Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Revisiting QRS Detection Methodologies for Portable, Wearable, Battery-Operated, and Wireless ECG Systems

  • Mohamed Elgendi ,

    Affiliation Department of Computing Science, University of Alberta, Edmonton, Alberta, Canada

  • Björn Eskofier,

    Affiliation Pattern Recognition Lab, Friedrich-Alexander-University Erlangen-Nuremberg, Bavaria, Germany

  • Socrates Dokos,

    Affiliation Graduate School of Biomedical Engineering, University of New South Wales, Sydney, New South Wales, Australia

  • Derek Abbott

    Affiliation School of Electrical and Electronic Engineering, University of Adelaide, Adelaide, South Australia, Australia

Revisiting QRS Detection Methodologies for Portable, Wearable, Battery-Operated, and Wireless ECG Systems

  • Mohamed Elgendi, 
  • Björn Eskofier, 
  • Socrates Dokos, 
  • Derek Abbott


Cardiovascular diseases are the number one cause of death worldwide. Currently, portable battery-operated systems such as mobile phones with wireless ECG sensors have the potential to be used in continuous cardiac function assessment that can be easily integrated into daily life. These portable point-of-care diagnostic systems can therefore help unveil and treat cardiovascular diseases. The basis for ECG analysis is a robust detection of the prominent QRS complex, as well as other ECG signal characteristics. However, it is not clear from the literature which ECG analysis algorithms are suited for an implementation on a mobile device. We investigate current QRS detection algorithms based on three assessment criteria: 1) robustness to noise, 2) parameter choice, and 3) numerical efficiency, in order to target a universal fast-robust detector. Furthermore, existing QRS detection algorithms may provide an acceptable solution only on small segments of ECG signals, within a certain amplitude range, or amid particular types of arrhythmia and/or noise. These issues are discussed in the context of a comparison with the most conventional algorithms, followed by future recommendations for developing reliable QRS detection schemes suitable for implementation on battery-operated mobile devices.


According to the World Health Organization, cardiovascular diseases (CVDs) are the number one cause of death worldwide [1]. An estimated 17.3 million people died from CVDs in 2008, representing 30% of all global deaths [1]. Moreover, it is expected that the number of mortalities due to CVDs, mainly from heart disease and stroke, will reach 23.3 million by 2030 and are projected to remain the single leading cause of death for several decades [2].

In 2010, the global direct and indirect cost of CVD was approximately $863 billion and is estimated to rise by 22% to $1,044 billion by 2030. Overall, the cost for CVD alone is projected to be as high as $20 trillion over the next 20 year period [3].

As a consequence of direct and indirect costs of CVD, medical researchers have placed significant importance on cardiac health research. This has led to a strong focus on technological advances with respect to cardiac function assessment. One such research pathway is the improvement of conventional cardiovascular-diagnosis technologies used in hospitals/clinics.

The most common clinical cardiac test is electrocardiogram (ECG) analysis. It represents a useful screening tool for a variety of cardiac abnormalities because it is simple, risk-free, and inexpensive [4]. Advances in technology have led to much change in the way we collect, store and diagnose ECG signals, especially the use of mobile phones to implement the clinical routine of ECG analysis into everyday life [5][9]. Thus, in the near future, it is expected that Holter devices, which are traditionally used for ECG analysis in the clinic, will be replaced by portable, battery-operated devices such as mobile phones in the near future [10]. The reason is that Holter devices do not detect arrhythmias automatically in real-time, and do not provide real-time information to the hospital/doctor/patient when a critical heart condition occurs.

Moreover, the advances in memory/storage technology have enabled us to store more ECG signals than ever before. Therefore, researchers are collecting more information in order to understand the mechanisms underlying CVDs, which is expected to ultimately lead to effective treatments. The trend towards using mobile smart phones for ECG assessment further speeds up this process, as the conveniently collected data can potentially be added to databases via the existing internet.

The analysis of ECG signals collected by a mobile phone needs to be fast and feasible in real-time, despite the existing limitations in terms of phone memory and processor capability. The same holds for the ability to analyse large ECG recordings collected over one or more days.

Recently, researchers have put an increased effort into developing efficient ECG analysis algorithms to run within mobile phones, including algorithms for determining the quality of collected ECG signals [11]. This increased effort is also evidenced in the 2011 PhysioNet/Computing in Cardiology Challenge [12], which has been established to encourage the development of ECG software that can run on a mobile phone, recording an ECG and providing useful feedback about its quality.

PhysioNet provided a large set of ECG records for use in their Cardiology Challenge, along with an open-source sample application for an Android phone (Google Inc., USA), and that can classify ECGs as acceptable or unacceptable. Therefore, the next step is to analyse the acceptable ECG signal for diagnosis, without relying on an expert for interpretation. If this possibility becomes a reality, it will help developing nations and rural populations, by benefitting from otherwise inaccessible expertise.

Note that ECG signals contain features that reflect the underlying operation of the heart. These features represent electrophysiological events that coincide with the sequence of depolarisation and repolarisation of the atria and ventricles. The signal of each heartbeat contains three main events: the P wave, the QRS complex, and the T wave (as shown in Figure 1). Each event (wave) has its corresponding peak. The analysis of ECG signals for monitoring or diagnosis requires the detection of these events. Once an event has been detected, the corresponding signal can be extracted and analysed in terms of its amplitude (peak), morphology, energy and entropy distribution, frequency content, intervals between events and other more complex parameters. The automatic detection of the P, QRS and T events is critical for reliable cardiovascular assessment, such as diagnosing cardiac arrhythmias [13][17], understanding autonomic regulation of the cardiovascular system during sleep and hypertension [18], [19], detecting breathing disorders such as obstructive sleep apnea syndrome [20], [21], and monitoring other structural or functional cardiac disorders. Once the QRS, P and T events are detected accurately, a more detailed analysis of ECG signals can be performed.

Figure 1. Main Events in ECG signals.

A typical ECG trace of the cardiac cycle (heartbeat) consists of a P wave, a QRS complex, and a T wave.

The detection of QRS complexes has been extensively investigated over the past two decades. Many attempts have been made to find a satisfying universal solution for QRS complex detection. Difficulties arise mainly because of the diversity of the QRS waveforms, abnormalities, low signal-to-noise ratio (SNR) and as well as artefacts accompanying ECG signals. Conversely, P and T event detection has not been investigated as much as QRS detection, and the P and T event detection problem is still far from being solved [22]. Reliable P and T wave detection is more difficult than QRS complex detection for several reasons, including low amplitudes, low SNR, amplitude and morphology variability, and possible overlapping of the P wave and the T wave. Any cardiac dysfunction associated with excitation from ectopic centres in the myocardium may lead to premature complexes (atrial or ventricular), which change the morphology of the waveform and the duration of the RR interval. The occurrence of multiple premature complexes is considered clinically important, as it indicates disorders in the depolarisation process preceding the critical cardiac arrhythmia. For all the above-mentioned reasons, the accurate detection of QRS complexes is clinically important. Prior to developing a fast-robust QRS detector that suits battery-driven applications and continuous 24/7 ECG monitoring, it is necessary to evaluate the performance of the current algorithms against the following three assessment criteria:

  1. Robustness to noise: there are several sources of noise (e.g. power line interference, muscle noise and motion artefacts). Therefore, the developed algorithms should be robust to these noise sources.
  2. Parameter choice: The choice of parameters must lead to accurate detection. Parameters must not have to be manually adjusted for different recordings.
  3. Numerical efficiency: The developed algorithm may have a large number of iterations, parameters to adjust, features extracted, or classification steps. It is desirable to provide numerically efficient (simple, fast, and fewer calculations) algorithms. Of course, computers have become very fast, and therefore numerical efficiency is less important than it used to be. However, if a simple and fast algorithm can achieve good results, there is no need for more complex algorithms. In particular, when the algorithm is used online (in a slightly modified form from the offline version) in a mobile phone embedded system, numerical efficiency is still relevant.

In the remainder of this review article, these proposed assessment criteria will be used to evaluate several well-known QRS algorithms in two important stages: QRS enhancement and QRS detection. The QRS enhancement stage is used to enlarge the QRS complex relative to the other ECG features (P, T, and noise). This stage is occasionally referred to as pre-processing or feature extraction. The QRS detection stage is used to demarcate the QRS complex by providing the onset and offset points of the QRS complex, and especially the location of the prominent R peak. The remainder of this paper is structured as follows: the next section delineates several types of QRS enhancements techniques, whilst Section 3 compares different QRS detection methods. Finally a discussion and concluding remarks are presented in Section 4.

In describing the algorithms for QRS enhancement and detection in this article, note that refers to the raw ECG signal collected from any ECG monitoring system, including battery-operated devices; while refers to the filtered signal.

QRS Enhancement

This section presents several signal processing techniques [23][26] that have been used to emphasise the QRS segment in time, frequency and time-frequency series, as shown in Table 1. Figure 2 demonstrates the importance of the QRS enhancement stage as a prerequisite for detecting the QRS complex.

Figure 2. QRS enhancement stage in ECG signals.

(a) ECG signal (top: from record 100 of the MIT-BIH Arrhythmia Database [62]), (b) amplitude from Eq.1 where , (c) first derivative from Eq.4, (d) first derivative and second derivative from Eq.7, and (e) digital filter from Ref. [33]. Signal amplitudes have been manipulated to fit all signals in one figure. Here, a red asterisk represents the annotated R peak.

Table 1. Comparison of QRS enhancement techniques based on algorithm usage and assessment criteria.


This algorithm is considered the oldest for detecting R peaks in ECG signals; however, for the last 30 years it is still useful and in common use. Recently, Sufi et al. [63] used the algorithm for detecting heart rate using mobile phone. In older algorithms, amplitude threshold was not used alone as in the case of Sufi et al. [63]; it was usually followed by a differentiation step to reduce the P and T wave influence relative to the R wave. The first derivative is applied after the amplitude threshold to accentuate the slope of the QRS complex. The amplitude threshold is calculated as a fraction of the measured ECG signal(1)where is the percentage of the ECG signal required to be removed and . Different amplitude thresholds have also been used. Moriet-Mahoudeaux et al. [23] developed a QRS detector using , which means that values below 30% of the maximum positive signal amplitude is truncated from the signal, while Fraden and Neuman [24] used .

First Derivative Only

In this class of QRS enhancement algorithms, a first-order differentiator is commonly used as a high-pass filter, to enhance base-line wander and eliminate any undesired high frequency noise, modify the phase of the ECG signals, and to create zero crossings in the location of the R peaks. Many first derivative QRS detection algorithms, introduced in literature [31] calculate the first derivative of the measured ECG signal according to:(2)In contrast, Holsinger [32] used a central finite-difference approach as:(3)whilst Okada [30] used a backward difference scheme:(4)

In these algorithms, a threshold criterion was subsequently applied to for QRS detection, as summarized in Table 1.

First and Second Derivative

Note that, QRS enhancement algorithms compute the first and second derivatives of the measured ECG signal independently. A linear combination of the magnitudes of these derivatives then used to emphasise the QRS complex area relative to the other ECG features. In a seminal paper, Balda et al. [26] calculated the first and second derivatives of the measured ECG signals according to:(5)(6)

They then formed a linear combination of both derivatives as follows:(7)

Ahlstrom and Tompkins [27] calculated the rectified first derivative of the ECG as:(8)

The rectified first derivative was then smoothed:(9)

A rectified second derivative was then calculated:(10)

Finally, the rectified smoothed first derivative was added to the rectified second derivative:(11)

For all these algorithms, a threshold criterion for QRS detection was applied to the linear combination of derivatives. A summary of these threshold criteria is given in Table 1.

Digital Filters

There have been many sophisticated digital filters for QRS enhancement published in the literature [28], [33], [64][71], as described briefly below. Algorithms utilizing more complex digital filters [28][30], [72][76] include Engelse and Zeelenberg [28], who first passed the ECG signal through a differentiator:(12)

This signal was then passed through a digital low-pass filter:(13)

A different digital filter algorithm was introduced by Okada [30], who first smoothed using a three-point moving-average filter:(14)

The output of this filter was then passed through a low-pass filter:(15)

The difference between the input and output of this low-pass filter was then squared, in order to suppress low amplitude waves relative to the R peak:(16)

This square difference was then filtered, in order to enlarge the QRS area compared to the other ECG features:(17)

In addition to the above filters, a multiplication of backward difference (MOBD) algorithm has also been proposed [77], [78] for QRS detection. In brief, this approach consists of an AND-combination of adjacent magnitude values of the derivative. The MOBD of order is defined by(18)where contains the extracted QRS features, which can subsequently be detected using an appropriate threshold. Another algorithm proposed by Dokur et al. [65] uses two different bandpass filters, subsequently multiplying the filter outputs and to form:(19)where contains the extracted QRS features. This procedure is based on the assumption that each QRS complex is characterised by simultaneously occurring frequency components within the passbands of each filter. The multiplication operation performs the AND-combination. In other words, the output of the AND-combination (the feature output) is ‘true’, and therefore indicates a QRS complex, only if both filter outputs are ‘high’. The location of the maximum amplitude is taken as the location of the R wave. Conversely, Pan and Tompkins [33] used a derivative after applying a bandpass digital filter to the ECG signals. The bandpass filter consisted of a low-pass filter () followed by a high-pass filter () as:(20)(21)

The first derivative () used after the bandpass filter was specified as:(22)

The bandpass filtered signal () was differentiated to emphasise high signal slopes, suppressing smooth ECG waves and baseline wander.

Mathematical Morphology

The use of mathematical morphology operators for QRS detection was described by Trahanias [79]. The mathematical morphology approach originates from image processing and was first proposed for ECG signal enhancement by Chu and Delp [80], who reported the successful removal of noise from the ECG using the approach. Mathematical morphology is based on the concept of erosion and dilation. Let and denote discrete functions, where the sets and are given by and . Here, is the set of integer numbers. The erosion of the function by the function is defined as [80]:(23)where is also referred to as the structuring element, and . The values of are always less than those of . The dilation of the function by the function is defined as [80](24)where in this case . The values of are always greater than those of . Erosion and dilation may be combined for additional operations. Opening, denoted by , is defined as erosion followed by dilation. Closing, denoted by , is defined as dilation followed by erosion. Both operators manipulate signals in a comparable way. That is, to open a sequence with a flat structuring element will remove all peaks. To close the sequence with the same structuring element will remove all pits (negative peaks). In Trahanias [79], opening and closing operations are used for noise suppression as proposed by Chu and Delp [80]; that is:(25)where is a flat structuring element (zero line). The generation of a feature signal for the QRS complexes is accomplished by the operation(26)Zhang and Lian [34] used the first derivative after multiscale mathematical morphology filtering to the ECG signal in order to remove motion artifacts and base line drifts. They used Okada's first-order differential equation, as shown in Equation 4.

Empirical Mode Decomposition

Empirical mode decomposition (EMD) was introduced by Huang et al. [81] for nonlinear and non-stationary signal analysis. The key part of this method is that any complex data set can be decomposed into a finite and often small number of intrinsic mode functions (IMFs), which admit well-behaved Hilbert transforms. Usually, when the raw ECG signals are decomposed into number of IMFs, the combination of IMFs produces a resulting signal where the QRS complex is more pronounced. This process can be considered as adaptive filtering, similar to the use of wavelet transform. The EMD is defined by a process called sifting. It decomposes a given signal into a set of components, the IMFs. modes and a residual term [82], [83] are obtained and expressed by:(27)

The EMD algorithm is summarised by the following steps:

  1. Start with the signal ; followed by the sifting process , .
  2. Identify all local extrema of .
  3. Compute the upper (EnvMax) and lower envelopes (EnvMin) by cubic spline interpolation of the maxima and minima.
  4. Calculate the mean of the lower and upper envelopes, .
  5. Extract the detail .
  6. If is an IMF, go to step 7; otherwise, iterate steps 2 to 5 on the signal , . (The definition of an IMF, although somewhat vague, consists of two parts: (a) the number of the extrema equals the number of zeros and (b) the upper and lower envelopes should have the same bsolute value).
  7. Extract the mode .
  8. Calculate the residual .
  9. If has less than two extrema, the extraction is finished ; otherwise, iterate the algorithm from step 1 on the residual , .

Hilbert Transform

The use of the Hilbert transform for QRS detection is proposed by Zhou et al. [84] and Nygards and Srnmo [85]. In the time domain, the Hilbert transform of the ECG signal is defined as:(28)(29)where denotes the convolution operator. In the frequency domain, the ECG signal can be transformed with a filter of response:(30)where the transfer function of the Hilbert transform is given by:(31)

Using the numerically efficient Fast Fourier Transform (FFT), the Hilbert transform can easily be computed. The Hilbert transform of the ECG signal is used for the computation of the signal envelope [85], which is given for band-limited signals by(32)

A computationally less expensive approximation to the envelope can be made by [85](33)

To remove ripples from the envelope and to avoid ambiguities in the peak level detection, the envelope is low-pass filtered in Nygards and Srnmo [85]. Additionally, they propose a waveform adaptive scheme for the removal of low-frequency ECG components is proposed. The method of Zhou et al. [84] is related to the algorithms based on the Hilbert transform. In their study, the envelope of the signal is approximated using(34)where and are the outputs of two orthogonal digital filters, namely:(35)(36)

In order to remove noise, the envelope signal is smoothed by a four-tap moving average filter. Some investigators use a first derivative before applying the Hilbert transform [35][37]. Differentiating the ECG modifies its phase, creating zero crossings at the presumed location of the R peaks. Thus, a transformation is required to rectify the phase in order to create a signal with marked peaks at the true location of the R peaks.

Filter Banks

Filter banks decompose the bandwidth of the input ECG signal into sub-band signals with uniform frequency bands. The sub-bands can be downsampled, since the sub-band bandwidth is much lower than the input signal. The sub-bands provide information from various frequency ranges; thus, it is possible to perform time- and frequency-dependent processing of the input signal.

As shown in Figure 3, a filter bank contains analysis filters, which decompose the input signal into sub-band signals with uniform frequency bandwidths, each of constant length. The analysis filters bandpass the input ECG signal to produce the sub-band signals:(37)

Figure 3. Filter bank schematic.

A filter bank contains a set of analysis filters that decompose the input signal into sub-bands with uniform bandwidths in order to extract ECG features. Here, is a downsampling process producing down-sampled signals .

The effective bandwidth of is and ; thus, they can be downsampled to reduce the total rate. The downsampling process (Fig. 3), keeps one sample out of all samples. The downsampled signal is(38)where . The sub-bands and are bandpassed versions of the input, and has a lower sample rate than . The filtering process can be efficiently conducted at the input rate by taking advantage of the downsampling. This process is referred to as polyphase implementation and it contributes to the computational efficiency of filter bank algorithms [47]. A variety of features indicative of the QRS complex can be designed by combining sub-bands of interest reported in Afonso et al. [47]. For example, a sum-of-absolute values feature can be computed using sub-bands, . From these sub-bands six features (, , , , , and ) can be derived as follows:(39)(40)These features have values that are proportional to the energy of the QRS complex. Finally, heuristic beat-detection logic can be used to incorporate some of the above features that are indicative of the QRS complex.

Wavelet Transform

Wavelets are closely related to filter banks. The wavelet transform (WT) [86] of a function is an integral transform defined by(41)where denotes the complex conjugate of the wavelet function . The transform yields a time-scale representation similar to the time-frequency representation of the short-time Fourier transform (STFT). In contrast to the STFT, the WT uses a set of analysing functions that allow a variable time and frequency resolution for different frequency bands. The set of analysing functions—the wavelet family —is deduced from a mother wavelet by:(42)where and are the dilation (scale) and translation parameters respectively. The scale parameter a of the WT is comparable to the frequency parameter of the STFT. The mother wavelet is a short oscillation with zero mean. The discrete wavelet transform (DWT) results from discretised scale and translation parameters; for example, and , where and are integers. This choice of and leads to the dyadic WT (DyWT):(43)(44)(45)

Although defined as an integral transform, the DyWT is usually implemented using a dyadic filter bank where the filter coefficients are directly derived from the wavelet function used in the analysis [87][89].

QRS Detection

After enhancing the QRS features using the previous algorithms, the next step is to detect the QRS complexes. Through the previous enhancement step, QRS complexes are filtered and magnified relative to other ECG features and noise. There are many detection techniques used in the literature, as shown in Table 2. This include thresholding, neural networks [91], [112][114], hidden Markov model [95], matched filters [115], [116], syntactic methods [104][106], zero-crossing [107], and singularity techniques [117][119]. In the summary of Table 3, all these algorithms are numerically inefficient except thresholding. As the main purpose of this article is to highlight suitable algorithms for ECG monitoring using battery-operated, portable devices, only thresholding will be considered for the detection phase for simplicity and efficiency. In this context, it has to be emphasised that thresholding can be applied to time-domain [23], [24], [120] as well as time-frequency [121][123] ECG signals. However, the use of a fixed threshold to detect QRS complexes is simple and only efficient for stationary ECG signals with similar beat-to-beat morphology. Due to severe baseline drift and movement of patients, an ECG waveform may vary drastically from one heartbeat to the next in mobile applications. Therefore, the probability of not accurately detecting QRS complexes is high. Using adaptive thresholding [59], [124][126], the probability of missing QRS complexes decreases. However, the main drawback of these adaptive-thresholding based algorithms is the setting of multiple thresholds empirically. Therefore, currently, these algorithms cannot provide a universal solution to the QRS detection problem, since they may work perfectly on some clean signals, but not those containing arrhythmias or noisy QRS complexes.

Table 2. Comparison of QRS detection techniques based on algorithm usage and assessment criteria.

Table 3. Comparison of ECG beat detection algorithms based on techniques for QRS enhancement and detection on the MIT-BIH arrhythmia database [62].


The performance of QRS detection algorithms are typically assessed using two statistical measures: sensitivity and positive predictivity , where is the number of true positives (QRS complexes detected as QRS complexes), is the number of false negatives (QRS complexes which have not been detected), and is the number of false positives (non-QRS complexes detected as QRS complexes). The sensitivity reports the percentage of true beats that were correctly detected by the algorithm, whilst the positive predictivity reports the percentage of beat detections that were true beats.

The performance of current QRS detection algorithms described in the literature has not been completely assessed in terms of robustness to noise, parameter choice, and numerical efficiency. Moreover, many of the QRS algorithms have not been tested against a standard database, or any database at all making the results difficult to compare and evaluate. Furthermore, many algorithms scored a high detection performance using the overall number of detected beats (i.e. QRS complexes), as shown in Table 3. It is worth noting that the algorithm of Li et al. [127] scored high overall performance with a sensitivity of 99.89% and a specificity of 99.94%. However, Li et al. excluded files 214 and 215 from the MIT-BIH arrhythmia database [62], and therefore their algorithm may not superior in terms of performance. In addition, their algorithm was based on wavelet feature extraction and singularity for classification, which is considered numerically inefficient.

As noted, some investigators have excluded records from the MIT-BIH arrhythmia database [62] for the sake of reducing noise in the processed ECG signals; consequently their algorithms appeared to achieve improved performance. Other researchers excluded segments with ventricular flutter [60] and signals from patients with paced beats [110] from their investigations. Therefore, a robust algorithm is required to analyse ECG signals without excluding any records or particular segments, especially if the main goal is to provide a robust algorithm for long-term ECG signals recorded over a few days.

Robustness to Noise

Robustness to noise is effectively tested as we use signals from the widely used MIT-BIH Arrhythmia Database [62] that contains signals with different noise sources and non-sinus beats. The MIT-BIH database is widely used to evaluate QRS detection algorithms. As demonstrated in Table 1, there are many algorithms used for denoising and enhancing the QRS complex in ECG signals.

Usually, denoising ECG signal requires a bandpass filter, which can be implemented on battery-driven devices and while reasonably preserving the clinical features of ECG signals (P, QRS, and T waves) at the same time. Perhaps, a more sophisticated algorithm may filter the ECG more effectively, for example Sameni et al. [128] proposed a Bayesian framework that filters ECG better than the conventional bandpass filtering [129][131], adaptive filtering [132], and wavelet denoising [133], [134] over different types of noise using highly realistic synthetic ECG. Recently, Sharma et al. [135] proposed a wavelet-based denoising method tested on real ECG data and synthetic ECG signals. However, both algorithms are numerically inefficient.

Table 3 shows that the Chiarugi et al. [136] as well as Christov [124], and Elgendi [137] algorithms are highly-numerically efficient, and the use of a first derivative with or without moving average in the QRS enhancement phase is promising, especially when it is followed by a proper QRS detection phase such as moving average and/or dynamic threshold. However, the only use of derivative in the QRS enhancement phase without a proper QRS detection phase is extremely sensitive to noise [29].

It is worth noting that Elgendi's algorithm [137] tested on the MIT-BIH Noise Stress Test Database and scored higher accuracy in detecting R peaks compared to Pan-Tompkins [33] and Benitez et al. [36].

Battery-Driven ECG Devices

Many QRS detection algorithms have been published, and a comparison between them needs to be conducted. An algorithmic comparison regarding numerical efficiency has been carried out empirically. As shown in Table 3, each algorithm has been categorised as low, medium or high in terms of its numerical efficiency, based on the number of iterations and the number of equations (e.g. multiplications, additions, differentiations) employed. The better the numerical efficiency, the faster the algorithm, and vice-versa. Consequently, the faster the algorithm, the more suitable it is for real-time monitoring.

With advances in computational power, the demand for numerical efficiency has decreased. However, this is still more the case when the ECG signals are collected and analysed in hospitals, but not for the case of portable ECG devices, which are battery driven. This leads to especially high demands on algorithms for use within a mobile phone for monitoring ECG signals of patients in a mobile, unobtrusive at home setting. Therefore, there is a need for developing numerically efficient algorithms to accommodate the new trend towards mobile ECG devices and to analyse long-term recorded signals in a time-efficient manner.

Typically, processing large databases is carried out on PC workstations with high-speed, multi-core processors and efficient memory. This advantage is still not available for battery-operated devices: even the current smartphone platforms have limited RAM and processing power [8], [10], [142]. In general, battery-driven ECG devices follow one of these schemes: 1) collect data for offline analysis; 2) collect data for real-time analysis within the device itself; or 3) collect data for real-time analysis via a remote connection to a separate server. Certainly, each scheme has its own advantage and disadvantage in terms of processing time and power consumption.

The Holter device is the most commonly-used ECG battery-operated platform, especially for monitoring and recording ECG signals to be processed offline. With the advancement of smartphones in terms of memory and processors, investigators are trying to replace the Holter devices by smartphones [8]. Furthermore, the use of a smartphone has extra advantages from the patient perspective such as mobility, familiarity and guaranteed usage [143]. Thus, recently, there have been some contributions in phone applications that analyse ECG signals collected wirelessly via Bluetooth [8], [10], [144] and Zigbee radio protocols [142].

The current advances in battery-driven devices such as smartphones and tablet computers have made these technologies invariably part of daily life, even in developing countries [12]. It has also increased the possibility of implementing more sophisticated algorithms such as the Pan-Tompkins method [33] on smartphones as shown in Figure 4. However, there is a significant trade-off as there will always be a power-consumption limitation in processing ECG signals on battery-operated devices. Therefore, prior to deploying any algorithm on modern mobile devices, comprehensive evaluation of the algorithm based on robustness to noise, parameter choice, and numerical efficiency is required to improve the quality of diagnosis with respect to processing time or power consumption. One of the recent studies that confirms this recommendation is done by Hyejung et al. [145] who developed a simple algorithm to detect QRS complexes for Holter devices. Their simple algorithm, which consists of bandpass filter followed by multiple thresholds, was faster and more efficient compared to relatively more complex methods [35], [146].

Figure 4. Screenshot showing the main interface of the ‘Hearty’ application implemented by Gradl et al. (2012) [8].

From top to bottom: Panel showing various clinically relevant parameters that are automatically detected including heart rate (HR) and RR interval; Panel showing the detected ECG signal, which is wirelessly streamed to the application; Panel showing the QRS detection with filled circle markers for the Q, R and S waves; Panel showing the detected beat-to-beat heart rate.

Mobile Telemedicine Systems

Mobile telemedicine systems often use mobile phones/PDAs to just collect the ECG data–wirelessly or wired–and send them to a central monitoring station using GSM or internet for further analysis [147], [148]. In such cases, some analysis can be done locally on the smartphone before transmission; however, it is not always recommended as the transmission can consume more power than the ECG analysis itself [149]. There is no doubt that the essential quality for any algorithm used for real-time analysis is its simplicity (numerical efficiency), provided this does not decrease accuracy. The simpler the algorithm (while retaining accuracy), the faster it will be in processing large databases [35], [150], and it will consume less power for battery-operated devices [63], [142], [151]. Moreover, a simple algorithm also offers low cost of hardware implementation in both power and size for body sensor networks [152]. Sufi et al. [163] investigated three simple QRS algorithms suitable for mobile phones. The QRS enhancement phase of these algorithms consisted of amplitude, first-derivative, and second-derivative methods, while the QRS detection phase was threshold-based. They used simple methodologies for QRS enhancement and detection for implementation over mobile phones. This simplicity has been confirmed in Table 3 where it is evident that the first derivative and threshold are an efficient combination for detecting QRS if developed properly.

Mobile Phone Applications

To demonstrate the importance of processing time on a mobile phone, a showcase is demonstrated in Figure 5 for three outdated mobile phones [63]. It can be seen that the Nokia 6280 consumes the least processing time, as shown in Fig. 5(c). As expected, the amplitude-based QRS enhancement technique was faster than the first-derivative and second derivative based techniques. In this study [63], the quality of ECG signals was discussed and the data used was relatively noise-free, as the ECG signal shown in Figure 2 illustrates. However, this does not mean that a simple (or faster) algorithm will be more accurate. For example, Figure 6 shows that a simple amplitude threshold or first derivative method does not emphasize the QRS complex for the case of paced beats (record 107) and inverted QRS complexes (record 108). Nevertheless, the Sufi et al. result is considered a foundational step for monitoring ECG signals using mobile phones, but their algorithm exhibited some limitations in terms of memory and processing time.

Figure 5. A showcase of realtime factors for three outdated mobile phones.

Three QRS detection algorithms were tested, as reported by Sufi et al. [63]. The QRS enhancement phase was based on amplitude, first-derivative, and second-derivative techniques, whilst the QRS detection phase employed thresholding. Realtime factor is the processing time needed to run the QRS detection algorithm for an individual ECG entry within one measurement window size of 60 seconds.

Figure 6. QRS enhancement stage in ECG signals.

(a) ECG signal (top: from record 107, bottom: from record 108 of the MIT-BIH Arrhythmia Database [62]), (b) amplitude from Eq.1 where , (c) first derivative from Eq.4, (d) first derivative and second derivative from Eq.7, and (e) digital filter from Ref. [33]. Signal amplitudes have been manipulated to fit all signals in one figure. Here, a red asterisk represents the annotated R peak.

Nowadays, smartphones possess advanced processing and storage capabilities, including a powerful CPU, more memory and GPUs with high-speed data access via Wi-Fi or mobile broadband [153]. Therefore, implementing a sophisticated QRS detection algorithm on a smartphone is becoming more feasible. Gradl et al. [8] implemented the Pan-Tompkins algorithm [33] on three smartphones: Samsung GT-I9000, Samsung™ GT-N7000, and HTC™ Wildfire S A510e. The authors showed that processing of the wirelessly streamed ECG signal is feasible in real-time on the mentioned devices; however, they stated that battery lifetime was affected negatively by running the monitoring application.

Certainly, the implementation of the Pan-Tompkins algorithm is more resource-demanding, and therefore consumes more time and power, than the three simple algorithms investigated by Sufi et al. [63]. Nevertheless, recent wearable devices can easily fulfil the real-time requirement. For example, the real-time factor for processing record 100 of the MIT-BIH Arrhythmia Database [62] using the Pan-Tompkins algorithm over three recent tablet computers/smartphones: the Asus Transformer Prime, the Samsung Galaxy S III, as well as the Samsung Galaxy S II was 0.14, 0.13, and 0.2, respectively. In contrast, the real-time factors for processing the same record using the first-derivative algorithm on the outdated phones: the Nokia N91, the Siemens C75, and the Nokia 6280 were 0.13, 0.1, 0.016, respectively.

Another aspect that has been ignored in the literature is the clinical utility of the ECG algorithms. It is rare to find a study that addresses the usefulness of the developed algorithm in a clinical setting. As far as we are aware, there is no evidence that shows whether the discussed algorithms are currently implemented and tested in clinical settings.


In conclusion, we provide a summary of the required algorithms for ECG detection based on the literature together with our own investigations. The use of the first-derivative of the filtered ECG with or without a moving-average filter is recommended, as this approach is highly numerically efficient for the QRS enhancement phase, but is sensitive to noise and arrhythmia; therefore, an adaptive thresholding or integration-based approach is needed in the detection phase. Both of these suggested methodologies are simple and computationally efficient for the detection of QRS complexes in mobile-phone applications. If more processing power is available, as is the case on modern tablet computers and smartphones, implementation of the classical Pan-Tompkins algorithm [33] is also a feasible choice. Overall, simplicity and efficiency are required in developing QRS detection algorithms for processing long-term recordings and large databases, as well as for expanding our telemedicine capabilities in the near future.


The authors would like to thank Patrick Kugler and Stefan Gradl for re-evaluating the Pan-Tompkins algorithm [33] result on modern wearable devices and the presentation of Fig. 4. Mohamed Elgendi would like to thank Dr Gari Clifford for helpful discussions.

Author Contributions

Conceived and designed the experiments: ME. Performed the experiments: ME BE. Analyzed the data: ME BE DA. Contributed reagents/materials/analysis tools: ME BE SD DA. Wrote the paper: ME BE SD DA.


  1. 1. Alwan A (2011) Global status report on noncommunicable diseases 2010. World Health Organization.
  2. 2. Goss J (2008) Projection of Australian health care expenditure by disease, 2003 to 2033. Australian Institute of Health and Welfare.
  3. 3. Bloom D, Cafiero E, Jané-Llopis E, Abrahams-Gessel S, Bloom L, et al.. (2011) The global economic burden of noncommunicable diseases. Technical report, Geneva: World Economic Forum.
  4. 4. Dilaveris PE, Gialafos EJ, Sideris SK, Theopistou AM, Andrikopoulos GK, et al. (1998) Simple electrocardiographic markers for the prediction of paroxysmal idiopathic atrial fibrillation. American Heart Journal 135: 733–738.
  5. 5. Ren-Guey L, Yih-Chien C, Chun-Chieh H, Chwan-Lu T (2007) A mobile care system with alert mechanism. IEEE Transactions on Information Technology in Biomedicine 11: 507–517.
  6. 6. Rasid MFA, Woodward B (2005) Bluetooth telemedicine processor for multichannel biomedical signal transmission via mobile cellular networks. IEEE Transactions on Information Technology in Biomedicine 9: 35–43.
  7. 7. Wen C, Yeh MF, Chang KC, Lee RG (2008) Real-time ECG telemonitoring system design with mobile phone platform. Measurement 41: 463–470.
  8. 8. Gradl S, Kugler P, Lohmuller C, Eskofier B (2012) Real-time ECG monitoring and arrhythmia detection using Android-based mobile devices. In: Proc. 2012 Ann Int. Conf. IEEE Engineering in Medicine and Biology Society (EMBC). San Diego, CA, USA, pp. 2452–2455.
  9. 9. Scully C, Jinseok L, Meyer J, Gorbach AM, Granquist-Fraser D, et al. (2012) Physiological parameter monitoring from optical recordings with a mobile phone. IEEE Transactions on Biomedical Engineering 59: 303–306.
  10. 10. Oresko J (2010) Portable heart attack warning system by monitoring the ST segment via smartphone electrocardiogram processing. Ph.D. thesis, University of Pittsburgh.
  11. 11. Hayn D, Jammerbund B, Schreier G (2011) ECG quality assessment for patient empowerment in mHealth applications. In: Computing in Cardiology, 2011. pp. 353–356.
  12. 12. Silva I, Moody GB, Celi L (2011) Improving the quality of ECGs collected using mobile phones: The Physionet/Computing in Cardiology Challenge 2011. In: Computing in Cardiology, 2011. pp. 273–276.
  13. 13. Tsipouras MG, Fotiadis DI, Sideris D (2002) Arrhythmia classification using the RR-interval duration signal. In: Proc. IEEE Computers in Cardiology. pp. 485–488.
  14. 14. Tran T, McNames J, Aboy M, Goldstein B (2004) Prediction of paroxysmal atrial fibrillation by analysis of atrial premature complexes. IEEE Trans Biomedical Engineering 51: 561–569.
  15. 15. Bashour C, Visinescu M, Gopakumaran B, Wazni O, Carangio F, et al. (2004) Characterization of premature atrial contraction activity prior to the onset of postoperative atrial fibrillation in cardiac surgery patients. Chest 126: 831S–a-832S.
  16. 16. de Chazal P, O'Dwyer M, Reilly RB (2004) Automatic classification of heartbeats using ECG morphology and heartbeat interval features. IEEE Trans Biomedical Engineering 51: 1196–1206.
  17. 17. Krasteva VT, Jekova II, Christov II (2006) Automatic detection of premature atrial contractions in the electrocardiogram. In: Electrotechniques Electronics E & E. volume 9-10 , pp. 49–55.
  18. 18. Scholz UJ, Bianchi AM, Cerutti S, Kubicki S (1997) Vegetative background of sleep: spectral analysis of the heart rate variability. Physiol Behavior 62: 1037–1043.
  19. 19. Trinder J, Kleiman J, Carrington M, Smith S, Breen S, et al. (2001) Autonomic activity during human sleep as a function of time and sleep stage. J Sleep Res 10: 253–264.
  20. 20. Zapanta L, Poon C, White D, Marcus C, Katz E (2004) Heart rate chaos in obstructive sleep apnea in children. In: Proc. 26th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (IEMBS '04). pp. 3889–3892.
  21. 21. Shouldice R, O'Brien L, O'Brien C, de Chazal P, Gozal D, et al. (2004) Detection of obstructive sleep apnea in pediatric subjects using surface lead electrocardiogram features. Sleep 27: 784–792.
  22. 22. Goutas A, Ferdi Y, Herbeuval JP, Boudraa M, Boucheham B (2005) Digital fractional order differentiation-based algorithm for P and T-waves detection and delineation. ITBM-RBM Elsevier 26: 127–132.
  23. 23. Morizet-Mahoudeaux P, Moreau C, Moreau D, Quarante JJ (1981) Simple microprocessorbased system for on-line e.c.g. arrhythmia analysis. Medical and Biological Engineering and Computing 19: 497–500.
  24. 24. Fraden J, Neuman M (1980) QRS wave detection. Medical and Biological Engineering and Computing 18: 125–132.
  25. 25. Gustafson D (1977) Automated VCG interpretation studies using signal analysis techniques. Technical Report R-1044, Charles Stark Draper Laboratory, Cambridge, MA, USA.
  26. 26. Balda R, Diller G, Deardorff E, Doue J, Hsieh P (1977) The HP ECG analysis program. In: JH van Bemmel and JL Willems, Editors, Trends in Computer-Processed Electrocardiograms: 197–205.
  27. 27. Ahlstrom ML, Tompkins WJ (1983) Automated high-speed analysis of Holter tapes with microcomputers. IEEE Transactions on Biomedical Engineering 30: 651–657.
  28. 28. Engelse WAH, Zeelenberg C (1979) A single scan algorithm for QRS-detection and feature extraction. Proc IEEE Computers in Cardiology 6: 37–42.
  29. 29. Friesen G, Jannett T, Jadallah M, Yates S, Quint S, et al. (1990) A comparison of the noise sensitivity of nine QRS detection algorithms. IEEE Transactions on Biomedical Engineering 37: 85–98.
  30. 30. Okada M (1979) A digital filter for the QRS complex detection. IEEE Transactions on Biomedical Engineering 26: 700–703.
  31. 31. Menrad A (1981) Dual microprocessor system for cardiovascular data acquisition, processing and recording. In: Proc. 1981 IEEE Int. Conf. Industrial Elect. Contr. Instrument. pp. 64–69.
  32. 32. Holsinger W, Kempner K, Miller M (1971) QRS preprocessor based on digital differentiation. IEEE Transactions on Biomedical Engineering 18: 212–217.
  33. 33. Pan J, Tompkins W (1985) A real-time QRS detection algorithm. IEEE Transactions on Biomedical Engineering 32: 230–236.
  34. 34. Zhang F, Lian Y (2007) Electrocardiogram QRS detection using multiscale filtering based on mathematical morphology. In: Proc. 29th Annual International Conference of the IEEE Engineering in Medicine and Biology Society. pp. 3196–3199.
  35. 35. Arzeno N, Deng Z, Poon C (2008) Analysis of first-derivative based QRS detection algorithms. IEEE Transactions on Biomedical Engineering 55: 478–484.
  36. 36. Benitez DS, Gaydecki PA, Zaidi A, Fitzpatrick AP (2000) A new QRS detection algorithm based on the Hilbert transform. In: Proc. IEEE Computers in Cardiology. pp. 379–382.
  37. 37. Arzeno N, Poon C, Deng Z (2006) Quantitative analysis of QRS detection algorithms based on the first derivative of the ECG. In: Proc. 28th Annual International Conference of the IEEE Engineering in Medicine and Biology Society. pp. 1788–1791.
  38. 38. Zhang F, Lian Y (2007) Novel QRS detection by CWT for ECG sensor. In: Proc. IEEE Biomedical Circuits and Systems Conference. pp. 211–214. doi: 10.1109/BIOCAS.2007.4463346.
  39. 39. Ulusar UD, Govindan RB, Wilson JD, Lowery CL, Preissl H, et al.. (2009) Adaptive rule based fetal QRS complex detection using Hilbert transform. In: Proc. Ann. Int. Conf. IEEE Engineering in Medicine and Biology Society (EMBC). pp. 4666–4669.
  40. 40. Lin CC, Hu WC, Chen CM, Weng CH (2008) Heart rate detection in highly noisy handgrip electrocardiogram. In: Proc. IEEE Computers in Cardiology. Bolognia, Italy, pp. 477–480.
  41. 41. Yongli C, Huilong D (2005) A QRS complex detection algorithm based on mathematical morphology and envelope. In: Proc. 27th Annual International Conference of the Engineering in Medicine and Biology Society. pp. 4654–4657.
  42. 42. Jing-tian T, Xiao-li Y, Jun-chao X, Yan T, Qing Z, et al.. (2008) The algorithm of R peak detection in ECG based on empirical mode decomposition. In: Proc. Fourth International Conference on Natural Computation. volume 5 , pp. 624–627.
  43. 43. Xing H, Huang M (2008) A new QRS detection algorithm based on empirical mode decomposition. In: Proc. 2nd International Conference on Bioinformatics and Biomedical Engineering. pp. 693–696.
  44. 44. Arafat A, Hasan K (2009) Automatic detection of ECG wave boundaries using empirical mode decomposition. In: Proc. IEEE International Conference on Acoustics, Speech and Signal Processing. pp. 461–464.
  45. 45. Zhang F, Lian Y (2009) Wavelet and Hilbert transforms based QRS complexes detection algorithm for wearable ECG devices in wireless body sensor networks. In: Proc. IEEE Biomedical Circuits and Systems Conference. pp. 225–228.
  46. 46. Zhou HY, Hou KM (2008) Embedded real-time QRS detection algorithm for pervasive cardiac care system. In: Proc. 9th International Conference on Signal Processing. pp. 2150–2153.
  47. 47. Afonso V, Tompkins W, Nguyen T, Luo S (1999) ECG beat detection using filter banks. IEEE Transactions on Biomedical Engineering 2: 192–201.
  48. 48. Zhang F, Wei Y, Lian Y (2009) Frequency-response masking based filter bank for QRS dection in wearable biomedical devices. In: Proc. IEEE International Symposium on Circuits and Systems. pp. 1473–1476.
  49. 49. Vai MI, Li-Gao Z (2004) Beat-to-beat ECG ventricular late potentials variance detection by filter bank and wavelet transform as beat-sequence filter. IEEE Transactions on Biomedical Engineering 51: 1407–1413.
  50. 50. Afonso VX, Tompkins WJ, Nguyen TQ, Trautmann S, Luo S (1995) Filter bank-based processing of the stress ecg. In: the 17th IEEE Annual Engineering Conference in Medicine and Biology Society. volume 2 , pp. 887–888.
  51. 51. Mengda L, Vinod AP, Samson SCM (2011) A new flexible filter bank for low complexity spectrum sensing in cognitive radios. Journal of Signal Processing Systems 62: 205–215.
  52. 52. Dinh HAN, Kumar DK, Pah ND, Burton P (2001) Wavelets for QRS detection. In: the 23rd Annual International Conference of the IEEE Engineering in Medicine and Biology Society. volume 2 , pp. 1883–1887.
  53. 53. Szilagyi L (1999) Wavelet-transform-based QRS complex detection in on-line Holter systems. In: Proc. First Joint 21st Annual Conference and the 1999 Annual Fall Meeting of the Biomedical Engineering Society (BMES/EMBS). Atlanta, GA, USA, volume 1 , p. 271.
  54. 54. Liang-Yu S, Ying-Hsuan W, Hu W (2004) Using wavelet transform and fuzzy neural network for VPC detection from the Holter ECG. IEEE Transactions on Biomedical Engineering 51: 1269–1273.
  55. 55. Zheng X, Li Z, Shen LL, Ji Z (2008) Detection of QRS complexes based on biorthogonal spline wavelet. In: Proc. Int. Symp. Information Science and Engineering (ISISE '08), Shanghai, China. volume 2 , pp. 502–506. doi:10.1109/ISISE.2008.61.
  56. 56. Alesanco A, Olmos S, Istepanian R, Garcia J (2003) A novel real-time multilead ECG compression and de-noising method based on the wavelet transform. In: Proc. IEEE Computers in Cardiology. pp. 593–596.
  57. 57. Ahmed S, Al-Shrouf A, Abo-Zahhad M (2000) ECG data compression using optimal nonorthogonal wavelet transform. Med Eng Phys 22: 39–46.
  58. 58. Szilagyi SM, Szilagyi L (2000) Wavelet transform and neural-network-based adaptive filtering for QRS detection. In: Proc. 22nd Annual Int. Conf. IEEE Engineering in Medicine and Biology Society, Chicago, Illinois, USA. volume 2 , pp. 1267–1270.
  59. 59. Xiaomin X, Ying L (2005) Adaptive threshold for QRS complex detection based on wavelet transform. In: Proc. 27th Annual International Conference of the Engineering in Medicine and Biology Society. pp. 7281–7284.
  60. 60. Martinez JP, Almeida R, Olmos S, Rocha AP, Laguna P (2004) A wavelet-based ECG delineator: evaluation on standard databases. IEEE Transactions on Biomedical Engineering 51: 570–581.
  61. 61. Bothe HH (1997) Neuro-Fuzzy-Methoden. Berlin, Germany: Springer-Verlag.
  62. 62. Moody GB, Mark RG (2001) The impact of the MIT-BIH arrhythmia database. IEEE Engineering in Medicine and Biology Magazine 20: 45–50.
  63. 63. Sufi F, Fang Q, Cosic I (2007) ECG R-R Peak detection on mobile phones. In: Engineering in Medicine and Biology Society, 2007. EMBS 2007. 29th Annual International Conference of the IEEE. pp. 3697–3700. doi:10.1109/IEMBS.2007.4353134.
  64. 64. Ligtenberg A, Kunt M (1983) A robust-digital QRS detection algorithm for arrhythmia monitoring. Computers and Biomed Res 16: 273–286.
  65. 65. Dokur Z, Olmez T, Yazgan E, Ersoy O (1997) Detection of ECG waveforms by neural networks. Med Eng Phys 19: 738–741.
  66. 66. Borjesson P, Palhm O, Sommo L, Nygards E (1982) Adaptive QRS detection based on maximum a posteriori estimation. IEEE Transactions on Biomedical Engineering 29: 341–351.
  67. 67. Fancott T, Wong D (1980) A minicomputer system for direct high speed analysis of cardiac arrhythmias in 24 h ambulatory ECG tape recording. IEEE Transactions on Biomedical Engineering 27: 685–693.
  68. 68. Keselbrener L, Keselbrener M, Akselrod S (1997) Nonlinear high pass filter for R-wave detection in ECG signal. Medical Engineering & Physics 19: 481–484.
  69. 69. Leski J, Tkacz E (1992) A new parallel concept for QRS complex detector. In: Proc. 14th Annual International Conference of the IEEE Engineering in Medicine and Biology Society. Paris, France, pp. 555–556.
  70. 70. Nygårds M, Hulting J (1979) An automated system for ECG monitoring. Computers and Biomedical Research 12: 181–202.
  71. 71. Srnmo L, Pahlm O, Nygards M (1982) Adaptive QRS detection in ambulatory ECG monitoring: A study of performance. In: Proc. IEEE Computers in Cardiology. pp. 201–204.
  72. 72. Christov I, Dotsinsky I, Daskalov I (1992) High-pass filtering of ECG signals using QRS elimination. Medical and Biological Engineering and Computing 30: 253–256.
  73. 73. Laguna P, Thakor N, Caminal P, Jane R (1990) Low-pass differentiators for biological signals with known spectra: application to ECG signal processing. IEEE Transactions on Biomedical Engineering 37: 420–425.
  74. 74. Thakor NV, Webster JG, Tompkins WJ (1984) Estimation of QRS complex power spectra for design of a QRS filter. IEEE Transactions on Biomedical Engineering 31: 702–6.
  75. 75. Thakor N, Zhu Y (1991) Applications of adaptive filtering to ECG analysis: noise cancellation and arrhythmia detection. IEEE Transactions on Biomedical Engineering 38: 785–794.
  76. 76. Chen H, Chen S (2003) A moving average based filtering system with its application to real-time QRS detection. In: Proc. IEEE Computers in Cardiology. pp. 585–588. doi: 10.1109/CIC.2003.1291223.
  77. 77. Sun Y, Suppappola S, Wrublewski T (1992) Microcontroller-based real-time QRS detection. Biomedical Instrumentation and Technology 26: 477–484.
  78. 78. Suppappola S, Sun Y (1994) Nonlinear transforms of ECG signals for digital QRS detection: a quantitative analysis. Biomedical Engineering, IEEE Transactions on 41: 397–400.
  79. 79. Trahanias P (1993) An approach to QRS complex detection using mathematical morphology. IEEE Transactions on Biomedical Engineering 40: 201–205.
  80. 80. Chu C, Delp E (1989) Impulsive noise suppression and background normalization of electrocardiogram signals using morphological operators. IEEE Transactions on Biomedical Engineering 36: 262–273.
  81. 81. Huang N, Shen Z, Long S, Wu M, Shih H, et al.. (1998) The empirical mode decomposition and hilbert spectrum for nonlinear and nonstationary time series analysis. Proc Roy Soc Lond A: 903–995.
  82. 82. Oukhellou L, Aknin P, Delechelle E (2006) Railway infrastructure system diagnosis using empirical mode decomposition and Hilbert transform. In: In: Proc. IEEE Int. Conf. Acoustics, Speech and Signal Processing (ICASSP), Toulouse, France. volume 3 , p. III.
  83. 83. Damerval C, Meignen S, Perrier V (2005) A fast algorithm for bidimensional EMD. Signal Processing Letters, IEEE 12: 701–704.
  84. 84. Zhou SK, Wang JT, Xu JR (1988) The real-time detection of QRS-complex using the envelope of ECG. In: the Annual International Conference of the IEEE Engineering in Medicine and Biology Society.
  85. 85. Nygards M, Sörnmo L (1983) Delineation of the QRS complex using the envelope of the ECG. Med Biol Eng Comput 21: 538–547.
  86. 86. Doyle TLA, Dugan EL, Humphries B, Newton RU (2004) Discriminating between elderly and young using a fractal dimension analysis of centre of pressure. International Journal of Medical Sciences 1: 11–20.
  87. 87. Burrus CS, Gopinath RA, Guo H (1998) Introduction to Wavelets and Wavelet Transforms. New Jersey: Prentice Hall.
  88. 88. Strang G, Nguyen T (1997) Wavelets and Filter Banks. Wellesley, MA, USA: Wellesley- Cambridge Press.
  89. 89. Elgendi M, Jonkman M, De Boer F (2009) R wave detection using Coiflets wavelets. In: Proc. IEEE 35th Annual Northeast Bioengineering Conference, Boston, MA, USA. pp. 1–2.
  90. 90. Krimi S, Ouni K, Ellouze N (2008) An approach combining wavelet transform and hidden Markov models for ECG segmentation. In: Proc. 3rd IEEE Int. Conf. Information and Communication Technologies: From Theory to Applications, Damascus, Syria. pp. 1–6.
  91. 91. Xue Q, Hu Y, Tompkins W (1992) Neural-network-based adaptive matched filtering for QRS detection. IEEE Transactions on Biomedical Engineering 39: 317–329.
  92. 92. Clifford GD, Azuaje F, McSharry P (2006) Advanced Methods And Tools for ECG Data Analysis. Norwood, MA, USA: Artech House, Inc.
  93. 93. Garca-Berdons C, Narvez J, Fernndez U, Sandovalm F (1997) A new QRS detector based on neural network. Springer-Verlag 1240: 1260–1269.
  94. 94. Dokur Z, lmez T (2001) ECG beat classification by a novel hybrid neural network. Computer Methods and Programs in Biomedicine 66: 167–181.
  95. 95. Coast A, Cano G (1989) QRS detection based on hidden Markov modeling. In: Proc. Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Seattle, WA, USA. pp. 34–35.
  96. 96. Coast D, Stern R, Cano G, Briller S (1990) An approach to cardiac arrhythmia analysis using hidden Markov models. IEEE Transactions on Biomedical Engineering: 826–836.
  97. 97. Cheng W, Chan K (1998) Classification of electrocardiogram using hidden Markov models. In: Proc. 20th Annual Int. Conf. IEEE Engineering in Medicine and Biology Society, Hong Kong. volume 1 , pp. 143–146.
  98. 98. Coast DA (1993) Segmentation of high-resolution ECGs using hidden Markov models. In: Proc. IEEE Conf. Acoustics, Speech, and Signal Processing (ICASSP-93), Minneapolis, MN, USA. volume 1 , pp. 67–70.
  99. 99. Kaplan D (1990) Simultaneous QRS detection and feature extraction using simple matched filter basis functions. In: Proc. IEEE Computers in Cardiology. pp. 503–506.
  100. 100. Ruha A, Sallinen S, Nissila S (1997) A real-time microprocessor QRS detector system with a 1-ms timing accuracy for the measurement of ambulatory HRV. IEEE Trans Biomedical Engineering 44: 159–167.
  101. 101. Hamilton PS, Tompkins WJ (1988) Adaptive matched filtering for QRS detection. In: Proc. Annual Int. Conf. IEEE Engineering in Medicine and Biology Society, New Orleans, LA, USA, Vol. 1 , pp. 14–18.
  102. 102. Yaosheng L, Ying X, Jiongfeng C, Zeguang Z (2008) A comparative study to extract the diaphragmatic electromyogram signal. In: Proc. Int. Conf. IEEE BioMedical Engineering and Informatics (BMEI 2008), Sanya, China, Vol. 2 , pp. 315–319.
  103. 103. Eskofier B, Kornhuber J, Hornegger J (2008) Embedded QRS detection for noisy ECG sensor data using a matched filter and directed graph search. In: Proc. 4th Russian-Bavarian Conference on Biomedical Engineering, Zelenograd, Moscow, Russia, pp. 48–52.
  104. 104. Belforte G, De Mori R, Ferraris F (1979) A contribution to the automatic processing of electrocardiograms using syntactic methods. IEEE Transactions on Biomedical Engineering 26: 125–136.
  105. 105. Ciaccio E, Dunn S, Akay M (1993) Biosignal pattern recognition and interpretation systems. IEEE Engineering in Medicine and Biology Magazine 12: 106–113.
  106. 106. Trahanias P, Skordalakis E (1990) Syntactic pattern recognition of the ECG. IEEE Transactions on Pattern Analysis and Machine Intelligence 12: 648–657.
  107. 107. Köhler B, Hennig C, Orglmeister R (2003) QRS detection using zero crossing counts. Progress in Biomedical Research 8: 138–145.
  108. 108. Mallat S, Hwang W (1992) Singularity detection and processing with wavelets. IEEE Transactions on Information Theory 38: 617–643.
  109. 109. Ayat M, Shamsollahi MB, Mozaffari B, Kharabian S (2009) ECG denoising using modulus maxima of wavelet transform. In: Proc. Annual Int. Conf. IEEE Engineering in Medicine and Biology Society, Minneapolis, MN, USA, pp. 416–419.
  110. 110. Moraes JCTB, Freitas MM, Vilani FN, Costa EV (2002) A QRS complex detection algorithm using electrocardiogram leads. In: Proc. IEEE Computers in Cardiology. pp. 205–208. doi: 10.1109/CIC.2002.1166743.
  111. 111. Chouhan V, Mehta S (2008) Detection of QRS complexes in 12-lead ECG using adaptive quantized threshold. International Journal of Computer Science and Network Security 8: 155–163.
  112. 112. Vijaya G, Kumar V, Verma H (1998) ANN-based QRS-complex analysis of ECG. Journal of Medical Engineering and Technology 4: 160–167.
  113. 113. Hu Y, Tompkins W, Urrusti J, Afonso V (1993) Applications of artificial neural networks for ECG signal detection and classification. Electrocardiology 26 (Suppl.): 66–73.
  114. 114. Strintzis M, Stalidis G, Magnisalis X, Maglaveras N (1992) Use of neural networks for electrocardiogram (ECG) feature extraction, recognition and classification. Neural Network World 3: 313–327.
  115. 115. Dobbs S, Schmitt N, Ozemek H (1984) QRS detection by template matching using real-time correlation on a microcomputer. Journal of Clinical Engineering 9: 197–212.
  116. 116. Ebenezer D, Krishnamurthy V (1993) Wave digital matched filter for electrocardiogram preprocessing. Journal of Biomedical Engineering 15: 132–134.
  117. 117. Di Virgilio V, Francaiancia C, Lino S, Cerutti S (1995) ECG fiducial points detection through wavelet transform. In: Proc. IEEE 17th Annual Conf. Engineering in Medicine and Biology Society, Montreal, Quebec, Canada, Vol.2 , pp. 1051–1052. doi: 10.1109/IEMBS.1995.579466.
  118. 118. Rao K (1997) Dwt based detection of R-peaks and data compression of ECG signals. IETE Jouranl of Research 43: 345–349.
  119. 119. Kadambe S, Murray R, Boudreaux-Bartels G (1999) Wavelet transform-based QRS complex detector. IEEE Transactions on Biomedical Engineering 46: 838–848.
  120. 120. Chen HC S, Chan H (2006) A real-time QRS detection method based on moving-averaging incorporating with wavelet denoising. Computer Methods and Programs in Biomedicine 82: 187–195.
  121. 121. Ghaffari A, Golbayani H, Ghasemi M (2008) A new mathematical based QRS detector using continuous wavelet transform. Computers & Electrical Engineering 34: 81–91.
  122. 122. Zheng H, Wu J (2008) Real-time QRS detection method. In: Proc. IEEE 10th Int. Conf. Real-time QRS detection method, e-health Networking, Applications and Services (HealthCom 2008), Singapore, pp. 169–170. doi:10.1109/HEALTH.2008.4600130.
  123. 123. Fard P, Moradi M, Tajvidi M (2007) A novel approach in R peak detection using hybrid complex wavelet (HCW). International Journal of Cardiology 124: 250–253.
  124. 124. Christov II (2004) Real time electrocardiogram QRS detection using combined adaptive threshold. Biomedical Engineering Online 3: 28.
  125. 125. Elgendi M, Mahalingam S, Jonkman M, De Boer F (2008) A robust QRS complex detection algorithm using dynamic thresholds. In: Proc. IEEE Int. Symp. Computer Science and its Applications (CSA'08), Hobart, Tasmania, Australia, pp. 153–158.
  126. 126. Elgendi M, Jonkman M, De Boer F (2009) Improved QRS detection algorithm using dynamic thresholds. International Journal of Hybrid Information Technology (IJHT) 2: 56–80.
  127. 127. Li C, Zheng C, Tai C (1995) Detection of ECG characteristic points using wavelet transforms. IEEE Transactions on Biomedical Engineering 42: 21–28.
  128. 128. Sameni R, Shamsollahi MB, Jutten C, Clifford G (2007) A nonlinear Bayesian filtering framework for ECG denoising. IEEE Transactions on Biomedical Engineering 54: 2172–2185.
  129. 129. Elgendi M, Jonkman M, De Boer F (2010) Frequency bands effects on QRS detection. In: Proc. International Joint Conference on Biomedical Engineering Systems and Technologies, BIOSIGNALS 2010, Valencia, Spain. pp. 428–431.
  130. 130. Elgendi M, Jonkman M, De Boer F (2009) P wave demarcation in electrocardiogram. In: Proc. IEEE 35th Annual Northeast Bioengineering Conference, Boston, MA, USA. pp. 1–2.
  131. 131. Elgendi M, Jonkman M, De Boer F (2009) Recognition of T waves in ECG signals. In: Proc. IEEE 35th Annual Northeast Bioengineering Conference, Boston, MA, USA. pp. 1–2.
  132. 132. Widrow B, Glover JJR, McCool J, Kaunitz J, Williams C, et al. (1975) Adaptive noise cancelling: Principles and applications. Proceedings of the IEEE 63: 1692–1716.
  133. 133. Donoho D (1995) De-noising by soft-thresholding. IEEE Transactions on Information Theory 41: 613–627.
  134. 134. Messer SR, Agzarian J, Abbott D (2001) Optimal wavelet denoising for phonocardiograms. Microelectronics Journal 32: 931–941.
  135. 135. Sharma L, Dandapat S, Mahanta A (2013) Kurtosis-based noise estimation and multiscale energy to denoise ECG signal. Signal, Image and Video Processing 7: 235–245.
  136. 136. Chiarugi F, Sakkalis V, Emmanouilidou D, Krontiris T, Varanini M, et al.. (2007) Adaptive threshold qrs detector with best channel selection based on a noise rating system. In: Proc. IEEE Computers in Cardiology. pp. 157–160. doi:10.1109/CIC.2007.4745445.
  137. 137. Elgendi M (2013) Fast QRS detection with an optimized knowledge-based method: Evaluation on 11 standard ECG databases. PLoS ONE 8: e73557.
  138. 138. Zidelmal Z, Amirou A, Adnane M, Belouchrani A (2012) QRS detection based on wavelet coefficients. Computer Methods and Programs in Biomedicine 107: 490–496.
  139. 139. Chouakri SA, Bereksi-Reguig F, Taleb-Ahmed A (2011) QRS complex detection based on multi wavelet packet decomposition. Applied Mathematics and Computation 217: 9508–9525.
  140. 140. Lee J, Jeong K, Yoon J, Lee JH (1996) A simple real-time QRS detection algorithm. In: Proc. 18th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 1996. Bridging Disciplines for Biomedicine. volume 4 , pp. 1396–1398. doi: 10.1109/IEMBS.1996.647473.
  141. 141. Afonso VX, Tompkins WJ, Nguyen TQ, Luo S (1996) Filter bank-based ECG beat detection. In: Proc. 18th Annual Int. Conf. IEEE Engineering in Medicine and Biology Society; Bridging Disciplines for Biomedicine, Amsterdam, Netherlands, Vol. 3: , 1037–1038. doi:10.1109/IEMBS.1996.652698.
  142. 142. Hii PC, Chung WY (2011) A comprehensive ubiquitous healthcare solution on an Android™mobile device. Sensors 11: 6799–6815.
  143. 143. Solar H, Fernndez E, Tartarisco G, Pioggia G, Cvetkovi B, et al.. (2012) A non invasive, wearable sensor platform for multi-parametric remote monitoring in CHF patients. In: Donnelly M, Paggetti C, Nugent C, Mokhtari M, editors, Impact Analysis of Solutions for Chronic Disease Prevention and Management, Publ: Springer, Berlin, Heidelberg, volume 7251 of Lecture Notes in Computer Science. pp. 140–147.
  144. 144. Muñoz Ramos O, Starostenko O, Alarcon-Aquino V, Cruz-Perez C (2013) Real-time system for monitoring and analyzing electrocardiogram on cell phone. In: Elleithy K, Sobh T, editors. Innovations and Advances in Computer, Information, Systems Sciences, and Engineering, Publ: Springer, New York, volume 152 of Lecture Notes in Electrical Engineering. pp. 327–338.
  145. 145. Hyejung K, Yazicioglu RF, Merken P, Van Hoof C, Hoi-Jun Y (2010) ECG signal compression and classification algorithm with quad level vector for ECG Holter system. IEEE Transactions on Information Technology in Biomedicine 14: 93–100.
  146. 146. Benitez D, Gaydecki PA, Zaidi A, Fitzpatrick AP (2001) The use of the Hilbert transform in ECG signal analysis. Computers in Biology and Medicine 31: 399–406.
  147. 147. Jasemian Y, Arendt-Nielsen L (2005) Evaluation of a realtime, remote monitoring telemedicine system using the bluetooth protocol and a mobile phone network. Journal of Telemedicine and Telecare 11: 256–260.
  148. 148. Jurik AD, Weaver AC (2008) Remote medical monitoring. Computer 41: 96–99.
  149. 149. Baig M, Gholamhosseini H, Connolly M (2013) A comprehensive survey of wearable and wireless ECG monitoring systems for older adults. Medical & Biological Engineering & Computing: 1–11.
  150. 150. Paoletti M, Marchesi C (2006) Discovering dangerous patterns in long-term ambulatory ECG recordings using a fast QRS detection algorithm and explorative data analysis. Computer Methods and Programs in Biomedicine 82: 20–30.
  151. 151. Martin-Clemente R, Camargo-Olivares JL, Hornillo-Mellado S, Elena M, Roman I (2011) Fast technique for noninvasive fetal ECG extraction. IEEE Transactions on Biomedical Engineering 58: 227–230.
  152. 152. Zhang CF, Tae-Wuk B (2012) VLSI friendly ECG QRS complex detector for body sensor networks. IEEE Journal on Emerging and Selected Topics in Circuits and Systems 2: 52–59.
  153. 153. Yan L, Shipeng L, Huifeng S (2011) Virtualized screen: A third element for cloud-mobile convergence. IEEE Multimedia 18: 4–11.