Diabetes ECG dataset

In this study, ECG signals are recorded from 51 healthy (21 females and 30 males) and 35 diabetic (12 females and 23 males) patients age group of 20–70 years. Diabetes has been present in all patients for 3–20 years. The patients involved in this study have type-2 diabetes. Before the acquisition, the objective of the research and experimental procedure are explained, and consent from the subject as per approval of the Ethical Committee of the Indian Institute of Information Technology Design and Manufacturing (IIITDM), Jabalpur, has been taken. The data are recorded under the close supervision of a specialist. All reading took place in a comfortable supine posture after the morning breakfast in an air-conditioned room at a temperature of 27 ◦C. During the ECG recording, all of the participants sat in a calm supine posture for 20–50 min. The data are recorded at the Biomedical Signal Acquisition Laboratory, PDPM IIITDM, Jabalpur. For recording ECG signals, clarity medical equipment with BT traveler acquisition software has been used. The machine is portable, user friendly, operates entirely via PC/laptop, and is used for both research as well as commercial purposes. The equipment has different montage configurations for recording. Here, the signals have been acquired using an ECG montage arrangement by placing two gel conductive electrodes on the wrist with a high-quality electrodes gel at a sampling frequency of 256 Hz and 24-bit ADC resolution. In this work, six basic physiological parameters, namely, blood glucose (random), blood pressure, pulse, oxygen level (SpO2), weight, and height of the participants, are measured before recording the ECG data. Both normal and diabetic participants have normal blood pressure and no heart disease. To provide a real-time robust wearable system, we have not involved skin tone and adipose thickness in our experiments; 50-Hz mains interference artifacts were eliminated using a band-reject filter with a 50-Hz center frequency. Table I gives the average value of physiological characteristics for both normal and diabetic subjects. In this work, recorded data are segregated based on the blood glucose values; recording that includes random glucose value ≥160 mg/dL is considered as abnormal or DM subject and less than (<) 160 mg/dL is treated as normal. Each ECG record is segmented into 5-s fragments and marked as either DM or normal.

Kapil Gupta, Varun Bajaj A Robust Framework for Automated Screening of Diabetic Patient Using ECG Signals, IEEE Sensors Journal, 24 (22) 24222-24229, SCI, 2022, Q1,  10.1109/JSEN.2022.3219554 

Dataset is available on request by submitted EULA

Emotion EEG Data-set

In this study, four basic emotional states are examined during the audio-video stimulus. The 20 healthy subjects of age group (22.5 ±2.5) have participated in the experiments, who were the undergraduates of PDPM, Indian Institute of Information Technology, Design and Manufacturing, Jabalpur. For evoking the subjects emotional states, the audio-video clips of Indian films are used as elicitors. The clips are selected based on the following criterions. (i) A clip should be relatively short to avoid the contaminated data recording of multiple emotions. (ii) The clip should be understood by subjects without any explanation. (iii) The clip should elicit a single targeted emotion. To further assess that the selected clips evoked the emotion properly or not, the questioner session on 60 volunteers is performed. The selected volunteers are not included in the final EEG recordings to assess the efficacy of selected elicitors (clips) for inducing the emotions. In the questioner session, 30 movie clips are shown to volunteers and they were asked to rate these from 1–10 ratings in four emotions categories. The three higher mean rating clips for each emotion are selected. The experiment is begun with providing the experiment manual to subjects, which explained the experimental procedure and rating scales for assessing the evoked emotional state. The induced emotional states are following the 2-D valence-arousal emotion model, including happy (high arousal and high valence), fear (high arousal and low valence), sadness (low arousal and low valence), and relax (low arousal and high valence) states .

For acquiring EEG signals, the 24-channel EEG traveller is used, which has options of different sampling frequencies and montages for recording. Here, the transverse bipolar montage arrangement is built up according to the 10–20 electrode system for acquiring the EEG signals at 256 Hz sampling frequency. In the 10–20 system, each electrode is represented by a letter and a number, which respectively identifies the lobe and hemisphere where it has to be placed. Corresponding to an impulse-stimulus, the frontal and tem- poral regions of the skull are playing an important role to execute any reaction in the human brain. So, electrode positions FP1, FP2, F3, F4, F7, F8, T3, T4, T5, and T6 EEG-recordings are considered for emotion recognition. The reference and ground electrodes are placed over the forehead one above the other so the disturbance created due to the eye movement was not recorded. The following conditions are the same for the EEG recordings of each subject: the subjects do not have any neurological diseases, not consume alcohol and medicines, and subjects have taken proper sleep last night.

Publication in the Data-set 

[1] Anala Hari Krishna, Aravapalli Bhavya Sri, KYVS Priyanka, Sachin Taran, and Varun Bajaj Emotion classification using EEG signals based on tunable-Q wavelet transform   IET Science, Measurement and Technology 13(3) 375-380 SCI 10.1049/iet-smt.2018.5237 2018 2.23 Q2 IET

[2] Varun Bajaj, Sachin Taran and Abdulkadir Sengur, Emotion classification using Flexible analytic wavelet transform for electroencephalogram signals, Health Information Science and Systems 06:12 1-7 SCImago https://doi.org/10.1007/s13755-018-0048-y, 2018,Q1 Springer

[3] Sachin Taran, Varun Bajaj, Emotion recognition from single-channel EEG signals using a two-stage correlation and instantaneous frequency-based filtering method Computer Methods and Programs in Biomedicine 173 157-165 SCI https://doi.org/10.1016/j.cmpb.2019.03.015 2019 3.632 Q1 Elesvier

[4]Smith K. Khare, Anurag Nishad, Abhay Upadhyay, and Varun Bajaj, Classification of emotions from EEG signals using time-order representation based on the S-transform and convolutional neural network, Electronics Letters, Vol. 56 No. 25 pp. 1370âĂŞ1372, DOI: 10.1049/el.2020.2380, 1.232 SCI Q2 

[5]Smith K. Khare, and Varun Bajaj, An Evolutionary Optimized Variational Mode Decomposition for Emotion Recognition, IEEE Sensors Journal, 21(2), 2035-2042, 2020 doi: 10.1109/JSEN.2020.3020915. 3.07 SCI Q1 

[6] Smith K. Khare, Varun Bajaj, and G. R. Sinha. 2020. Adaptive Tunable Q Wavelet Transform Based Emotion Identification. IEEE Transactions On Instrumentation And Measurement, 69, 12, Dec. 2020, pp 9609 - 9617 doi:10.1109/tim.2020.3006611. 3.67 SCI Q1 

[7] Smith K. Khare, and Varun Bajaj. 2020. Time-Frequency Representation And Convolutional Neural Network-Based Emotion Recognition. IEEE Transactions On Neural Networks And Learning Systems, 1-9. doi:10.1109/tnnls.2020.3008938. 8.793 SCI Q1 

Data-set Available on request: