Computer Based Model to Filter Real Time Acquired Human Carotid Pulse

Dipali Bansal
Signal Processing: An International Journal (SPIJ), Volume (7) : Issue (1) : 2013 42
Computer Based Model to Filter Real Time Acquired Human
Carotid Pulse
Dipali Bansal dipali.bansal@yahoo.co.in
Faculty of Engineering & Technology, Department of EEE
Manav Rachna International University
Aravali Hills, Faridabad, 121001, Haryana, India
Abstract
Acquisition and reliable parameter extraction of human bio-signals is extremely sensitive to
interferences. Analogue filters have been used in the past to remove artefacts but they only help
to suppress these contaminations and are difficult to realize. Their simulated computer based
digital equivalents are however a viable and effective solution to this. But these filters are mostly
tested on stored database and so there is a need to develop a system where real time acquired
human signal is filtered and analyzed online on a computer. Present paper deals with design and
simulation of digital techniques like FIR filter, IIR notch filter, spectrum analysis and convolution
implemented on real time, non-invasively acquired carotid pulse wave on a computer based
environment named Simulink using FDA tool. Results obtained show the accuracy of the
designed techniques and gives an easy model to online acquire and filter bio-signals sitting at
home.
.
Keywords: Carotid Pulse, FIR Filter, IIR Filter, Non-Invasive, Real Time, Simulink
1. INTRODUCTION
Power line interference of 50 Hz considerably corrupts the bio-signals when detected using a
computer based acquisition system. Human physiological parameters are also affected by EMG
signals due to muscle movements. Filter algorithms can be implemented on microprocessors for
conditioning the bio-signals. In a certain application, real time digital filters are used in cascade
and are programmed on a microprocessor to obtain desired filtering task [1]. This is an old and
cumbersome method of realizing digital filters. As an alternative, computer based digital signal
processing has been widely utilized now-a-days to combat these artefacts. The clean ECG data
is then analyzed quantitatively [2]. The study seems to be made on stored ECG data-base as no
mention is done on the acquisition procedure of bio-signal. Self-adapting correlation method has
been explored in another PC-based system [3]. The system discussed, digitally records ECG
signal and claims fast elimination of 50-Hz interference from it. Adaptive digital notch filters are
also investigated and their performance is verified on a noisy EEG signal [4]. A novel approach to
eliminate the power line frequency from the ECG signal is designed using wavelets in MATLAB.
An orthogonal basis is chosen to differentiate between signal and noise to obtain a clean ECG
[5]. The work is relatively simple to implement in real time cases but is a simulated effort.
Subtraction procedure has also been used in computer simulation to eliminate power line
interference. In this work the filter is tested on animals using a signal generator to inject additive
noise interference [6]. In another work, varied digital schemes have been used to remove power
line noise but are tested on stored ECG tape [7]. Fast computation techniques are investigated
for real-time digital filtration in home monitoring autonomous ECG systems [8]. The technique
however, does not give satisfactory result in real time cases. Another paper compares the
efficiency of notch filters and subtraction procedure for removal of power-line interference in ECG
signals [9]. Yet another work has modified the subtraction procedure for use in ECG instruments

Dipali Bansal
and computer based applications, but gives only the analysis based on stored ECG database
[10].
It is evident that biomedical instruments require inherent filtering stages and customized algorithm
for parameter extraction [11]. Notch filtering of power line frequency is also considered essential
in ECG signals so that they can be analyzed for calculation of heart rate variability (HRV) [12].
Digital FIR filters are explored in Simulink for removal of power line interference in ECG signal in
yet another paper [13]. The results obtained are not so satisfactory. However computer simulation
provides an easy means to test biomedical applications [14] and digital methods can be further
investigated for better results. Summarized literature survey is presented in Table I.
Most of the literature work listed, have used stored bio-signal data for analysis in simulation
environment and do not talk about real time acquisition and online digital filtering. An attempt has
therefore been made here to implement and analyze the effect of various digital techniques like
FIR Equiripple Bandpass filter, IIR notch filter, spectrum analysis and convolution filter on a
computer based real time acquired bio-signal. The work is tested on carotid pulse waveform
acquired online and the entire model is implemented using Simulink in MATLAB. Carotid pulse
waveform is a tested physiological parameter, vastly explored these days for its amplitude and
contour that helps in the assessment of cardiac details. The digital filter model is tested on carotid
data as its acquisition does not require a complex electronic arrangement for research
applications [15].
TABLE 1 : Literature Survey.
AUTHOR YEAR SIGNAL METHODOLOGY COMMENTS
M. L. Ahlstrom et al.
[1]
1985 ECG
Digital adaptive 60-Hz interference
filter, low-pass and a high-pass filter
Microcontroller based system
T. Starr et al. [2] 2005 ECG IIR notch filter & type I FIR filter
Time & frequency domain analysis
of bio-signal
B. F. Li et al. [3] 2002 ECG Adaptive correlation method Computer based digital system
M. Ferdjallah et al.
[4]
1994 EEG
Constrained least mean-squared
(CLMS) algorithm
Comparatively complex adaptive
process
S. Poornachandra et
al. [5]
2008 ECG
Wavelet coefficient threshold based
shrinkage function using MATLAB
Simulation work to remove power
line hum
R. M. Lu et al. [6] 1993 ECG
Subtraction procedure used to
design digital notch filter
Can be used as filters in implantable
pulse generators
C. D. McManus et al.
[7]
1993 ECG Digital notch filter & adaptive filter Tested on stored signal database
S. Tabakov et al. [8] 2008 ECG Online digital filters QRS wave shifts leading to error
I. Dotsinsky et al. [9] 2005 ECG
Subtraction process & digital notch
filters
Software digital filters
recommended
C. Levkov et al. [10] 2005 ECG Modified Subtraction process
An online computer based
application
A. Shukla [11] 2008 ECG
Filters in Xilinx system generator
simulation environment
Parameter extraction done for bio-
signals
L. Hejjel [12] 2004 ECG Analogue notch filters HRV analysis done
M. S. Chavan et al.
[13]
2008 ECG Digital FIR equiripple notch filter
Results obtained can be further
improved

Dipali Bansal
Rest of the paper is organized as follows: Section 2 briefly covers the various digital filtering
techniques analyzed in this work. Section 3 details the material and methodology involved,
followed by section 4 which discusses the results obtained. Section 5 finally concludes the work.
2. DIGITAL TECHNIQUES
Digital filters are broadly categorized into infinite impulse response (IIR) and finite impulse
response (FIR) filter types. Convolution operation also plays an important role in various electrical
signal filter algorithms. Spectrum estimation provides supplementary information about the signal
as it is dealt in frequency-domain.
2.1 IIR Filter Design
The most important filter required to be designed for the task of removing noise from the bio-
signal is a notch filter because it helps in removing the specific 50 Hz power line interference. The
digital IIR notch filter is obtained from its analogue counterpart by applying bilinear
transformation. Figure 1 shows the magnitude characteristic of the proposed IIR notch filter. A
stable single notch IIR filter of order 2 with notch frequency 50 Hz and quality factor 100 is
designed for this application.
FIGURE 1: Magnitude Response Curve of Designed 50 Hz IIR Notch Filter.
2.2 FIR Filter Design
A FIR filter settles to zero in a finite duration in response to an impulse input and possesses some
very desirable characteristics. All the poles are placed inside the unit circle and so these filters
show stability and have linear phase characteristics. FIR filters are realized without any feedback,
so are simple to implement and result in reduced rounding errors. However they require more
computation power especially in low frequency applications like bio-signal processing. The
magnitude response curve of the designed FIR Bandpass filter as obtained in Simulink is shown
in Figure 2 which clearly indicates the filter specifications. The Equiripple design method is used
to find the coefficients from frequency specifications [16]. A stable Equiripple band-pass FIR filter
of order 10 is designed to filter the carotid signal. The FIR Bandpass filter parameters are:
Fstop1=0.03 Hz; Fpass1=0.05 Hz; Fpass2=150 Hz; Fstop2=160 Hz.

Dipali Bansal
FIGURE 2: Magnitude Response Curve of Designed Equiripple Bandpass FIR Filter.
2.3 Convolution
Convolution is widely explored in engineering applications now-a-days and has become an
integral part of signal processing algorithms. Of all the functions of convolution operation, filtering
& windowing are the most useful ones in bio-signal processing. Biological signals when acquired
using computer based systems are corrupted by undesirable frequencies which results in
disturbed signal output. One of the methods to drop out unwanted frequencies is the
multiplication of Fourier transform of bio-signal with a gate function having a width equal to the
required cut off frequency. The inverse Fourier transform of this product gives the new filtered
signal which represents convolution.
2.4 Spectrum Analysis
Representation in frequency-domain gives additional information about bio-signals since one of
the most important advantages of the spectrum estimation is to determine and analyze small
signals in the presence of large ones. Spectrum details can be obtained by calculating the power
spectral density (PSD) either by parametric or nonparametric methods. Parametric methods use
estimated model parameters to calculate the PSD. However, non-parametric is the classical
method of calculating direct PSD using the fast Fourier transform (FFT) and is called the
Periodogram approach which is used in this model. Selection of appropriate window function is
important in calculating the Fourier transform using Periodogram approach as they help in
deciding the spectral resolution and side lobes level. Various window functions like Rectangular,
Hanning, Hamming, Gaussian etc are designed for the purpose [17].
3. MATERIAL & METHOD
The experimental set up for recording and digital filtering of carotid data mainly consists of Piezo-
electric sensor and a MATLAB (7.4) based Simulink model. The pressure sensitive sensor has a
metallic plate of diameter 2.0 cm and thickness 0.25 mm. This metal plate is coated with a
ceramic layer of diameter 1.30 cm with a thickness of 0.1 mm. The arrangement also consists of
computer sound card connector jack, windows sound card and a personal computer. The
Simulink model designed to acquire and filter real time carotid signal is shown in Figure 3.
Simulink blocks used in the model includes analogue input block, amplifiers, scopes, vector
scope, IIR notch filter, FIR Equiripple Band-pass filter, convolution block, Periodogram and the
Buffer block.

Dipali Bansal
FIGURE 3: Simulink Model To Acquire and Filter Real Time Carotid Signal.
Measurements were made non-invasively on a healthy human subject using piezoelectric
pressure sensor placed at the neck over the carotid artery to acquire the carotid data. The
analogue output from the transducer is interfaced directly through the sound card into a computer
in real time. Analogue input block in the data acquisition toolbox is used to acquire online real
time analogue carotid data into Simulink model. The sound card is selected as the input device in
the analogue input block. The analogue input data is acquired in asynchronous mode at a sample
rate of 8000 samples/sec from the microphone-in (winsound) port of a computer. The
asynchronous mode initiates the acquisition when the simulation starts and runs while data is
being acquired into a FIFO buffer. No hardware amplifier is required for acquisition; however an
amplifier block with a gain of 40 is used in Simulink to enhance the view of the acquired raw bio-
signal. Scopes are used to view the real time obtained signal. Digital techniques are then
employed in the model to filter out the interference in the carotid signal.
FDA tool in the filter design block of signal processing block set is used to design digital IIR Notch
filter and FIR Equiripple Band pass filter as explained in section 2.1 and 2.2 respectively. The
frequency response curves for both the filters are shown in Figure 1 and 2. Convolution block is
also used to filter the bio-signal for comparison. The inputs to the convolution block are the raw
carotid data and the FIR filtered output.
The Simulink model in this work also implements a spectral analysis method to compute the
spectral content of the carotid signal. Non-parametric spectral estimation using the Periodogram
method with Hanning window function is done in this work. The signal is buffered into frames,
with 128 samples per frame using the buffer block. Each frame is then windowed using a Hanning
window function, and the FFT for the windowed frame is computed. Fast Fourier transforms for
successive frames are collected and plotted on the frequency vector scope [18]. A GUI front
panel as shown in Figure 4 is created that provides a link to the Simulink model for easy viewing.
The GUI can run the simulation and plot the results.

Dipali Bansal
FIGURE 4: GUI Front Panel That Links to The Designed Simulink Model.
4. RESULTS & DISCUSSIONS
The Simulink model designed successfully acquires the human carotid pulse in real time. The
carotid pulse waveform obtained is the result of the analogue voltage produced at the output of
the piezoelectric sensor and quantifies the pulsation of the carotid artery. The raw carotid data
obtained is shown in Figure 5 which however does not exactly match the theoretical carotid wave
contour. The positive peak on the contour represents contraction during systole and the negative
peak signifies the sudden reduction in arterial pressure during diastole. The other signal seen is
basically the noise interference due to the 50 Hz power line hum. This noise interference is
reduced to a great extent by using digital IIR notch filters and Equiripple FIR band-pass filters as
can be visually observed in Figures 6 and 7 respectively. Unwanted spikes can be seen at some
places in these plots which arise because sensor body interface is not stationary with time. Figure
8 shows the convolved output of the raw carotid data and the filtered carotid waveform. Figure 6,
7 and 8 represents the real time acquired carotid signal in time domain using different processing
techniques. The result can be utilized to measure the peak to peak period and also to analyze
HRV as the peaks are highlighted and noise is greatly reduced in this plot. However, further
algorithm can be developed for mathematical analysis and comparison. Frame 1001 of the
spectrum scope is captured in Figure 9.
FIGURE 5: Raw Carotid Waveform and Noise Acquired in Real Time Using Simulink Model.

Dipali Bansal
FIGURE 6: IIR Notch Filtered Carotid Waveform and Noise Acquired Using Simulink.
FIGURE 7: FIR Digitally Filtered Carotid Waveform and Noise Acquired Using Simulink.
FIGURE 8: Convolved Carotid Waveform Obtained Using Simulink.

Dipali Bansal
FIGURE 9: Spectrum Details of Real Time Acquired Carotid Wave in Simulink.
Computerized bio-signal processing has the potential to remove interference easily and
effectively. T. Starr has worked on 10 seconds of real clinical stored ECG signal sampled at
360Hz and has applied IIR notch filter and three Type 1 FIR filters of varying order to analyze
filter parameters to conclude their effectiveness and practical considerations. The work lacks its
implementation on real time acquired data [2]. M. S. Chavan et. al. have explored Digital FIR
filters in Simulink environment for removal of power line interference from real ECG data acquired
using proprietary data acquisition unit. The results obtained are however, not so satisfactory [13].
In yet another work varied digital schemes have been used to remove power line noise but are
tested on stored ECG tape [7]. S. Poornachandra et. al. have attempted a novel and less
complex adaptive approach to distinguish between signal and noise using orthogonal basis,
where signal enhancement is obtained using shrinkage function which discards values below a
predetermined threshold in real time applications. The technique is however, a simulation effort
tested in MATLAB [5]. R. M. Lu et. al. have explored subtraction procedure to adaptively eliminate
power line interference injected using a signal generator in animals [6]. I. Dotsinsky et. al. have
modified the subtraction procedure for computer based ECG systems, but the analysis is based
on stored ECG database [10]. Thus, it can be said that computer simulation provides an easy
platform to test bio-signal processing applications and digital techniques can be further explored
for better results.
5. CONCLUSIONS
Analogue hardware notch filters and their microprocessor based digital equivalents have their
own limitations in removing noise components from bio-signals. The aim of proposed virtual
instrument in Simulink was to utilize the knowledge of digital signal processing in removing
interference from the real time acquired carotid pulse waveform. The present work describes a
simple simulation method that helps in online digital filtering in personal home monitoring systems
since such systems are prone to artefacts. The blocks used in the model are from the main
Simulink library that gives satisfactory results and has no hardware limitations. The system
provides a PC based testing platform for acquisition, viewing and digital processing of bio-signals
and provides the flexibility of being used with different data acquisition units. A model may further
be designed using LabVIEW, where Adaptive Digital Filters with enhanced efficiency may be
used along with proprietary data acquisition units for improved display and analysis of real time
bio-signal.

Dipali Bansal
6. REFERENCES
1. M. L. Ahlstrom and W. J. Tompkins, Digital Filters for Real-Time ECG Signal Processing Using
Microprocessors, IEEE transactions on Biomedical Engineering, vol. BME-32, no. 9, sept. 1985,
pp 708-13.
2. T. Starr, Filtering A Noisy ECG Signal Using Digital Techniques, April 19, 2005.
3. B. F. Li, B. S. Yin, D. K. Yu, Fast elimination of 50-Hz noise in PC-based application system, Di
Yi Jun Yi Da Xue Xue Bao, 2002 Dec;22(12):1131-2, PMID: 12480596.
4. M. Ferdjallah, R. E. Barr, Adaptive digital notch filter design on the unit circle for the removal of
powerline noise from biomedical signals, IEEE Trans Biomed Eng. 1994 Jun; 41(6):529-36.
5. S. Poornachandra, N. Kumaravel, A novel method for the elimination of power line frequency in
ECG signal using hyper shrinkage function, Digital Signal Processing archive Volume 18, Issue 2
(March 2008). 10.1016/j.dsp.2007.03.011.
6. R. M. Lu, B. M. Steinhaus, A simple digital filter to remove line-frequency noise in implantable
pulse generators, Biomed Instrum Technol. 1993 Jan-Feb; 27(1):64-68.
7. C. D. McManus, K. D. Neubert, E. Cramer, Characterization and elimination of AC noise in
electrocardiograms: a comparison of digital filtering methods, Comput Biomed Res. 1993 Feb;
26(1):48-67.
8. S. Tabakov, I. Iliev, V. Krasteva, Online digital filter and QRS detector applicable in low
resource ECG monitoring systems, Ann Biomed Eng. 2008 Nov; 36(11):1805-15.
9. I. Dotsinsky, T. Stoyanov, Power-line interference cancellation in ECG signals, Biomed Instrum
Technol. 2005 Mar-Apr; 39(2):155-62.
10. C. Levkov, G. Mihov, R. Ivanov, I. Daskalov, I. Christov and I. Dotsinsky, Removal of power-
line interference from the ECG: a review of the subtraction procedure, BioMedical Engineering
OnLine 2005, 4:50 doi:10.1186/1475-925X-4-50.
11. A. Shukla, Exploring and Prototyping Designs for Biomedical Applications, Xcellence in
automotive & ISM, Xcell Journal, Third Quarter 2008, pp 18-21.
12. L. Hejjel, Suppression of power-line interference by analogue notch filtering in the ECG signal
for heart rate variability analysis: to do or not to do? Med Sci Monit. 2004 Jan; 10(1):MT6-13.
13. M. S. Chavan, R. A. Agarwala, M. D. Uplane, Design and implementation of digital FIR
equiripple notch filter on ECG signal for removal of power line interference, WSEAS transactions
on signal processing, issue 4, volume 4, april 2008, pp- 221 – 230.
14. M. Khan & A. K. Salhan, Biofeedback controlled anti-G suit: a computer simulation, Int.J. of
Aviation, Space & Env. Med. Mar 2007, Vol.78, Issue 3.
15. D. Bansal, M. Khan, A. K. Salhan, Real time acquisition and PC to PC wireless Transmission
of Human Carotid pulse Waveform, Published in the International Journal ‘Computers in Biology
& Medicine’, Elsevier, Science Direct, 39 (2009) 915-920.
16. J. R. Johnson, Introduction to Digital Signal Processing, Prentice Hall of India Private Limited,
Englewood Cliffs, NJ, 1992.

Dipali Bansal
17. M. Tarvainen, Estimation Methods for Nonstationary Biosignals, Doctoral dissertation, Kuopio
university, KUOPIO 2004, ISSN 1235-0486.
18. MATLAB: http://www.mathworks.com.

Computer Based Model to Filter Real Time Acquired Human Carotid Pulse

More Related Content

What's hot (19)

Similar to Computer Based Model to Filter Real Time Acquired Human Carotid Pulse (20)

Recently uploaded (20)

Computer Based Model to Filter Real Time Acquired Human Carotid Pulse