Feature Selection for Nonstationary Data: Application to Human Recognition Using Medical Biometrics

Abstract

Electrocardiogram (ECG) and transient evoked otoacoustic emission (TEOAE) are among the physiological signals that have attracted significant interest in biometric community due to their inherent robustness to replay and falsification attacks. However, they are time-dependent signals and this makes them hard to deal with in across-session human recognition scenario where only one session is available for enrollment. This paper presents a novel feature selection method to address this issue. It is based on an auxiliary dataset with multiple sessions where it selects a subset of features that are more persistent across different sessions. It uses local information in terms of sample margins while enforcing an across-session measure. This makes it a perfect fit for aforementioned biometric recognition problem. Comprehensive experiments on ECG and TEOAE variability due to time lapse and body posture are done. Performance of the proposed method is compared against seven state-of-the-art feature selection algorithms as well as another six approaches in the area of ECG and TEOAE biometric recognition. Experimental results demonstrate that the proposed method performs noticeably better than other algorithms.

ECG Biometric Authentication: A Comparative Analysis

“…Furthermore, what is common to the literature shown in the tables is that single-channel ECG (one lead of sensor) contains sufficient information to be discriminated between different subjects for the support of biometric recognition. There are different types of feature extraction modalities [8], [15], [18], [19], [27] and various classifiers [12], [21], [23], [24], [28] have been utilized for ECG-based recognition. In the following section, we summarize the methodologies based on the features and classification schemes.…”

“…This method aims to extract discriminative information from the ECG waveform without having to localize fiducial points. Several subsets of these non-fiducial features have been used in the literature such as autocorrelation [19], [27], discrete cosine transform [18], [44], [46], [47], NCN [8], [28], [35], and wavelet transform [8], [22], [40], [47]. Algorithms Based on Non-handcrafted Fiducial Features: Most handcrafted feature extraction approaches involve a pre-processing phase for preparing the ECG (e.g., a statistical analysis such as fiducial or non-fiducial features extraction).…”

“…During authentication, the score is compared to a predefined threshold, and the claimed identity is accepted, if the score is greater. There are different types of classifiers such as Euclidean distance, support vector machines (SVMs), dynamic time warping (DTW), and hamming distance utilized in the literature [8], [15], [18], [19], [21], [23], [24], [27], [35], [41], [44], [46]. ECG identification Most of the deep learning approaches for ECG biometric systems have been studied in the literature review used a fully connected layer and softmax for a classifier.…”

A Comprehensive Survey on ECG Signals as New Biometric Modality for Human Authentication: Recent Advances and Future Challenges

“…Generally, the non-fiducial feature-based algorithms extract discriminative information from the ECG waveform and eliminate the need for fiducial point localization for biometric recognition. Existing works have employed diverse subsets of these non-fiducial features like Wavelet transform [11], [302]- [304], autocorrelation [296], [297], [304], DCT [32], [298], [305], and normalizenonvoluted normalize (NCN) [302], [306], [307].…”

A Diagnostic Model for Identification of Myocardial Infarction from Electrocardiography Signals

“…On the other hand, many researchers have used numerous techniques to reduce the dimension of feature space and to determine the most relevant features (Yang and Zhidong 2017). To this end, mutual information (MILCA, mRMR, NMIFS) (Valenzuela et al 2013), genetic algorithms (GAs) (Silva Teodoro, Peres, and Lima 2017), multisession feature selection (MSFS) (Komeili et al 2017), and dynamic programming (DP) algorithm (Acir 2005) have been employed. In addition, it is clear that many popular classification methods (Cömert and Kocamaz 2017a), such as fractal analysis (Lai and Chan 1998), chaotic modeling (Owis et al 2002), bispectral coherence analysis (Khadra, Al-Fahoum, and Binajjaj 2005), and radial basis networks (Maglaveras et al 1998) In this study, we obtained several features from morphological and statistical domains to describe ECG signals after the preprocessing stage.…”