Numerous authentication mechanisms have been reported
in literature. We focus our attention on authentication
mechanisms which are closely related or applicable to
wearable devices. Two dominant methods exist: the first
approach is based on vision sensors using either color or
depth. The second approach is based on sensors such as
accelerometers.
The first type of authentication mechanisms such as [1, 4,
5] rely on image sensors and employ color and depth
information to track or segment out hand or fingertip of a
person. This is followed by post processing to extract
trajectories and features such as position, velocity and
acceleration. Finally, signature is matched against pre-stored
signatures for authentication. Although these systems produce
accurate results, their usages are limited to a well-controlled
indoor environment. Also, they are computationally expensive
with requirement of depth in addition to color information. It
is important to note that most existing wearable devices do not
have any built-in depth sensor and would therefore incur
additional hardware cost, making this approach unsuitable for
resource constraint wearable devices.
The second type of authentication mechanisms such as [6,
7] are based on readily available accelerometers. This
approach requires additional hardware and circuit to capture
the hand movement and extract trajectory, which is then
transmitted to main device. More recent smart phones have
built in accelerometers and can perform gesture recognition
without additional hardware. However, accelerometer is a
much coarser device compared to camera and is capable only
to differentiate simple gestures. For authentication, the user is
required to remember lengthy gesture sequences, which are
not as straightforward and natural as gestures that are based on
the hand-written signature.
The most natural way to introduce a robust authentication
mechanism in wearable devices is to exploit the built-in
hardware among which color camera is the most common
sensor and is found in almost every wearable device. The
proposed approach is based on this theme