ECG Processing & Analysis

Starting from filtering, detection and delineation, to compression and classification, we focus on all techniques of ECG signal processing. Our work encompasses high-frequency ECG analysis and ECG quality evaluation, for which, among others, we utilize motion sensor data. We also investigate all possibilities on how to utilize wearable and smart devices in monitoring human health and activity. This would include both selection and combination of suitable devices for data acquisition and analysis followed by interpretation. Smartphones, being a commonly available device, show a great potential in this respective area.


Human health and activity monitoring using smart and wearable devices

Monitoring human health and activity during the day using smart and wearable devices has become an emerging trend. Our team has also joined in on the effort to work in this advanced field of research. In light of this the team has decided to utilize all the great potential that smartphones, being a commonly available device, provide.
By using smartphones we are able to determine; a pulse rate by three diverse approaches (by using a camera [Němcová et al. 2017], a microphone or an accelerometer), oxygen saturation or blood pressure. We also monitor and classify human activity recorded during the day. To be able to perform such monitoring and classification we use diverse smartphone sensors or specialized devices, as for example, wrist-worn accelerometer or commercial smart wristband.
At the same time we compare the ability of these devices to distinguish between different human activities (walking, jogging, a stand up position, a sitting position or driving a car) and the accuracy they provide. We are also developing a robust pedometer. In our research we use methods like wavelet transform, empirical decomposition or machine learning. A long-term goal of our research group is for our algorithms to take form of smartphone applications. At the moment we are focusing on the following research topics:

ECG signal filtering and quality assesment

Large volumes of data are collected during a long-term ECG monitoring. Such volumes of data are stored and further analyzed. The informational content in signals may differ significantly in diverse segments. Contamination of ECG signals with noises may be generated by various environmental or biological resources (patient’s movement, breathing, sweating, surrounding electrical devices, electrode contact noise). Not addressed such noises may lead to a wrong interpretation and consequently diagnosis.
In order to maximize the information yield, we need to learn the signal quality and utilize this knowledge to pre-process and analyze different segments of ECG. To evaluate the ECG signal quality we use various advanced filtering methods combining a wavelet transform with a Wiener filter [Smital et al. 2013], an adaptive thresholding, a signal-to-noise ratio estimation and a signal segmentation based on various quality levels. Our current work involves optimization and development of these methods and designing novel algorithms to ensure even more efficient ECG signal quality evaluation.

ECG classification and analysis

In terms of the ECG analysis we apply our own sophisticated QRS complex detector [Smíšek et al. 2017]. The detector combines three separate detectors based on a phase transformation, a continuous wavelet transform and a Stockwell transform. In order to achieve a more detailed analysis we use two dimensional algorithms based on a wavelet transform [Hejč et al. 2015], [Vítek et al. 2009]. To be able to determine an accurate clinical diagnosis based on ECG recordings it is essential to first classify individual cardiac cycles and rhythms from collected data. Due to the large volumes of data, especially in a long-term ECG recording, it is not possible to only analyse the data manually but automatic classification methods have to be applied as well.
Currently, there do not exist such algorithms and software that would be able to evaluate the recordings without errors. Our team is working, for example, on a P-wave detection [Maršánová et al. 2018]. The determination of the P-wave location permits an easier classification of large volumes of pathologies (2nd and 3rd degree AV block, ventricular extrasystoles, atrial fibrillation, supraventricular tachycardia). We also work on designing novel algorithms for a detection of bundle-branch block. We also made P wave annotations for MIT-BIH Arrhythmia Database [Physionet] and new annotations and recommendations for CSE Database [Maršánová et al. 2016].

ECG compression

ECG signal compression with loss declines the volume of data. This leads to a more efficient data archiving and to a faster and less energy-intensive data transfer. Such features are especially beneficial in the field of telemedicine with the common use of wireless ECG monitoring systems. In order to compress data we use the wavelet transform and a SPIHT algorithm [Hrubeš et al. 2008] and design new fractal-based algorithm. An essential part of the compression process is an evaluation of signal quality after compression followed by a reconstruction. We also developed several novel methods in this field of focus (for example, based on ECG measuring or dynamic time warping) [Němcová et al. 2018]. We also proposed a combination of already published methods for a complex view on the perception of signals.


Follow Us

Get the latest news about our teaching and learning activities or get excited about scientific discoveries in biomedical engineering.

Partner With Us

We are committed to promote collaboration across biomedical disciplines creating mutual benefits with companies and research organizations. Explore opportunities to engage with our research teams, or find out more about our current research partners.


Donate to support research and education that translates into improved healthcare and people’s wellbeing. With a number of giving options, you can support the area you care about most. Gifts of any size help expand our positive impact throughout the society.