This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.
Technological progress in bio-signal processing has enabled computer-based analysis of ECG which is an adjunct to the physician, and helps them in feature extraction and measurement of ECG signal for diagnostic purposes. However, acquisition and reliable parameter extraction of ECG signals using a computer based acquisition system is extremely sensitive to power line interference even though its magnitude is imperceptible. Therefore, commonly 50 Hz active analog notch filter is used to reduce the effect of power line interference. An attempt has been made in this paper to design different 50 Hz active analog notch-filter topologies in P-Spice that produce a notch as opposed to band rejection using minimum number of high precision components. Simulation results of these filters are evaluated for their amplitude response, phase response and group delay to frequency sweeps. Based on the simulation results, add-on notch filter hardware are developed and used in conjunction with real time computer based ECG amplifier system. The front end hardware of amplifier system includes biosensors, cascaded amplifiers, right leg drive circuits, active filters and 50 hertz notch filters. Analog output of the front end system is then interfaced to the computer for viewing. The purpose of this work is to develop computer based compact and enhanced signal acquisition system to yield ECG signal of greatest clinical use.
Keywords: Biosignal processing, ECG, Notch Filter, Real Time, P-Spice, MATLAB
A patient in proximity to power buses gets capacitively coupled through the stray capacity between them, to the extent that if one square metre of the human body is coupled to the power lines one metre away, then the stray capacitance is of the order 8.85 pF. Therefore, 50 Hz power line voltage coupled with stray capacitance, exceeds the bio-potential level, and interferes with the measurement of ECG signal. Power line interference may be reduced by using differential amplifier of high common mode rejection ratio (CMRR) in designing the cascade data acquisition unit. Modern biomedical amplifiers have very high CMRR but are still contaminated by power-line interference. This is due to differences in the electrode impedances and stray currents through the patient and the cables. Hardware solutions have been developed to increase the actual CMRR by equalization of the cable shield and the right leg drive circuits . This reduces the influence of stray currents through the body, but the efficiency obtained is not significant.
Enormous efforts have been made to address this issue in the past. Li BF et al have attempted a self-adapting correlation method for fast elimination of 50-Hz noise from the ECG signal in PC-based digital ECG recording system . Dotsinsky I et al assessed the efficiency of notch filters and a subtraction procedure for power-line interference cancellation in ECG signals. However, the hardware used was incapable for power-line frequency cancellation, so a software measurement of the power-line interference period was developed . As a software approach, the interference reference signal and its quadrature, are linearly combined to be subtracted from any one of the channels to reduce line interference . Tabakov S et al described fast computation methods for real-time digital filtration and QRS detection in computerized ECG systems for long-term monitoring. Tests made on standardized ECG database were capable in removing baseline wanderings and 50 Hz interferences. The performance of the QRS detector had mean sensitivity of 99.65% and positive predictive value of 99.57% . Further to this, M.S. Chavan et al used Chebyshev type I and Chebyshev type II digital filters on the ECG Signal using MATLAB FDA tool. ECG data was acquired from the self designed instrumentation amplifier but, the interfacing of ECG amplifier to the computer was done using advantech PCL 711B add on card . They also presented the application of Butterworth and Elliptic notch and high pass digital IIR filter on the raw ECG signal in real time . Kaur M et al combined the moving averages technique and IIR notch filter to reduce the power line interference having fewer coefficients and hence lesser computation time, hence applicable to real time processing. However, the results have been concluded using Matlab and MIT-BIH database . Mbachu C.B worked on FIR digital filters with Kaiser Window to remove the power line interferences using FDA tool . But, linear filters designed to reduce power line interferences are not very effective and also have long impulse response which causes large delays in ECG monitoring and analysis. However, the non-linear filters cater to these limitations and effectively reduce these interferences as stated by Leski J et al . Their analysis was however done on synthetic signal. Literature also reports the study of the effects of AC interference and its filtering on the accuracy of heart rate detection for heart rate variability analysis .
ECG data is analyzed mainly based on the shape of the PQRST wave, so it is important to consider how the filter changes its shape. Analog filters like Wien and twin Tee or digital filters like Butterworth, Chebychev and Elliptic can be used but they not only attenuate a desired frequency, but affect adjacent frequencies as well, thus changing the basic ECG wave shape . Adaptive filtering introduces unacceptable transient response time, and digital notch filters are computer based solutions to this problem. Thus to reduce the 50 Hz noise in a PC based system, a hardware Notch filter can be a solution which completely attenuates a particular frequency and affects none other .
Therefore, various 50 Hz Notch filter designs are tested in PSpice before implementing the filter hardware to be used with the amplifier unit. The front end hardware thus includes biosensors and self developed tested cascaded amplifiers, right leg drive circuits, active filters and 50 hertz notch filters. Analog output of the front end system is then interfaced to the sound port of the computer and can be viewed on a software oscilloscope . The purpose of this work is to understand how modern ECG is derived and displayed and to ascertain standards that will improve the precision and utility of the ECG in practice. Emphasis is placed on compact and enhanced signal acquisition system and computer-based signal processing, which provide automated measurements that lead to computer-generated diagnostic inferences.
Analysis Of ECG Signal
This section examines the relation of the resting ECG to its technology and shall help clinicians to understand the missing link between technology and its consequences for clinical ECG interpretation. ECG nowadays are recorded by digital, automated machines equipped with software that measures ECG intervals and amplitudes, provides a virtually immediate analysis, and also compares the tracing to those recorded earlier by the same system. Automated ECG Signal analysis and processing include signal acquisition, filtering, waveform recognition, feature extraction, and diagnostic classification. This requires adequate understanding of ECG signal frequency range, its filtering and proper lead placement.
Band pass filtering in the range of 1 Hz to 30 Hz removes artefacts from the ECG signal, but distorts its high- and low-frequency components and is therefore not suitable for recording purposes. Inadequate high-frequency response may cause incorrect assessment of signal QRS peak amplitude and improper low-frequency response can result in distortions of repolarization. Thus, cut off frequency range for filtering ECG acquired on a computer must be chosen so that it suppresses low-frequency noise resulting from baseline wander, movement, and respiration and higher-frequency noise that results from muscle artefact, power-line or radiated electromagnetic interference . Diagnostic interpretation of ECG using computer-based system initially require signal preparation like filtering, and template formation followed by detection and feature extraction. It therefore becomes essential to understand various ECG signal frequency ranges and its effect for proper filtering. Table I below gives the frequency details of the ECG signal.
Primary wave in ECG signal
1 - 2 Hz
Diagnostic information in QRS (Adults)
Diagnostic information in QRS (infants)
250 - 500 Hz
Low-frequency cut-off for diagnostic ECG using regular filters
Does not eliminate the baseline drift
Low-frequency cut-off for diagnostic ECG using linear digital filters with zero phase distortion
Reduces ST-segment distortions
High-frequency digital filter cut-off in adults
100 - 150 Hz
Reduces error to nearly 1%.
High-frequency digital filter cut-off in infants
Maintain diagnostic accuracy of ECGTable 1. The ECG Signal frequency
Design & Development of ECG Acquisition System Using Hardware Notch Filters
The design and implementation of hardware analog circuits is however, complex and computationally intensive. So, computer software like P-Spice proves out to be very helpful in virtually designing these filters . The section below gives the design considerations for 50 Hz analog notch filters.
3.1. Design of 50 Hertz Notch Filters
Active notch filters have been used in the past for eliminating components of 50-Hz hum due to their enormous advantages over passive filters and because they provide excellent low frequency filtering, which is the prime requirement in ECG signal analysis. Active filters eliminate bulky inductors so require less space, are cheap, and provide good isolation between source and load with suitable gain. In designing these filters care needs to be taken in deciding the centre frequency tuning, stability, and repeatability as selectivity and centre frequency depends on the gain. A number of notch-filter topologies are therefore explored with design goals that:
generates a notch in-place of rejecting band of frequencies
provides a good notch depth
produces ideal frequency response, phase response and group delay
uses minimum operational amplifiers and high precision components
allows easy tuning for centre frequency & Q factor
Using P-Spice, notch filters are made by standard resistors, capacitors and UA741 operational amplifiers, as shown in Figure 3.1 (a), 3.2 (a), 3.3 (a), 3.4 (a), 3.5 (a), 3.6 (a) & 3.7 (a). The component value details for various circuits and calculation for center frequency is detailed in Table 2. Analyzing the filters in P-Spice with a frequency sweep resulted in frequency response plots of the filters and is given in Figure 3.1 (b), 3.2 (b), 3.3 (b), 3.4 (b), 3.5 (b), 3.6 (b) & 3.7 (b) followed by their phase response and time delay plots in subsections (c) and (d) respectively. A twin T-network is most commonly used notch filter, capable of generating infinitely deep notch. Twin T-network is a passive network consisting of 2 T-shaped networks. One T-network is made up of 2-resistor & a capacitor while the other is made of two capacitors & a resistor. The frequency at which maximum attenuation occurs is called notch out frequency (denoted as f0) and is given in Eq (1) as:
The major disadvantage of twin T-network is that it has a relatively low 'Q' and affects the selectivity of filter. This can be improved by increasing the Q value by using a voltage follower whose output is fed back to the junction of R/2 and 2C. The C value of notch is chosen less than 1 microfarad & then R is calculated from Eq. (1) for the desired notch out frequency, which is 50 Hz in our case (India). In place of voltage followers we can also use an emitter follower but advantage of voltage follower in that it has much higher input resistance and the output amplitude is exactly equal to the input in magnitude and in phase. The combination of twin 'T' arrangement with an operational amplifier in voltage follower mode improves the behavior further by improving the Q factor from 0.3 to little greater than 50. The depth and frequency of notch remains unchanged by adding a voltage follower in the filter circuit. The voltage follower acts as a buffer that provides a low output resistance. The use of operational amplifier provides high input impedance and so it becomes possible to use large resistance values and low capacitance values in the 'T' network for low frequency applications.
C (in nF)
Table 2. Component value details for various circuits and calculation for center frequency
Figure 3.1(a) is a Twin-T 50 Hz notch filter configuration where resistor R5 is half in value of resistors R1 and R2. Resistance value of R3 is ideally designed using combinations of R1 and R2. Similarly, capacitor C4 and C5 are half the value of C3. Capacitance value of C3 is obtained by parallel combination of C4 and C5, which are same in value. The circuit in figure 3.2 (a) uses two operational amplifiers, but the number of high precision components required is more. Figure 3.3 (a) is a very simple notch filter design made using a single amplifier and requires the minimum number of R and C components. In figure 3.4 (a), the junction at R2 and C1 is bootstrapped to the output of the voltage follower. It is a single op-amp filter circuit that raises the Q in proportion to the signal fed back to R2 and C1. Figure 3.5 (a) is a modified version of the notch filter in figure 3.4 (a), where two voltage followers are used. A second follower is used to further stabilize the Q factor and allows rejection over a wide input frequency ranges. A fraction of the output is fed back to the junction at R3 and C1 by the second voltage follower. The amount of signal fed back affects the notch Q. A potentiometer can be used in place of the resistor R4 in figure 3.5 (a) to have variable Q.
Figure 3.6 (a) is a single op-amp notch filter circuit that gives a very good notch depth as compared to other single op-amp notch circuits. The circuit however is not very flexible, as adjustments of center frequency becomes difficult. Trimming the center frequency requires adjustments of the three resistors R1, R4 and R5. The circuit requires six high precision components for tuning, of which two are the ratios of others. It is required that R2 and R3 should be very small as compared to R4 and R5. This raises the spread of resistor values which affects the depth of notch and the center frequency. Figure 3.7 (a) notch filter circuit is similar to the circuit of Figure 3.4 (a) with different values of R and C.
Figures 3.1 to 3.7 (sections a, b, c, d)
The following observations can be made from the response curves of these notch filter circuits:
In the pass-band, the magnitude response curve is almost flat
Phase response changes in the stop band. Largest change is observed at the center frequency
The group delay is high at center frequency of 50 Hz
Notch depth hugely depends on component matching
The depth of the notch obtainable in simulation results is theoretical and is not the depth that can be achieved with real-world components due to their tolerance values. The objective to design various 50 hertz notch filters for medical monitoring devices is the rejection of a specific interfering frequency. As can be seen in Table 2, the component values chosen are such, which produce a notch at the center frequency of 50.02 for most of the circuits analyzed.
Comparison is done for various theoretical results obtained in figures 3.1 to 3.7 sub-sections b, c and d. It clearly shows that sharp drop-offs are not possible using designs in Circuit 2, 3 and 6 of figure 3.2, 3.3 & 3.6 respectively. However, the notch depth achievable is remarkable in these circuits. Circuit designs 4 and 5 of figure 3.4 and 3.5 respectively, give sharp cut off notch filters, but the notch depth is not very good. Circuit of figure 3.7 does not give a good notch frequency. The analysis clearly shows that the result of Circuit 1 of figure 3.1, if implemented can give a good notch depth with sharp cut-off.
Practically, strong attenuation and moderately sharp drop-offs are possible using notch filters. Thus, this study supports the fact that analog notch filters, if constructed based on simulation results, can be used as add-on circuit for removal of 50 Hz hum in ECG amplifier units. Circuit 1 when analyzed using P-spice gave the best notch depth and sharp cut-off. Thus this circuit was implemented in the laboratory using two 9 volt batteries as shown in the figure 3.8. The results obtained in figure 3.8 using precise values of Resistors & Capacitors almost matched the theoretical results and so was used in conjunction with the amplifier system to obtain better bio-signal by reducing the effect of power line interference.
Material And Method
Functional block diagram of the developed enhanced system for ECG acquisition is shown in Figure 4.1. Disposable Ag-AgCl electrodes are positioned on both wrists and the right leg of the subject after applying electrolytic gel to pick the ECG signal.
The ECG signal acquired by these non-invasive electrodes is 1mV peak to peak so an amplification of about 500 is done in cascade to make ECG signal usable for detection of heart details. Front end of amplifier block A1 and A2 in the block diagram of Fig. 4.1 along with blocks A3, A4, A5 and A6 that provides overall gain of 500, improved CMRR and do not get saturated . This is followed by a 50 Hz Notch filter developed using standard resistors, capacitors and UA741 Operational Amplifiers. Further, sound card of the computer is used as the interface unit, to display the signal in MATLAB .
Result And Discussion
Multiple stage amplification, proper lead placement, right leg drive circuit, analog filter, hardware 50 Hz notch filter and MATLAB based virtual oscilloscope with digital filter program helped in obtaining human ECG signal. Figure 5.1 shows the snap shot of rest ECG signal without using the Notch Filter. X-axis in the result represents the number of ECG samples acquired w.r.t time. The sampling rate was set to 8000 samples per second in this work. Thus, the results obtained showing 80,000 samples actually represents 10 seconds of data. Y-axis gives the amplitude of the ECG signal. Interference present in Figure 5.1 has been reduced by using 50 Hz notch filter, which is clear from Figure 5.2.