Motion Artifact Detection Model using Machine Learning Technique for Classifying Abnormalities in Human Being
D.S Subhagya1, Keshavamurthy. C2

1D.S Subhagya, Research Scholar, Department  of Electronics and Communication Engineering, Jain University Bangalore, Karnataka.

2Keshavamurthy. C, Professor, Department  of Electronics and Communication Engineering, Shri Revana Siddeshwara Institute of Technology, Chikkajala, Bangalore.

Manuscript received on 05 March 2019 | Revised Manuscript received on 17 March 2019 | Manuscript Published on 22 March 2019 | PP: 334-340 | Volume-8 Issue-5S April 2019 | Retrieval Number: ES3438018319/19©BEIESP

Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open-access article under the CC-BY-NC-ND license (

Abstract: Obtaining an exact measurement of oxygen saturation (SpO2) using a finger-probe based pulse oximeter is dependent on both artifact-free infrared (IR) and red (R) Photo plethy smographic signals. However, in actual real-time environment condition, these Photo plethy smographic signals are corrupted due to presence of motion artifact (MA) signal that is produced due to the movement/motion from either hand or finger. To address this motion artifacts interference, the cause of the contamination of Photo plethy smographic signals by the motion artifacts signal is observed. The motion artifact signal is established to resemble similar to an additive noise. Motion and noise artifacts enforce constraints on the usability of the Photo plethy smographic, predominantly in the setting of sleep disorder detection and ambulatory monitoring. Motion and noise artifacts can alter Photo plethy smographic, resulting wrong approximation of physiological factors such as arterial oxygen saturation and heart rate. For overcoming research challenges, this manuscript present a novel hybrid approach for detection of artifacts. Firstly, this work present an accurate SpO2 measurement model. Secondly, present an adaptive filter and adaptive threshold model to detect artifact and obtain derivative of correlation coefficient (CC) for labelling artifacts respectively. Lastly, Enhanced Support Vector Machine (ESVM) Model is presented to perform classification. Experiment are conducted on both real-time and simulated dataset set. Our hybrid approach attain significant performance in term of accuracy, sensitivity, specificity and positive prediction.

Keywords: Artifact Detection, Machine Learning, PPG, SVM.
Scope of the Article: Machine Learning