Dry ECG Dataset: Normal and Interference-Affected Records (NSIR Dataset)

The dataset consists of 149 ECG signals that were sampled at a frequency of 500 Hz using a single lead. 

The main objective is to identify and differentiate normal-recorded ECG signals from those affected by two common types of interference:

50/60 Hz power line interference .

 electrode contact noise caused by unstable movement.

The data was recorded at the LINS Laboratory within the University of USTHB in Algeria using the Orbital 90 dry electrode, which has a 25 mm conductive area diameter based on the three electrodes configurations to measure the potential difference between the two arms, with the right leg serving as the reference. In order to improve signal quality, a fourth-order Butterworth low-pass filter was utilized for its smooth frequency response and minimal phase distortion, followed by a notch filter to eliminate 50 Hz power line noise. These filtering processes significantly enhanced the ECG signal quality by reducing noise and interference while preserving important cardiac information. The recorded data was stored in CSV format using UART communication via a USB connected to a computer.

Dataset Summary

Type of Record 
Number of Records

well-recorded ECG
42

50/60 Hz power line interference
44

electrode contact noise
63

Total
149

Note :

You can find in the attachment a set of scalogram representations of the data with different wavelets, converted using the CWT technique.

These data were used in transfer learning and gave a remarkable result, highlighting the unnecessary need for a huge dataset to train a transfer learning algorithm when the data is well recorded and the classes are distinguished.

For more details please feel free to contact us :

[e-mail]

[e-mail]

[e-mail]

[e-mail]