ECG Signal Processing Using Digital Signal Processing Techniques - IJSER

International Journal of Scientific & Engineering Research, Volume 4, Issue 12, December-2013ISSN 2229-55181624ECG Signal Processing Using Digital Signal Processing TechniquesS.Thulasi Prasad (Phd).,Associate Professor.,CVSCE,Tirupati.Email: stprasad123@yahoo.co.inAbstract: This work describes theimplementation of wavelet-based de noisingalgorithm on electrocardiogram (ECG)signal and detection of important parametersuch as heart rate, amplitude, timings of theECG, etc. The algorithm is implemented inDSP based starter kit (DSK) with a twoelectrode ECG preamplifier. The signal fromthe ECG preamplifier is acquired throughthe Codec input of the DSP starter kit. Theacquired data is subjected to signalprocessing techniques such as removal ofpower line frequencies and high frequencycomponent removal using wavelet-denoisingtechnique. ECG component analysis such asQRS peak detection, heart rate calculation,etc is performed using nonlinear filtertechnique called first order derivative andmoving average filter. The performance ofthe algorithm is studied in the DSPenvironment as well as MATLABenvironment for comparison. The results ofthis study reveal the potentiality of the DSPsystem for routine clinical use.Dr. S. Varadarajan, comprocessing. In this design, high-speedfloating point digital signal processorTMS320C6711andTLC320AD535dualchannel voice/data codec based DSPstarter kit (DSK) was employed forprocessing the ECG.Electrocardiogram (ECG) signal frequencyrange varies between 0 Hz-300 Hz and mostof the information available in the signal liesin the range 0.5-150 Hz “Ref. [1-4]”.Therefore, the removal of higher frequenciesis necessary to eliminate the unwantedsignals, which reduces only less than 1% ofthe useful information. ECG signalprocessing comprises of two steps viz. (i)preliminary processing and (ii) primaryprocessing. In preliminary processing,artifacts like higher peaks due to electrodemotion and power line interference areremoved through the application of suitablesoftware filters in the DSK system.IJSERIndex Terms: ECG, DSP, Denoising,Wavelet, Heart rate, Power line interference1. IntroductionIn recent years, there has been increasinginterest in the design and implementation ofDSP systems for real time ECG signalIn primary processing, techniques likedenoising, baseline wandering and detectionof P, QRS, and T waveforms are performedthrough the implementation of suitablealgorithms in the DSK system. Foranalyzing the ECG signal in DSK system,the ECG signal is sampled at the frequencyof 1 kHz and the sampled data is stored inDSP buffer for processing. This sampledECG data are subjected to various signalIJSER 2013http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 4, Issue 12, December-2013ISSN 2229-5518processing algorithms to obtain a noise freeand clear ECG waveform for analysis.2. Hardware DetailsFig. 1, depicts the complete setup for DSPbased ECG system, which comprises of a setof electrodes, ECG pre-amplifier board,TMS320C6711 DSP Starter Kit (DSK) with3.5mm audio jack, and Pentium IV DesktopPC. The DSP based ECG system has beenbuilt around the TMS320C6711 DSK.1625environment (IDE), called Code ComposerStudio (CCS). The CCS is a high-levellanguage, which has built-in FFT, Wavelet,and other functions for signal processingapplications. Also, we can develop our ownfunctions in C for dedicated and novelapplications. The TMS320C6711 DSK has abuilt in CODEC, which has 16-bit register toacquire the ECG signal directly. Theacquired ECG data can be viewed in thedisplay before processing. Here 2048samples can be viewed at a time but the datacan be updated in a circular for real timedisplay of ECG. The DSK can process upto5000 samples at a time, but the display canstore and view only 2048 samples at a timedue to its limitation in video buffer memory.Hence a circular memory technique isemployed in this design to view theprocessed data in a sequential manner.IJSERFig.1 Block Diagram of the DSP Starter Kitbased ECG Analysis Experimental SetupA two-electrode ECG preamplifier “Ref.[5]” is constructed using op-ampsMCP607/OPA2336. A set of standard stickon disposable electrodes are placed in thetwo arms of the subject (patient) picks-upthe ECG signal from the body and the signalis amplified to 1 V level by the ECGpreamplifier circuit. The output of theamplifier is directly connected with theCodec input of the DSK system. The Codecsampled the ECG signal at the sampled of 1kHz. The data is stored in DSK buffermemory for processing.3. Description Of the DSK EnvironmentFor Algorithm DevelopmentFor the development of algorithms in theDSK system, the Texas Instrument DSK isprovided with an integrated developmentFig. 2 is the display of the 2048 raw samplesof a typical ECG signal acquired using thepreamplifier hardware and the DSK system.Here the X-axis represents the samplenumber and the Y-axis represents theamplitude of the signal. This view shows 7QRS peaks and other components of theECG waveforms such as P and T, which areburied in artifacts and noises.Original signalFig. 2 Plot of the actual ECG signal withoutany processingIJSER 2013http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 4, Issue 12, December-2013ISSN 2229-55184. Implementation Of Wavelet TransformFor DenoisingDenoising is the primary processing toremove all the high frequency as well aspower supply interference from the ECGsignal. Several researches have beenattempting wavelets for denoising ofbiomedical signals “Ref. [6, 7, 8]”. Toestimate the performance of wavelet indenoising, biomedical researchers havemade several attempts employing variouswavelet basis functions like Coiflets, Haar,etc in the “Ref. [9, 10]” The outcome of thisstudy revealed that the performance of theDaubechies (DB4) wavelet basis function indenoising is extremely well and has thebasis function graphically shown in Fig 3.Also, the Daubechies wavelet was chosenfor this work on the basis of the ics of the DB4 basis functionwith the ECG waveform. The denoised ECGsignal using the Daubechies DB4 wavelet isshown in the Fig. 4, which clearly indicatesthat the Daubechies DB4 Wavelet denoisingis an efficient method to completely removeall high frequency as well as power-lineinterferences.1626in the signal. Almost all other unwantedinformations are removed.Fig. 4 Denoising of ECG signal usingDaubechies wavelet5. Analyses Of ECG Waveforms UsingFiltered Derivative Operator And MovingAverage FilterMost of the diseases and complaints arereflected in the waveform intervals,morphology of the waveforms, and theiramplitude values. Most important diagnosticinformation can be obtained by the detectionof QRS complex, calculation of variousintervals and heart rate measurement.Hence, analysis of ECG waveform isperformed mainly on the detection of QRSwaveform “Ref. [11, 12]” and thelocalization of other waveforms (S, T, P)with respective to the QRS complex. In thiswork, the detection of QRS complex isperformed using weighted and squared firstderivative operator (Filtered- Derivative)and Moving Average (MA) filter proposedby Murthy et al “Ref. [13]”. A FilteredDerivative Operator can be represented as,IJSERFig. 3 Daubechies WaveletDaubechies DB4 wavelet has the exactresponse characteristics like the ECG signalexcept the baseline wandering, which hasvery low frequency of the order of 0.05 Hzwhere x(n) is the digitized ECG sample, N isthe width of a window within which firstorder differences are computed, squared,and weighted by the factor (N-i 1). Theweighting factor provides a smoothingIJSER 2013http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 4, Issue 12, December-2013ISSN 2229-5518effect. Further smoothing was performed bythe moving average filter defined by thefollowing equation.This algorithm provides a single peak foreach QRS complex and suppresses the othercomponents in the ECG such as P, T, etc.Fig. 5 shows the QRS peak detection afterapplying the Filtered Derivative Operatorand Moving Average Filter. Here the QRSdetection is achieved by applying athreshold and finding out the number ofpeaks that cross the threshold value, fortypercentage of the maximum amplitude of thesignal is chosen as threshold value for thedetection of QRS peaks.1627Using the above relationship, the heart ratecalculated for normal case as 73 beats perminute (bpm) for the R-R interval of 825 msec, which is well agreed with thecommercial machine readings. The samemethodology was employed forotherimportant interval calculation by fixingdifferent threshold for identifying the othercomponents of the ECG such as P, T, etc.Important intervals and durations of thewaveforms are found after identifying andlocating the P, QRS and T waves.6. Calculation EfficiencyFig. 6 shows the timing diagram of theprocessor in terms of the execution cycle.The total processing time becomesT t T proc N Tist . (3)IJSERIn this process, the position of each QRS hasbeen identified and from the QRS values,the R-R interval is measured by calculatingthe number of samples between adjacent Rpeaks multiplied with the sampling time.Fig. 6 Processor efficiency AnalysisWhere N is the number of interruptions thatoccur in Tt, given by,N T t / T T t T sam . (4)Therefore,T t (1- fsam T ist ) T proc . (5)Fig. 5 Application of first-Derivative andMoving AverageFrom the Fig 5, the time duration calculatedbetween the adjacent R peaks as 825 m sec.From the R-R interval, the heart rate iscalculated using the formulaHeart Rate (1/RR interval in sec.) * 607. ConclusionIn this work, the power of the memory onchip DSP was realized by implementing thevarious signal processing algorithms forECG signal processing and analysis. Novelsignal processing techniques such asDaubechies Wavelet and Filtered-Derivativeoperator are tested for ECG denoising andQRS peak detection respectively. De-IJSER 2013http://www.ijser.org