Ear recognition based on local information fusion
Li Yuan
⇑
, Zhi chun Mu
School of Automation and Electrical Engineering, University of Science and Technology Beijing, Beijing 100083, China
article info
Article history:
Received 3 August 2010
Available online 20 October 2011
Communicate by M.S. Nixon
Keywords:
Ear recognition
Partial occlusion
Neighborhood preserving embedding
Sub-classifier fusion
abstract
Ears have rich structural features that are almost invariant with increasing age and facial expression vari-
ations. Therefore ear recognition has become an effective and appealing approach to non-contact biomet-
ric recognition. This paper gives an up-to date review of research works on ear recognition. Current 2D
ear recognition approaches achieve good performance in constrained environments. However the recog-
nition performance degrades severely under pose, lighting and occlusion. This paper proposes a 2D ear
recognition approach based on local information fusion to deal with ear recognition under partial occlu-
sion. Firstly, the whole 2D image is separated to sub-windows. Then, Neighborhood Preserving Embed-
ding is used for feature extraction on each sub-window, and we select the most discriminative sub-
windows according to the recognition rate. Each sub-window corresponds to a sub-classifier. Thirdly, a
sub-classifier fusion approach is used for recognition with partially occluded images. Experimental
results on the USTB ear dataset and UND dataset have illustrated that using only few sub-windows we
can represent the most meaningful region of the ear, and the mult i-classifier model gets higher recogni-
tion rate than using the whole image for recognition.
Ó 2011 Elsevier B.V. All rights reserved.
1. Introduction
As an emerging biometrics technology, ear recognition is
attracting more and more attention in biometrics recognition.
Human ears offer some distinct advantages over other biometric
modalities: they have a wealthy of structural features that are per-
manent with increasing age from about 8–70 years old, and they
are not affected by the expression variations (Burge and Burger,
2000). Ear image is smaller under the same resolution, which can
be favorable in some situations, such as the audio-visual person
authentication using speech and ear images for mobile phone
usage. According to the evaluations in Choras
´
(2006), the ear is a
kind of highly accepted biometrics, and subjects to be identified
feel more comfortable with ear images enrollment compared to
face images enrollment. Ear recognition is user-friendly and can
be used in non-intrusive recognition and surveillance scenarios.
Ears have played a significant role in forensic science for many
years (Nixon et al., 2010), especially in the United States, where an
ear classification system based on manual measurements has been
developed by Iannarelli, and has been in use for more than 40 years
(Iannarelli, 1989). The United States Immigration and Naturaliza-
tion Service (INS) has a form giving specifications for the photo-
graph that indicate that the right ear should be visible (INS Form
M-378 (6-92)). During crime scene investigation or airplane
crashes, earmarks are often used for identification (Alberink and
Ruifrok, 2007; Choras
´
, 2007). The history of using ear images or
ear prints shows their potential value for human identification
applications such as access control, security monitoring and video
surveillance (Hurley et al., 2008).
An ear recognition system usually involves ear detection, fea-
ture extraction and ear recognition/verification. As the first stage
of the ear recognition system, real-time ear detection and tracking
is a key component for the whole system. It mainly focuses on
detecting and tracking human ear from the input video images of
a scene and then returning the location and extent of each ear in
the image if one or more ears are present. The next step is to rep-
resent the ear by appropriate features and design effective classi-
fier. Most of the present ear recognition papers are focused on
this step. But in real scenarios, the performance of ear recognition
will be affected by illumination variation, pose variation and par-
tial occlusion.
In this paper, we deal with ear recognition under partial occlu-
sion. The main contribution of this paper is to propose a fusion
method for component based ear recognition based on our previ-
ous work (Yuan et al., 2010). In the previous work, we proposed
a multi-classifier fusion scheme for ear recognition with partially
occluded images. Top discriminative sub-classifiers ware com-
bined on the decision level for ear recognition. This paper has made
the following further improvements compared with our previous
work: (1) we add the ear detection step to form a more complete
ear recognition system, and give more comparisons with other
up-to-date references on ear recognition under partial occlusion;
(2) recent advances in ear recognition have been reviewed in this
0167-8655/$ - see front matter Ó 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.patrec.2011.09.041
⇑
Corresponding author. Tel.: +86 10 62334995.
E-mail address: yuanli64@hotmail.com (L. Yuan).
Pattern Recognition Letters 33 (2012) 182–190
Contents lists available at SciVerse ScienceDirect
Pattern Recognition Letters
journal homepage: www.elsevier.com/locate/patrec