An Affordable Cuff-Less Blood Pressure Estimation Solution*
Monika Jain, Niranjan Kumar, Sujay Deb
IIIT-Delhi, Department of Electronics and Communication, New Delhi, India

Abstract—This paper presents a cuff-less hypertension prescreening device that non-invasively monitors the Blood
Pressure (BP) and Heart Rate (HR) continuously. The
proposed device simultaneously records two clinically
significant and highly correlated biomedical signals, viz.,
Electrocardiogram (ECG) and Photoplethysmogram (PPG).
The device provides a common data acquisition platform that
can interface with PC/laptop, Smart phone/tablet and
Raspberry-pi etc. The hardware stores and processes the
recorded ECG and PPG in order to extract the real-time BP
and HR using kernel regression approach. The BP and HR
estimation error is measured in terms of normalized mean
square error, Error Standard Deviation (ESD) and Mean
Absolute Error (MAE), with respect to a clinically proven
digital BP monitor (OMRON HBP1300). The computed error
falls under the maximum standard allowable error mentioned
by Association for the Advancement of Medical
Instrumentation; MAE < 5 mmHg and ESD < 8mmHg. The
results are validated using two-tailed dependent sample t-test
also. The proposed device is a portable low-cost home and clinic
bases solution for continuous health monitoring.
Keywords— Cuff-less blood pressure monitoring; Biomedical
signal acquisition; Wireless health monitoring
I. INTRODUCTION
A survey conducted by the World Health Organization
shows that by the year 2030, 1.56 billion adults will be living
with high blood pressure (hypertension), which accounts for
about 1/3rd of the total death count [1]. Hypertension,
although, does not have any obvious symptoms or signs, but
it does result in many pulmonary and cardiac diseases like
myocardial infarction, transient ischemic attack, dementia,
mild cognitive impairment, strokes, kidney failure (damage
to glomeruli) and many more [2]. Clearly, hypertension is
responsible for a number of the cardiovascular diseases,
which accounts for about 17 million deaths per year.
Hypertension pre-screening seems to be of crucial
importance in stemming the rapidly increasing number of
deaths from such condition. As a result, hypertension
screening based on cardiovascular data analysis has become
an active area of research in past few decades; researchers
have rigorously analyzed signals like Electrocardiogram
(ECG), Photoplethysmogram (PPG), impedance cardiogram
and phonocardiogram for monitoring the Blood Pressure
(BP). One of the major challenges in such studies is
developing a low-cost robust mechanism for collection and
automated analysis of cardiovascular data; the data collection
*Research supported by Indraprastha Institute of Information
Technology- Delhi (IIIT-D), Okhla Phase-III, New Delhi, Delhi- 110020.
Monika Jain and Niranjan Kumar are Research Assistants at IIIT-D.
(email: [email protected] ; [email protected]).
Sujay Deb was with Intel labs, Hillsboro, OR. He is now an Assistant
Professor at IIIT-D in Department of Electronics and Communication. (email: [email protected])
978-1-4577-0220-4/16/$31.00 ©2016 IEEE
procedure should be easy, non-invasive and safe enough to
gain the confidence of the masses. Further in the collection
and analysis phase, the role of a trained medical practitioner
should be as small as possible; partially solving the problem
of paucity of qualified medical experts in impoverished
regions of the world.
This study introduces a compact, portable and low-cost
health monitoring platform that can collect, process and
communicate non-invasive data (ECG and PPG) related to
the cardiovascular system and extract BP and HR based on
them. In this paper, the platform is shown to work with
PC/laptop, smart phone/tablet and Raspberry-pi (R-pi). The
device is easy to use and does not need an assistance of a
trained medical practitioner during the home based pre- and
post-hospitalization care. The high accuracy of BP and HR
estimation makes this solution suitable for clinical use as
well. Apart from this, the platform is also capable of
communicating the data using existing communication
infrastructure to cloud-based medical databases for further
analysis and reporting. This feature can be used to develop
the databases and patients’ follow-ups, which can be
accessed by the health centers, hospitals and researchers for
further analysis.
II. RELATED WORK
Many researchers are working towards developing a
reliable, affordable, accurate and easy to use BP monitor.
Mafi et al. [3] developed a cuff-based hardware that records
the oscillometric wave and proposed a pulse morphology
based approach for estimating BP. Samria et al. [4] designed
a finger PPG sensor using an infra-red LED and Ahmed et al.
[5] developed a prototype that records PPG from the user’s
head region (temple). In both the cases, features extracted
from the recorded PPG are used to estimate the BP. However,
estimation of both systolic and diastolic BP based on only
PPG may not be a reliable option since PPG has a higher
correlation with diastolic as compared to systolic BP [6]. A
complementary information must be added in order to
support a reliable systolic BP estimation. To overcome this,
ECG is generally used due to its high clinical importance.
Anisimov et al. [7] introduced a hardware that records
user’s ECG (from arm or chest) and PPG (from finger-tip or
earlobe). Pulse wave velocity extracted from the recorded
ECG-PPG is used for estimating BP. Puke et al. [8] proposed
a wearable prototype that records the ECG and PPG from
user’s chest (sternum). Pulse wave transit time extracted from
ECG-PPG is used to estimate BP. Ahmad et al. [9] developed
a cuff-based InBeam prototype in which ECG-assisted
oscillometric and Pulse Transit Time (PTT) analysis are
seamlessly integrated into the oscillometric BP measurement
paradigm. Franco et al. [10] introduced a wearable BP
monitoring system that records ECG (from the chest) and
PPG (from the forehead). BP is estimated based on pulse
5294
Figure 1. Proposed health monitoring system
arrival time which is computed within the ECG sensor node
by the means of wireless body sensor network. Most of these
devices are cuff based, bulky and requires the patient to keep
the hand at the heart level. Hence, these cannot be used for
continuous monitoring.
Apart from this, some significant advancement can be
seen in the field of smartphone-based health monitoring.
Visvanathan et al. [11] collected PPG using the built-in
camera of the mobile phone. BP is estimated using a large
number of PPG based features. Junior et al. [12] proposed
smartphone based calculation of PTT using its inbuilt camera
and standard microphone. BP estimation is performed using
PTT. Lin et al. [13] introduced a mobile phone based body
sensor network for continuous BP monitoring. This system
consist of three parts: Wrist band (for PPG), HR belt (for
ECG) and Smart Phone; PTT extracted from simultaneously
recorded ECG-PPG is used in BP estimation.
Most of these existing BP monitoring devices either used
insufficient signals or utilized very less features from these
signal. Also, the methods are not entirely developed, suitably
integrated or tested. In this work, we propose a cuff-less
health monitoring system that continuously monitors BP and
HR based on a large number of parameters extracted from the
ECG and PPG. All the components are designed from the
first principle and optimized at each step to achieve
affordability and accuracy. The estimated cost of the
developed system is around 50$. It is also validated against
clinically approved BP monitors.
III. PROPOSED HEALTH MONITORING SYSTEM
We propose a low-cost health monitoring system that can
continuously record and communicate the electrical
functioning of heart (ECG) and condition of arteries (PPG)
concurrently in a non-invasive manner, and analyze them to
extract the real-time BP and HR. The system can be seen in
Fig. 1 which is subdivided into four parts, viz. (A)
Transducer Unit, (B) Preprocessing Unit, (C) Interfacing Unit
and (D) Post-Processing. Each unit is subsequently explained
as follows:
the subject, using pain-free velcro bands. It also has an
optical reflective type PPG sensor, designed using LED
(5mm, 562-575nm wavelength) and a phototransistor (L14G2
NPN). PPG sensor is placed on the right-hand index finger.
PPG sensor and ECG electrodes are shown in Fig. 2.
Figure 2. PPG sensor (left) and ECG electrodes (right)
B. Preprocessing Unit
The ECG and PPG sensor outputs from the transducer
unit are noise-prone and weak, and hence, cannot be fed
directly to an Analog to Digital Converter (ADC). To
precondition the sensor output, filtering and amplification
circuits are designed (shown in Fig. 3) using high input
impedance operational amplifier MCP6004. Motion artifacts
and fluctuating DC baseline are removed using real-time
adaptive DC filtering.
After preconditioning, each signal is digitized using an
ADC. The ADC is selected according to the capabilities of
the computing platform. In this study, we have tested two
such solutions, one targeting high-speed computing platforms
like PC/laptop and another one targeting energy efficient, and
highly portable platforms like Android mobile/tablet and R-pi
etc. The detail of the ADCs used is given as follows:
Atmel AVR XMEGA 128A1 (ATxmega128 A1): The ECG
and PPG signals which are to be acquired on PC/laptop or
similar high-compute platforms are digitized using the
A. Transducer Unit
The transducer unit consists of 3-lead non-polarizable
ECG electrodes, which are placed on the wrists and left leg of
5295
Figure 3.
ECG and PPG preprocessing unit
inbuilt 12-bit ADC of ATxmega128 A1 board. This highspeed ADC (2 million samples per second) is used to
achieve the maximum data rate of digitization/acquisition for
multiple cardiac signals simultaneously – around 1000
samples per second with high resolution. Such ADC will
enable the concurrent acquisition of more than two signals as
well in future.
MCP3208: The analog signals can be digitized and stored
using MCP3208 (100-kilo samples per second, 12-bit
resolution). It is a low power and standalone ADC, which can
be easily interfaced with smart phone or R-Pi based
computing platforms. Overall this ADC makes the proposed
hardware more affordable and compact.
C. Interfacing Unit
The interfacing unit support multiple protocols to send the
data across different platforms, as shown in Fig. 4. To
provide wired USB interface with PC, laptop, tablet and
smart phones FT232RL (USB to serial UART interface) is
used. The data acquisition on PC/laptop is done at the rate of
1000 samples per second with the baud rate 230400 for ECG
and PPG simultaneously.
using the detected peaks. HR and BP are calculated based on
these parameters and are displayed on the GUI continuously.
Since the signals were preprocessed during acquisition, no
extra filters were incorporated during the post-processing.
IV. EXPERIMENTAL SET-UP
ECG and PPG of 72 subjects (aged between 20 to 60
years) are acquired and stored on a PC. The ground truth BP
and HR is simultaneously recorded using a clinically proven
BP monitor-OMRON HBP1300 (Measured BP and HR). A
database is created using the 10 parameters extracted from
each ECG-PPG set and the BP-HR measured corresponding
to it. 10 subjects in this database gave their readings
immediately after a physical exercise. Fig. 5 shows the lab
set-up during the data collection procedure. In order to
perform three-fold cross-validation, the entire database is
randomly divided into three equal parts (24 subjects each).
The basic framework followed by most of the researchers
for BP prediction is almost the same. All used the regression
framework,
yx

To connect to low power computing devices such as Rpi, Serial Peripheral Interface (SPI) protocol is used [14]. The
data acquisition on this R-pi is done at the rate of 1000
samples per second through two channels simultaneously for
ECG and PPG.
where ‘y’ represents the vector of expected BP (systolic or
diastolic), ‘A’ is the matrix of parameter vectors – the rows
of ‘A’ correspond to different individuals and the columns
correspond to the various parameters and ‘x’ is the regression
weights (to be estimated). The regression framework is
simple to interpret; the regression weights tell us the relative
importance of the various parameters in BP prediction. In this
study, we will solve (1) using Radial Basis Function kernel
regression [17]. To extract the HR, we simply use: HR (bpm)
= 6*(number of R peaks in 10 seconds). This enables the
device to update the HR in every 10 seconds on the real time
GUIs.
D. Post-processing
A Graphical User Interface (GUI) is designed for each
acquisition platform. The received ECG and PPG are dumped
into the GUI and plotted in real time on the respective
screens. In the back-end, significant peaks like P, Q, R, S, T
peaks in ECG and peaks, foots and dicrotic notches in
corresponding PPG are detected (using the peak detection
algorithm in [15]). Based on prior domain knowledge, 10
parameters (PTT, Pre-ejection period [16], QRS duration, R
peak amplitude, PPG pulse width and height, crest and delta
time, augmentation and reflection index [11]) are extracted
A. Results
The BP and HR estimation accuracy of the proposed
device (as compared to OMRON) is measured in terms of
normalized mean square error, Mean Absolute Error (MAE)
and Error Standard Deviation (ESD) as shown in Table I. The
computed error falls under the standard allowable error
mentioned by Association for the Advancement of Medical
Instrumentation (AAMI); MAE < 5 mmHg and ESD < 8
mmHg [18], and is better as compared to many state-of-theart approaches [5, 7]. The proposed device gives a reliable
diastolic as well as systolic BP estimation, unlike others.
Figure 4. Interfacing of the developed hardware with different
platforms
Figure 5. PC based data acquisition set-up; Box 1, 2, 3 shows the ECG
sensors and PPG sensor is placed on right hand index finger. 4 shows
the acquisition on PC and 5 shows the real time GUI that continuously
plots the signal and displays estimated BP, PTT and HR alongside)
FT232RL is replaced with IEEE standard 802.15.1
Bluetooth module HC-05 while interfacing wirelessly over
the mobile phone. The sampling rate and baud rate for
android mobile/tablet, during wired transmission mode, is
200 samples per second and 28800 respectively.
5296
Two-tailed dependent samples t-test [19] is also
conducted to verify that there is no statistically significant
difference between BP-HR estimated using the proposed
device and BP-HR measured using OMRON. The test is
conducted for each fold. Prior to conducting the t-test, the
assumption of normally distributed BP-HR value differences
between the proposed device and OMRON is examined. The
assumption is considered true as the skew and kurtosis levels
are found to be lesser than the maximum acceptable range for
conducting the t-test (skew <│2.0│ and kurtosis <│9.0│)
[20]. The null hypothesis: “There is no statistically significant
difference between the BP-HR estimations made using the
developed device and BP-HR measured using OMRON”, is
accepted; the calculated t-value is lesser than the critical tvalue and p>0.01. Table II shows the 99% confidence
interval obtained using t-test.
TABLE I.
Fold
I
II
III
Avg.
Diastolic BP
E2*
E3*
E1
E2
E3
E1
E2
E3
3.68
4.84
4.21
4.24
3.56
4.71
4.14
4.14
4.48
5.95
5.15
5.19
5.94
5.19
6.28
5.80
3.89
3.67
4.44
4.00
4.29
4.06
4.85
4.40
2.61
3.31
2.95
2.96
1.89
2.29
2.04
2.07
2.23
2.84
2.51
2.53
*E1: Normalized Mean Square Error (%); E2: Mean Absolute Error (mmHg);
E3:Error Standard Deviation (mmHg)
Fold
I
II
III
99% CONFIDENCE INTERVALS OBTAINED USING T-TEST
Systolic BP
( -1.88 , 3.26 )
( -2.97 , 3.84 )
( -3.32 , 2.58 )
Diastolic BP
( -4.21 , 0.71 )
( -1.45 , 3.20 )
(-1.41 , 4.15 )
[2]
[3]
[4]
[5]
HR
E1*
TABLE II.
[1] V. Perkovic, R. Huxley, Y. Wu, D. Prabhakaran, and S. MacMahon.
[6]
BP - HR ESTIMATION ERROR WITH RESPECT TO OMRON
Systolic BP
REFERENCES
[7]
[8]
[9]
HR
(-1.88 , 0.67)
(-1.42 , 1.84)
(-1.65 , 1.24)
[10]
V. CONCLUSION
[11]
This paper presents a portable health monitoring solution
that records and transmits the clinically significant
cardiovascular signals (ECG and PPG) to different
acquisition platforms. It also supports immediate and
automated analysis of the recorded signals on the acquisition
device itself, in order to extract the real-time BP and HR. The
validation of device is done against a clinically proven BP
monitor (OMRON HBP1300); the BP and HR estimation
error falls under the maximum standard allowable error
mentioned by AAMI. The device is developed using low cost
and locally available off the shelf ICs and discrete
components wired on a double-layer printed circuit board. It
is an affordable, easy and reliable solution for home or clinic
based continuous health monitoring. Also, the system is
capable of continuously storing the recorded data on cloudbased medical databases for further analysis.
In future, we will explore miniaturization of the proposed
device and will modify it into a wearable solution.
[12]
[13]
[14]
[15]
[16]
[17]
ACKNOWLEDGMENT
This work is partially supported by Indo-US Grand
Challenge Initiative- Affordable BP Measurement
Technologies for Low Resource Setting. We would also like
to acknowledge the support of ITRA project, funded by
DEITy, Government of India, under grant with Ref. no.
ITRA/15(57)/Mobile/HumanSense/01
[18]
[19]
[20]
5297
"The burden of blood pressure-related disease a neglected priority for
global health." Hypertension 50, no. 6, 2007, pp 991-997.
S. Mendis, P. Puska, and B. Norrving. Global atlas on cardiovascular
disease prevention and control. World Health Organization, 2011.
M. Mafi, S. Rajan, M. Bolic, V.Z. Groza, H.R. Dajani, "Blood
pressure estimation using oscillometric pulse morphology", Annual
International Conference of the IEEE Engineering in Medicine and
Biology Society, EMBC, 2011, pp.2492-2496.
R. Samria, R. Jain, A. Jha, S. Saini, S.R Chowdhury, "Noninvasive
cuff'less estimation of blood pressure using Photoplethysmography
without electrocardiograph measurement”, IEEE Region 10
Symposium, April 2014, pp.254-257.
N. Ahmed, R. Rohan, A. Ghose, and A. Sinharay. "Feasibility
Analysis for Estimation of Blood Pressure and Heart Rate using A
Smart Eye Wear." In Proceedings of the 2015 workshop on Wearable
Systems and Applications, ACM, 2015, pp. 9-14.
J. S Kim, K. K. Kim, H. J. Baek, and K. S. Park. "Comparable
parameter related to arterial stiffness in blood pressure estimation
method", International Conference on Information Technology and
Applications in Biomedicine, 2008, pp. 410-411.
A. A. Anisimov, A. Ananyeva, "Development of system for
noninvasive cuffless estimation of blood pressure”, Proceedings of
the 2014 IEEE NW Russia Young Researchers in Electrical and
Electronic Engineering Conference (ElConRusNW), 2014, pp.19-22.
S. Puke, T. Suzuki, K. Nakayama, H. Tanaka, S. Minami, "Blood
pressure estimation from pulse wave velocity measured on the chest",
35th Annual International Conference of the IEEE Engineering in
Medicine and Biology Society (EMBC), 2013, pp.6107-6110.
S. Ahmad, C. Silu, K. Soueidan, I. Batkin, M. Bolic, H. Dajani, V.
Groza, "Electrocardiogram-Assisted Blood Pressure Estimation",
IEEE Transactions on Biomedical Engineering, vol.59, no.3, March
2012, pp.608-618.
J. Franco, J. Aedo, F. Rivera, "Continuous, Non-invasive and Cufffree Blood Pressure Monitoring System", Andean Region
International Conference (ANDESCON), 2012 VI, pp.31-34.
A. Visvanathan, A. Sinha, and A. Pal. "Estimation of blood pressure
levels from reflective Photoplethysmograph using smart phones",
IEEE 13th International Conference on Bioinformatics and
Bioengineering (BIBE), 2013, pp.1-5.
A. D. Junior, S. Murali, F. Rincon, D. Atienza, "Estimation of Blood
Pressure and Pulse Transit Time Using Your Smartphone”, Euromicro
Conference on in Digital System Design (DSD), 2015, pp.173-180.
L. Hao, X. Wenyao, G. Nan, J. Dong, W. Yangjie and Y. Wang,
"Noninvasive and Continuous Blood Pressure Monitoring Using
Wearable Body Sensor Networks”, in Intelligent Systems, IEEE,
vol.30, no.6, 2015, pp.38-48.
P. G. Davis, Q. B. Huynh, and J. C. Peck Jr. "Serial peripheral
interface." U.S. Patent 8,589,717, issued November 19, 2013.
M. Jain, N. Kumar, S. Deb and A. Majumdar, “A sparse regressionbased approach for cuff-less blood pressure measurement”, The 41st
IEEE international conference on acoustics, speech and signal
processing (ICASSP), 2016, unpublished.
S. Deb; C. Nanda, D. Goswami, J. Mukhopadhyay and S. Chakrabarti,
"Cuff-Less Estimation of Blood Pressure Using Pulse Transit Time
and Pre-ejection Period", International Conference on Convergence
Information Technology, Nov. 2007, pp.941-944.
Y. W. Chang, C. J. Hsieh, K. W. Chang, M. Ringgaard, and C. J. Lin.
"Training and testing low-degree polynomial data mappings via linear
SVM." The Journal of Machine Learning Research 11 2010, pp 14711490.
AAMI, Arlington, VA, 1–70 Manual, Electronic or Automated
Sphygmomanometers, 2003.
M. R. King & N. A. Mody, Numerical and statistical methods for
bioengineering: applications in MATLAB, Cambridge University
Press, 2010.
H. O Posten, Robustness of the two-sample t-test. In Robustness of
statistical methods and nonparametric statistics, Springer Netherlands,
1984, pp. 92-99.