IEEE Consumer Electronics Magazine - January/February 2020 - 18

Smart Home

Table 1. Technology on IoT applications.

intelligent applications, and easily interacts
with various Apps.4 However, the cost of Kinect is high and its physical size is relatively
large, compared with some home appliances.
In addition, pure Complementary MetalOxide-Semiconductor (CMOS) image sensors are
mature and can be significantly involved in IoTbased home applications. CMOS-based sensors
can be easily integrated to other circuits on a
system on-chip. Thus, the system price can be
reduced along with the size of the system.

PROPOSED SYSTEM
Design Considerations
The development of IoT-based applications
with object recognition has been realized in several
implementation strategies, as shown in Table 1.
The recognition of user gestures can be used to
derive many intelligent applications. The common
approach to this is the use of gyro sensors. However, these additional devices need some integration effort, such as the data glove. They are always
inconvenient and very expensive.5,6
Video surveillance systems can be integrated
into any machine being used as a home appliance
in an IoT environment. This basically follows the
client-server structure. Chacon-Murguia and
GonzalezDuarte7 developed an object detection
system, which identifies and tracks objects of
interest in visual surveillance media. The most
common risk for this technology is privacy.
Kinect mainly consists of IR sensors to detect
depth information and recognize hand gestures.8
On the other hand, Kinect sensor can construct a
skeleton, which points toward the use of Kinect
in more applications related to vision-based senlu et al.9,10 proposed an augmented
sors. Yelog
reality visualization interface to simultaneously
present visual and laser sensor information

further enhanced by stereoscopic viewing and
three-dimensional graphics.
Since depth information is always the fundamental information for gesture recognition, it can
be derived with a two-camera mechanism to create stereo information and include machine learning to recognize hand gestures.11 Stereo image
sensor modules are not large, but they are still relatively expensive, and the related computation is
still high. Kim et al.12 proposed a home appliance
control system based on a smart set-up box to provide some interaction. It is easier to integrate into
current home appliances. However, a more elaborate algorithm for vision tasks is required.

Overview of Proposed System
This article proposes an IoT-based intelligent
homecare appliance system (see Figure 2). A single-camera approach is applied in order to
reduce both the size and cost of the system. An
elaborate algorithm is also proposed to accomplish this application automatically. The proposed design has several advantages as follows:

automatic base height identification,
vision-based object recognition,
event trigger on sensor instead of server,
implemented as an embedded system.

ALGORITHM-LEVEL DEVELOPMENT
Base Height Identification
The detection of the object is influenced by
the position of the camera installed in an appliance. The initial setting requires that the camera
be set up at a height of 1.4 meters in order to
have an initial direction and angle. Once the camera angle has been saved, the initial step does not
need to be repeated. After this, users can move
the camera to different heights and use a white
object to allow the algorithm recognize the

IEEE Consumer Electronics Magazine

IEEE Consumer Electronics Magazine - January/February 2020

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - January/February 2020