Comparative Study of Biomedical Physiological based ECG Signal heart monitoring for Human body Pushparaj1, Amod Kumar2and Garima Saini3 1,2,3 National Institute of Technical Teachers Training and Research, Sector-26, Chandigarh, 160019, India Abstract Cardiovascular disease the major challenges in the current 21st century in terms of health care and related to diagnostic developments. In this pandemic COVID-19 scenario, the cardiovascular disease or non-cardiovascular disease has been increased like cardiac arrest or silent heart attack. According to WHO has guidelines, it is set to reduce 25% overall mortality rate due to cardiovascular disease upto 2025 on the priority basis kept as prevention and control. Some techniques developed for heart rate estimation from multimodal physiological signals namely ECG, AB, and PPG, EEG, EMG and EOG etc. are the part of cardiovascular and non-cardiovascular signals have been reviewed. Keywords1 ECG, Arterial Blood Pressure (ABP), Electroencephalogram (EEG), Electro-oculogram (EOG) and Electromyogram (EMG) 1. Introduction Human body consists of various major parts and heart [1] plays the main important role for the body to be alive else it is dead body. If other part of the body is infected but not related to heart then with that problem the person can survive or alive with special care as per the infected disease. The heart is the main system of the body to pump the blood throughout the body due to contraction of muscles. Electrical signals [2,3] generated through the body, provoking variations in the electrical potential of skin surface. These signals (ECG related to heart) are captured with the help of electronics devices like electrode and related measurement equipment. At first these signals captured in 1887 for human ECG [4] signals, the electrical activity of the heart is recorded. But now we have so many approaches to measure or record ECG signals in three ways as ECG methods [3,5]- 1.in-the-person 2.on-the-person 3.off-the-person When a person report for treatment or surgery then if certain designed equipments are used inside the human body or ingested in the form of pills are well known as in-the-person category. If some electrodes or sensors are used over the human body or touch the outer skin for ECG measurements is well known to be on-the-person category and majorly used devices for ECG monitoring. In case of off-the-person category it is found that minimal skin contact or ECG without skin contact. It is based on the capacitive devices which measures the electric field changes induced by body allow ECG measurement at distances of 1cm or more even with the help of sensors[6,7]. Various tests are used to diagnose heart disease[6-8] and doctor enquires about personal and family medical history [8,9]and records the present and past symptoms. Based on the assessment enquiries doctor recommend for the related laboratory test. These tests may be invasive or non- invasive test. If laboratory tests are taken with the help of instrument inserted into the body is called non-invasive otherwise it is understood invasive tests[8-10]. International Conference on Emerging Technologies: AI, IoT, and CPS for Science & Technology Applications, September 06–07, 2021, NITTTR Chandigarh, India EMAIL: pushparajpal.ece20@nitttrchd.ac.in (A. 1); csioamod@yahoo.com (A. 2); garima@nitttrchd.ac.in (A. 3) ORCID: 0000-0002-9035-5870 (A. 1); 0000-0003-1177-3191 (A. 2); 0000-0001-6525-5192 (A. 3) ©2021 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). CEUR Workshop Proceedings (CEUR-WS.org) Types of Test: 1. Invasive test 2. Non-Invasive Test Invasive Test: It is a medical practice of cut skin and insert into the body opening for test using instruments while doing such types of test [10, 11] biopsy, endoscopy, and cryotherapy etc is required. Non-Invasive test: Non-invasive disease does not damage other organs of human body [2,5]. In this case when there is no need of tools that break the skin physically enter the body the skin. For instances [4, 5, 9] x-ray, CT Scan, MRI, ECG. Hearing aids, external splints and casts fall under this category as a related device. Cardiac Catheterization: common procedure that helps to diagnose heart disease [7,8] it can also be used to treat heart disease by opening blocked arteries with balloon angioplasty and stent placement. Figure 1: Layout of Physiological signals with ECG artifacts detection for cardiovascular/non- cardiovascular [7, 8] Some physiological [9-11] signal recognizes emotions due to its instant reaction and human as body structure is complex. So various related physiological [12-14] signals like HR, HVR, ECG, EEG, blood pressure etc. This type of signals travel into multichannel [15-19] as single channel has its own limitation. Figure 2: Anatomy of Heart (internet sources) [8-10] These signals can be classified through the classifier [16,18] as SVM-support vector machine. Multimodal physiological [10-12] signals are helpful to monitor continuously the health of patients either in ICU or at home at night. In high-tech hospitals [9, 11] there are the availability of intelligent bed system [11-13] for the patients whom automatically health monitors[12,14] and send the signal to the cloud server. Mainly focused signals are Electrocardiogram (ECG), Arterial Blood Pressure (ABP), Electromyogram (EMG), Electroencephalogram (EEG) and Photoplethysmogram (PPG)[11,12,15] etc. The signals directly helpful to monitor the heartbeat are related to cardiovascular and these are ECG, ABP and PPG etc and those physiological signals are indirectly connected or helpful are called non-cardiovascular signals such as EEG,EMG and EOG etc. Figure 3: Cardiac cycle in an ECG Signal [11, 14] In Figure 3, it represents the heart signal observed as ECG signal and shows some up-down lines like half sinewave and sharp peak like sawtooth waveform as monitored in electrons analog signals. Basically these have been termed as the English alphabets namely P; Q; R; S; T; U. These are the heart monitored ECG signals and termed as P-wave, T-wave, U-wave, and QRS complex wave. If we go through all the waves shows unique patterns. So these have defined as P-wave as atrial depolarization, T-wave as ventricular repolarization, U-wave as Papillary muscle repolarization and the last one is QRS complex wave as ventricular depolarization. 2. Literature Eduardo Jose da S. Luz, et al [1], authors described in their work about the current scenario state- of-art-methods of ECG-based automatic abnormalities heart beat classification, ECG signal processing heartbeat segmentation techniques, feature description methods and learning algorithms used. Tian Zhou, et al [3], results have compared the performance of different classifier, sensor fusion schemes, physiological modalities such as heart rate variability, electrodermal, and Electroencephalogram (ECG), (single vs multiple based on the prediction). In this case multisensors approach is used and got the individual performance which predicts cognitive workload levels of 83.2% of the time during basic and complex surgical skills tasks. Wei Wei, et al, [4], proposed a decision level weight fusion strategy for emotion recognition in multichannel physiological signals. Study focused on the four types of physiological signals includes EEG, ECG, Respiration Amplitude RA, and Galvanic Skin Response GSR. Various analysis domains have been used in physiological emotion features extraction. Secondly, adopted feedback strategy for weight definition, according to recognition rate of each emotion of each physiological signal based on Support Vector Machine, SVM classifier independently. Luis J.Mena et.al [5], developed wearable ECG monitoring device integrated with self-designed wireless sensor for ECG signal acquisition. It is native purposively designed smartphone application based on machine learning techniques, for automated classification of captured ECG beats from aged people. Deger Ayata, et al [8], proposed a novel emotion recognition algorithm from multimodal physiological signals for emotions aware healthcare systems. The physiological signals are collected from respiratory belt RB, photoplethysmography PPG, and fingertip temperature FTT sensors. The collection of the signal is done by wearable technologies. The results shows improved accuracy for arousal (69.86 to 73.08%) and from (69.53 to 72.18%) for valence. This is due to the multiple sources of physiological signals and their fusion increases the accuracy rate of emotion recognition. Xiaowei Zhang, et al [9], proposed a regularized deep fusion of kernel machine framework for emotion recognition based on multimodal physiological signals. Experimental study shows two benchmark datasets which improve the subjects-independent emotion recognition and compared to single-modal classifiers or other fusion methods. Chen Wang et al [10], proposed various methodologies to monitor or detection of heart rate (HR) using human face recording. This is done as there is motion in the cardiovascular activities or subtle color changes as the method used. Syem Ishaque et al [11], restricted the physiological signals ECG, EDA (Electrodermal Activity), PPG (photoplethysmography), and respiration (RESP) analyzed the heart rate variability between each heartbeat w.r.t time. Various analysis of various research work have been analyzed HRV associated with morbidity and stress. Detection of HRV in motion is far from perfect situations involving exercise or driving reported accuracy as high as 85% and as low as 59%. This can be improved with the advancement of machine learning techniques. Fanny Larradet et al [12], focused on the methodological issues in real life data collection in the wild based on physiological signals like emotion recognition. In the paper common technique used to induce emotions for the physiological dataset creation to compare with existing dataset of real-life applications. Author also proposed a set of categories visual tool called Graphical assessment of real- life application-focused emotional dataset (GARAFED) and compared with existing physiological datasets for EMSR in the wild. Quentin Meteier et al [13], discussed about the workload and task of drivers requested to regain the control of the car while performing the secondary task. In future, no longer of primary task for the driver as the automation of cars introduced. Measuring drivers workload continuously essential to support the driver and hence to limit the no. of accidents in takeover situations. For this purpose machine learning techniques is used to evaluate and classify the workload in real-time. Haoran Xu et al [17], prepared the system for the hospital general ward to monitor wirelessly (wearable and artificial intelligence) using physiological signals (ECG, RESP and transmit data wireless). The system consists of multi-sensor wearable device, database servers and user interface. Hospital needs to be highly integrated with the existing hospital information system and then explored the set of processes of physiological signals acquisition, storage, analysis and combination with the e- health records. Once the system is implemented in the general ward of the hospitals and starts collecting more than 1000cases from the clinics and so continues the whole system repeatedly to give the clinic feedback. With this system helps in reliable physiological monitoring of patients of general ward in the hospitals and generates more personalized pathophysiological signals/information to diagnose the disease and treatment form continuously monitored physiological data. Throughout the latest 20 years, beat screens (HRMs) have become a for the most part used guide for a grouping of employments [18]. The improvement of new HRMs has in like manner progressed rapidly during the latest two decades. Despite beat (HR) responses, look at has starting late revolved more around beat capriciousness (HRV). Existing approach for separating heart-related signs fall under two classes: wearable sensor advancements [19] and non-contact (fusing) structures. The conventional framework for checking beat surveyed the spikes of the potential made by heart at each weight beat for example electrocardiogram, or ECG. Physiological estimations, for example, ECG signals, are unbelievable considering the way that they are obliged by means of modified actuations of the autonomic unmistakable structure (ANS). In any case, existing sensors that can separate these signs require physical contact with subject's body, and causes impedance with the client experience. Various wearable advances, for example, a versatile ECG contraption, worn as a chest band holds the sensor focused over the heart, may move the readings remotely to the host PC. Unfortunately, the chest band is ungainly when worn for extended periods. It has a task to do in helpful and practice settings, at any rate it's unquestionably not a feasible choice for consistent use. Subsequently, wristband or a sharp have inadequacies [20]. 3. Inferences from Literature Eduardo Jose da S. Luz, et al [1], The depth and detailed discussion has been concise related to limitations, drawback methods found the literature which concludes and remarks the same and future challenge. Finally authors proposed the evaluation workflow process. Table 1 Typical feature of a normal ECG signal, with a cardiac frequency of 60beats per minute of healthy adult Feature Normal Value Normal Variation (+ve/-ve) P-wave 110ms 20ms PQ/PR interval 1160ms 40ms QRS width 100ms 20ms QT interval 400ms 40ms Amplitude of P 0.115mv 0.005mv Amplitude of QRS 1.5mv 0.5mv ST level 0mv 0.1mv Amplitude of T 0.3mv 0.2mv Source: [2] G. D. Clifford, F. Azuaje, P. McSharry, Advanced Methods and Tools for ECG Data Analysis, 1st ed., Artech House Publishers, 2006. Tian Zhou, et al [3], Authors discussed about the frame work leverages wireless sensors to monitor surgeons’ cognitive load and predicts their cognitive load and predict their cognitive states. Continuous multiple physiological modalities had simultaneously recorded for 12 surgeons performing surgical skill task on the validated da Vinci Skills Simulator. Table 2 Performance of six different classifiers and chance as base predictor Classifier Accuracy Precision Recall F1 Score Chance (flip a coin) 0.487 0.435 0.509 0.470 Support Vector Machine, SVM 0.773 0.810 0.641 0.716 Decision Trees, DT 0.714 0.694 0.641 0.667 Random Forest, RF 0.739 0.739 0.641 0.687 Extra Trees, ET 0.756 0.800 0.603 0.688 Gradient Boosting Trees, GBT 0.731 0.784 0.547 0.644 Adaboost, AB 0.681 0.653 0.603 0.627 Table 3 Performance of different modalities for Individual case Classifier Accuracy Precision Recall F1 Score EEG 0.782 0.755 0.755 0.755 EDA 0.479 0.395 0.321 0.354 HRA 0.504 0.442 0.434 0.438 MOV 0.664 0.710 0.415 0.524 EMG 0.580 0.526 0.566 0.545 Wei Wei, et al, [4], The experiments on the MAHNOB-HCI database show the highest accuracy. The result also provide evidence and suggest a way for further developing a more specialized emotion recognition system based on multichannel data using weight fusion strategy. Table 4 The size of each set of each emotion Emotion Sample Set Training Set Test Set Sadness 69 31 98 Happiness 86 31 55 Disgust 57 31 26 Neutral 112 31 81 Fear 39 31 8 Table 5 Detailed number of correctly recognized data and recognition rate under various physiological signals Signal/ Sadness Happiness Disgust Neutral Fear Emotion Test Set 38 55 26 81 8 EEG 25 (65.79%) 42 (76.36%) 13 (50.00%) 70 (86.42%) 5 (62.50%) ECG 21 (55.26%) 37 (67.27%) 18 (69.23%) 63 (77.78%) 4 (50.00%) RA 17 (44.74%) 30 (54.55%) 12 (46.15%) 51 (62.96%) 3 (37.50%) GSR 18 (47.37%) 29 (52.73%) 13 (50.00%) 56 (69.14%) 4 (50.00%) Identity Matrix 25 (65.79%) 43 (78.18%) 18 (69.23%) 72 (88.89%) 5 (62.50%) Diagonal Matrix 28 (73.68%) 47 (85.45%) 20 (76.92%) 75 (92.59%) 6 (75.00%) Table 6 MAHNOB –HCI database recorded signals Emotion Data Modalities Frequency 32-channel EEG 256 Hz 3-channel ECG 256 Hz 1-channel RA 256 Hz 1-channel GSR 256 Hz 1-channel Skin Temp (SKT) 256 Hz Face and Body Video 6 cameras, 60f/s Eye Gaze 60 Hz Audio 44.1KHz Luis J.Mena et.al [5], When tested on 100older adults, the monitoring system discriminated normal and abnormal ECG signals with a high degree of accuracy 97%, sensitivity 100%, and specificity 96.6% With further verification, the system could be useful for detecting cardiac abnormalities in the home environment and contribute to prevention, early diagnosis, and effective treatment of cardiovascular diseases, while keeping costs down and increasing access to healthcare services for older persons. Table 7 Performance summary of the ECG sensor device Technology Low-power microchip 8-bit AVR RISC-Based Microcontroller Supply Voltage 3.3V Input Impedance 100M-ohm Frequency Response Range 0.1Hz and Internal 8Mhz calibrated Oscillator Common Mode Rejection Ratio More than 90dB Gain 45 Sampling Rate 9.6KHz Data Bit-width 8bits Supply Voltage 3.3V Table 8 Total test performance of the mobile personal health monitor (PHM) system Technology Low-power microchip 8-bit AVR RISC-Based Microcontroller Supply Voltage 3.3V Input Impedance 100M-ohm Frequency Response Range 0.1Hz and Internal 8Mhz calibrated Oscillator Common Mode Rejection Ratio More than 90dB Gain 45 Sampling Rate 9.6KHz Data Bit-width 8bits Table 9 Total test performance of the mobile personal health monitor (PHM) system Evaluation Metrics Values (%) Sensitivity 100 Specificity 96.6 Accuracy 97 Precision 81.3 Deger Ayata, et al [8], The study shows a framework for emotion recognition using multimodal physiological signals from RB, PPG, and FTT. The paper concluded that the decision level fusion from multiple classifiers improved the accuracy rate of emotion recognition both for arousal and valence dimensions. Table 10 Selected classifier and optimized hyper-parameters for PPG, RB, and FFT signals Data source Classifier Feature set Window Window overlap duration RB Random Forest FS-14 3s Overlap PPG Random Forest FS-10 8s No-overlap FTT Random Forest FS-10 5s Overlap Table 11 Single modality and decision level fusion accuracy results Signal source Arousal Accuracy % Valence Accuracy % Respiratory Belt 71.32 70.62 PPG 68.59 70.23 FTT 69.68 67.73 Fused 73.08 72.18 Xiaowei Zhang, et al [9], Final fusion representation exhibits higher class-separability power for emotion recognition. Table 12 Data dimension output of each module of the RDFKM Framework for each modality during the training phase on DEAP and DECAF datasets Signal source Arousal Accuracy % Valence Accuracy % Modalities EEG,EMG,GSR,RES MEG,EMG,EOG,ECG Kernel 1200 x 1200 924 x 924 Embedding Layer 1200 x 100 924 x 100 FCN 64 x 1200 64 x 924 Intermediate Fusion layer 64 x 1200 64 x 924 Modalities EEG,EMG,GSR,RES MEG,EMG,EOG,ECG Chen Wang et al [10], Several other approaches have been proposed as signal processing and machine learning. The algorithms analyzed and compared with the public database MAHANOB-HCI. Table 13 Obtained performance for the best method at each stage Stage M (SD)/bpm RMSE/bpm Face video processing (extracted skin area) 5.34 (14.98) 15.05 Blood volume pulse signal extraction 4.09 (13.37) 13.56 (independent component analysis) Heart rate computing (peak detection) 3.01 (12.14) 12.23 Note: Mean error (M)-Performance measured; Standard Deviation (SD), RMSE(Root mean squared error) ; ρ-correlation co-efficient Quentin Meteier et al [13], The author concluded a high level of drivers’ mental workload can be accurately detected while driving in conditional automation based on 4-min recordings of respiration and skin conductance. Table 14 Best score for each combination of selected signals (with a segmentation level of 1) Selected signals Best Classifier Best Accuracy [Mean (SD)] EDA RF 0.73 (0.15) ECG RF 0.89 (0.09) RESP RF 0.88 (0.15) EDA+ECG RF 0.89 (0.09) EDA+RESP RF 0.89 (0.13) ECG+RESP MLP 0.94 (0.06) EDA+ECG+RESP SVC/MLP 0.92 (0.09) 4. Conclusion I have studied last 6years paper and their linked referenced papers too previously cited paper in the literature papers. From all those papers it has been concluded that various types of physiological signals, various parameters as per the papers have been reviewed. The physiological signal (ECG) has been used to detect for some abnormalities in heartbeat classifications, segmentation techniques, and feature descriptions and learning algorithms used. In some papers different datasets used, different modalities, signal extraction and classifiers used, discussion of wearable technologies, and surgical skill task validated with da Vinci Skills Simulator. With different modalities signals found results to individual accuracy and precision four types of physiological signals includes EEG, ECG, Respiration Amplitude RA, and Galvanic Skin Response GSR. Various analysis domains have been used in physiological emotion features extraction. 5. Acknowledgment We are thankful to the department of Electronics and Communication Engineering, National Institute of Technical Teachers Training and Research, Chandigarh for their interest in this work and useful comments to draft the final form of the paper. The support of AICTE QIP (Poly) Schemes department, Government of India is gratefully acknowledged. We would like to thank NITTTR Chandigarh for lab facilities and research environment to carry out this work. 6. References [1] Luz, E. J. D. S., Schwartz, W. R., Cámara-Chávez, G., & Menotti, D. (2016). ECG-based heartbeat classification for arrhythmia detection: A survey. Computer methods and programs in biomedicine, 127, 144-164 [2] Clifford, G. D., Azuaje, F., & McSharry, P. (2006). Advanced methods and tools for ECG data analysis (p. 12). Boston: Artech house [3] Tian Zhou, Jackie S. Cha, Glebys Gonzalez, Juan P. Wachs, Chandru P. Sundaram, and Denny Yu. 2020. Multimodal Physiological Signals for Workload Prediction in Robot-assisted Surgery. ACM Trans. Hum.-Robot Interact. 9, 2, Article 12 (January 2020), 26 pages. https://doi.org/10.1145/3368589. [4] Wei, W., Jia, Q., Feng, Y., & Chen, G. (2018), Emotion recognition based on weighted fusion strategy of multichannel physiological signals. Computational intelligence and neuroscience, 2018. [5] Mobile personal health monitoring for automated classification of electrocardiogram signals in elderly. Computational and mathematical methods in medicine, 2018. [6] Mena, L. J., Félix, V. G., Ochoa, A., Ostos, R., González, E., Aspuru, J.,& Maestre, G. E. (2018). [7] Bassett, D.R. Device-based monitoring in physical activity and public health research. Physiol. Meas.2012, 33, 1769. Sensors 2020, 20, 6778 14 of 16 [8] [8] Ayata, D., Yaslan, Y., & Kamasak, M. E. (2020). Emotion recognition from multimodal physiological signals for emotion aware healthcare systems. Journal of Medical and Biological Engineering, 40(2), 149-157. [9] Zhang, X., Liu, J., Shen, J., Li, S., Hou, K., Hu, B., ... & Zhang, T. (2020). Emotion recognition from multimodal physiological signals using a regularized deep fusion of kernel machine. IEEE transactions on cybernetics. [10] Wang, C., Pun, T., & Chanel, G. (2018). A comparative survey of methods for remote heart rate detection from frontal face videos. Frontiers in bioengineering and biotechnology, 6, 33. [11] Ishaque, S., Khan, N., & Krishnan, S. (2021). Trends in Heart-Rate Variability Signal Analysis. Frontiers in Digital Health, 3, 13. [12] Larradet, F., Niewiadomski, R., Barresi, G., Caldwell, D. G., & Mattos, L. S. (2020). Toward emotion recognition from physiological signals in the wild: approaching the methodological issues in real-life data collection. Frontiers in psychology, 11, 1111 [13] Meteier, Q., Capallera, M., Ruffieux, S., Angelini, L., Abou Khaled, O., Mugellini, E., ... & Sonderegger, A. (2021). Classification of drivers' workload using physiological signals in conditional automation. Frontiers in psychology, 12, 268 [14] Chen, L.Y.; Tee, B.C.K.; Chortos, A.L.; Schwartz, G.; Tse, V.; Lipomi, D.J.; Wong, H.S.P.; McConnell, M.V.; Bao, Z. Continuous wireless pressure monitoring and mapping with ultra- small passive sensors for health monitoring and critical care. Nat. Commun. 2014, 5, 1–10. [15] Yang, J.C.; Mun, J.; Kwon, S.Y.; Park, S.; Bao, Z.; Park, S. Electronic Skin: Recent Progress and Future Prospects for Skin-Attachable Devices for Health Monitoring, Robotics, and Prosthetics. Adv. Mater. 2019, 31, 1904765. [16] Lim, H.R.; Kim, H.S.; Qazi, R.; Kwon, Y.T.; Jeong, J.W.; Yeo, W.H. Advanced soft materials, sensor integrations, and applications of wearable flexible hybrid electronics in healthcare, energy, and environment. Adv. Mater. 2020, 32, 1901924. [17] Xu, H., Li, P., Yang, Z., Liu, X., Wang, Z., Yan, W., ... & Zhang, Z. (2020). Construction and application of a medical-grade wireless monitoring system for physiological signals at general wards. Journal of Medical Systems, 44(10), 1-15 [18] Boutry, C.M.; Beker, L.; Kaizawa, Y.; Vassos, C.; Tran, H.; Hinckley, A.C.; Pfattner, R.; Niu, S.; Li, J.; Claverie, J.; et al. Biodegradable and flexible arterial-pulse sensor for the wireless monitoring of blood flow.Nat. Biomed. Eng. 2019, 3, 47–57. [19] Bandodkar, A.J.; Molinnus, D.; Mirza, O.; Guinovart, T.; Windmiller, J.R.; Valdés-Ramírez, G.; Andrade, F.J.; Schöning, M.J.; Wang, J. Epidermal tattoo potentiometric sodium sensors with wireless signal transduction for continuous non-invasive sweat monitoring. Biosens. Bioelectron. 2014, 54, 603–609. [20] Huang, C.Y.; Chan, M.C.; Chen, C.Y.; Lin, B.S. Novel wearable and wireless ring-type pulse oximeter with multi-detectors. Sensors 2014, 14, 17586–17599.