Deep Arm/Ear-ECG Image Learning for Highly Wearable Biometric Human Identification

Abstract

In this study, to advance smart health applications which have increasing security/privacy requirements, we propose a novel highly wearable ECG-based user identification system, empowered by both non-standard convenient ECG lead configurations and deep learning techniques. Specifically, to achieve a super wearability, we suggest situating all the ECG electrodes on the left upper-arm, or behind the ears, and successfully obtain weak but distinguishable ECG waveforms. Afterwards, to identify individuals from weak ECG, we further present a two-stage framework, including ECG imaging and deep feature learning/identification. In the former stage, the ECG heartbeats are projected to a 2D state space, to reveal heartbeats' trajectory behaviors and produce 2D images by a split-then-hit method. In the second stage, a convolutional neural network is introduced to automatically learn the intricate patterns directly from the ECG image representations without heavy feature engineering, and then perform user identification. Experimental results on two acquired datasets using our wearable prototype, show a promising identification rate of 98.4% (single-arm-ECG) and 91.1% (ear-ECG), respectively. To the best of our knowledge, it is the first study on the feasibility of using single-arm-ECG/ear-ECG for user identification purpose, which is expected to contribute to pervasive ECG-based user identification in smart health applications.

Sleep Stage Classification Using Time-Frequency Spectra From Consecutive Multi-Time Points

“…Deep learning is one popular kind of artificial intelligence algorithm. It has been widely applied in many fields and has achieved many remarkable results (Cai et al, 2016;Kim et al, 2017;Liu et al, 2017Liu et al, , 2018Sharma et al, 2017;Zhang and Zhou, 2018). Recently, studies of deep learning-based automatic sleep stage classification have also frequently been published (Fiorillo et al, 2019).…”

Off-Person ECG Biometrics Using Spatial Representations and Convolutional Neural Networks

“…The final image is obtained by an initial appropriate scale adjustment followed by binarization of the resulting trajectory. Much similar to [37], each time-delayed version of the (filtered, R-peak centered, normalized, fixed-length) ECG segment is first scaled into a common range, then a 200x200 pixels binary image is obtained by setting black color on all pixels hit by at least one point of the phase space trajectory, while labeling by white color all the rest of the pixels. As a consequence, the subsequent CNN based classifier may use single channel bidimensional inputs (as analyzed in [29]) or three (generally, multiple) channel inputs obtained by projecting the 3D (generally multidimensional) representation onto corresponding orthogonal planes.…”

“…The topology of the convolutional neural network is indicated in Table 2 and is much similar to the architectures used in [37], [40]. The actual structure resulted after repeated experiments using various dimensions of the convolution filters and number of neurons in the fully connected layers.…”

Unobtrusive and Continuous BCG-Based Human Identification Using a Microbend Fiber Sensor

“…Several algorithms have been developed in the literature for ECG-based human identification [7], [15], [18]- [20]. Time domain ECG features, saying the normalized distance between the two fiducials, are identified from P, R, and T complexes to identify individuals [18].…”