis, “a blend of
nineteenth century sphygmography with cuff sphygmo-
manometry” [30].
[1] Gilford, S.R. Automatic sphygmomanometer. U.S. Patent
2,827,040, March 18, 1958.
[2] Geddes, L.A., Spencer, W.A. and Hoff, H.E. (1959) Graphic
recording of the Korotkoff sounds. Am. Heart J. 57, 361–370.
[3] Geddes, L.A. and Moore, A.G. (1968) The efficient detection of
Korotkoff sounds. Med Biol. Eng. Comput. 6, 603609.
[4] Edwards, W.C. Apparatus for measuring blood pressure. U.S.
Patent 3,405,707, October 15, 1968.
[5] Hurwitz, M. Automatic recording sphygmomanometer. U.S.
Patent 3,771,515, November 13, 1973.
[6] Aisenberg, S. and Chabot, R.W. Noise rejecting electronic
sphygmomanometer and methods for measuring blood pressure.
US Patent 4,005,701, February 1, 1977.
[7] Trimmer, G.A. and Slechta, E.W. Mthod and apparatus for
measurement of blood pressure and pulse rate. US Patent
4,245,648, January 20, 1981.
[8] Swearingen, J.D. and Watson, R.C. Methods of and apparatus
for the measurement of blood pressure. US Patent 4263918, April
28, 1981.
[9] Geddes, L.A, Voelz. M, Combs, C., Reiner. D. and Babbs, C.F.
(1982) Characterization of the oscillometric method for
measuring indirect blood pressure. Ann. Biomed. Eng. 10,
[10] Ramsey, M. Automatic mean blood pressure reading device. U.S.
Patent 4,360,029, November 23, 1982.
[11] Nunn, D.E. and Beveridge, R.W. Apparatus and method for
measuring blood pressure. U.S. Patent 4,427,013, January 24,
[12] Miyawaki, Y., Matumoto, K. and Shirasaki, O. Electronic blood
pressure meter incorporating compensation function for systolic
and diastolic blood pressure determinations. US Patent 4,860,760,
August 29, 1989
[13] Chio, S. Method and apparatus for determining blood pressure
and cardiovascular condition. U.S. Patent 4,880,013, November
14, 1989.
[14] Hata, H. and Souma, T. Automatic sphygmomanometer. US
Patent 5,099,851, March 31, 1992.
84 K. R. Naqvi et al. / J. Biomedical Science and Engineering 1 (2008) 79-84
SciRes Copyright © 2008 JBiSE
[15] Shinomiya, T. Electronic sphygmomanometer. U.S. Patent
5,103,830, April 14, 1992.
[16] Brinton, T.J., Cotter, B., Kailasam, M.T., Brown, D.L., Chio, S.,
O’Connor, D.T. and DeMaria, A.N. (1997) Development and
validation of a noninvasive method to determine arterial pressure
and vascular compliance. Am. J. Cardiol. 80, 323–330.
[17] Pinto, L., Dhanantwari, A., Wong, W., Stergiopoulos, S. and
Maris, M. (2002) Blood pressure measurement in noise intensive
environment using adaptive interference cancellation. Ann.
Biomed Eng. 30, 657–670.
[18] Hersh, L. Automatic indirect non-invasive apparatus for
determining diastolic blood pressure. U.S. Patent 6,517,495,
Februay 11, 2003.
[19] Jazbinsek, V., Luznik, J. and Trontelj, Z. (2005) Non-invasive
blood pressure measurements: separation of the arterial pressure
oscillometric waveform from the deflation using digital filtering.
IFBME proceedings of EMBEC’05.
[20] Jazbinsek, V., Luznik, J. and Trontelj, Z. (2005) Averaging
oscillometric non-invasive blood pressure recordings:
Transformation into the normalised view. IFBME proceedings of
EMBEC’05. (http://fizika.imfm.si/jazbinsek/konf2005/2271F.pdf)
[21] Korns, H.M. The nature and time relations of the sounds of
Korotkov in man. (1926) Am. J. Physiol. 44, 247–263.
[22] Erlanger, J. (1916) Studies in blood pressure estimations by
indirect methods: I. The mechanism of the oscillatory criteria. Am.
J. Physiol. 39, 401–446.
[23] Kroeker E.J. and Wood, E.H. (1955) Comparison of
simultaneously recorded central and peripheral arterial pressure
pulses during rest, exercise and tilted position in man. Circulation
Res. 111, 623–632.
[24] Karr, S.G., Karwoski, T., Jacobs, J.E. and Mockros, L.F. (1985)
Transducer system for the noninvasive recording of arterial
pressure contours. Ann. Biomed Eng. 13, 425–442
[25] G. Drzewiecki and J. J. Pilla, Noninvasive measurement of the
human brachial artery pressure-area relation in collapse and
hypertension. (1998) Ann. Bi omed. Eng. 26, 965–974.
[26] Middelhoek, S. and Noorlag, D.J.W. (1981) Silicon micro-
transducers. J. Phys. E: Sci. Instrum. 14, 1343–1352.
[27] Weber, F., Analuf, M., Hirche, H., Roggenbuck, U. and Phillip,
Th. (1999) Differences in blood pressure values by simultaneous
auscultation of of Korotkoff sounds inside the cuff and in the
antecubital fossa. J. Hum. Hypertens. 13, 695700.
[28] Naqvi, K.R., Jávorfi, T. and Vellani, C.W. (2008) A method for
retrieving the tions from the output of an electronic oscillometer.
JBiSE 2, 75-78.
[29] Amoore, J.N. (2006) Extracting oscillometric pulses from the cuff
pressure: does it affect the pressure determined by oscillometric
blood pressure monitors? Blood Press. Monit. 11, 269–279.
[30] O'Rourke, M.F., Pauca, A. and Jiang, X. (2001) Pulse wave
analysis. Br. J. Clin. Pha harmacol. 51, 507–522