Cloud based multivariate signal based heart abnormality detection

Abstract

This paper discusses the rate of heartbeat monitoring with the help of face tracking, extraction of forehead region and separation of blind source. Separation of blind source is applied for three RGB color channel. Independent Component Analysis (ICA) is a powerful tool for such acquisitions. There are mainly four ICA algorithms and these algorithms have been described in the literature. In this paper contribution of two main common ICA algorithms has been studied. These methods are compared to each other in terms of their ability to obtain independent signal from standard RGB signal of forehead region. These methods are Joint Approximate Diagonalization of Eigen matrices (JADE) and Fixed-point ICA (FastICA). Same RGB data set have been applied to these common ICA algorithms and compared with the results of blood volume pulse (BVP) sensor readings. Both methods provide equally consistent results. However, FastICA has shown better results for heart rate measurement compared to JADE.

Subject Classification:

• 92C50 Medical applications (general)

Keywords:

• Independent component analysis (ICA)
• Blood volume pulse (BVP)
• Blind source separation (BSS)
• Fixed point independent component analysis (FastICA)
• Standard deviation (SD)
• Root mean square error (RMSE)