Instrumentation & Measurement Magazine 24-6 - 39

[10]. Furthermore, several novel dedicated protocols are developed
for applications that require specific processing [11].
Regardless of the selected protocol, the employment of
wireless links involves some issues. As a matter of fact, a
power supply for the sensors should be provided for their
operation. This can be provided either by a battery or by an energy
harvesting system. The employment of a battery can limit
the system's operative life and increase its maintenance costs
due to the necessity for replacing / recharging periodically the
sensor batteries. The employment of the energy harvesting approach,
instead, makes the sensors virtually autonomous from
any external power source. However, it increases the cost of
the sensors themselves due to the additional electronics required,
and at the same time, it limits the type of sensors which
can be employed. In fact, typically the energy harvesting approach
is able to provide a quite small amount of power (up
to few milliwatts), and this could be insufficient for the proper
operation of several kinds of sensors. Of course, other alternative
approaches are also available; wireless sensors powered
by electric grid, power-line sensors, radio frequency powered
sensors (such as the RFID (Radio Frequency Identification
tags) and self-powered sensors [12] are some of them.
Therefore, the selection of the most suitable network is a
critical point in the design of a smart home system and should
be carefully carried out to find the optimal trade-off between
hardware and maintenance costs, performance and ease of
deployment.
Apart from the selected approach, the sensors and network
should guarantee suitable reliability, responsiveness and data
security for the specific application.
After measurements have been acquired by the sensors,
they are aggregated in some specific way and one or more data
receivers/collectors should be available in the monitored environment.
Such devices gather the sensor data and send them
to a central data processing unit which will process them detect
what event or activity is actually happening in the home.
In order to properly carry out such a task, it is required that the
data acquired by different sensors, often placed in different areas
of the home, are time-aligned, within few seconds. Indeed,
aggregation is typically carried out considering a data timestamp,
or in general, a time reference for each collected data
provided by the sensor within the data itself. A synchronization
protocol [13] is therefore necessary for the sensor network
to provide usable data.
The central data processing unit can be located in the home
itself (local processing) or can be located either on a remote
server or on a dedicated cloud infrastructure. In the latter case,
of course, an Internet connection should be available in the
home.
Usually, a statistical approach based on expert systems and
machine learning algorithms [14], [15] is used to process the
data and " understand " what is happening in the home. This
processing typically requires high computing capabilities and
is carried out by extracting specific features from the acquired
aggregated data, and by employing a suitable type of classifier,
deciding which among the expected events or behaviors
September 2021
is the one occurring with higher probability. Sometimes, data,
scheduled actions and, optionally, some sort of user feedback,
can be logged in a suitable storage, and they can be used to
continuously tune the decision algorithms with the aim of improving
the quality and the reliability of the classifier and the
action scheduling.
According to the detected situation, the central processing
unit schedules suitable actions and sends to the local actuator
controller some specific commands to carry out the scheduled
actions. The actuators, which typically use the same network
infrastructure of the smart sensors, operate on specific devices
to carry out the required action. As an example, they
can switch on and off lights or domestic appliances, or they
can change the setting of the conditioning system. As shown
in Fig. 1b, actuators, similar to smart sensors, embed several
blocks, like the network interface, power supply unit and processing
unit, a Digital-to-Analog Converter (DAC) and an
analog conditioning circuit for driving the actuator. In some
cases, the sensor and actuator are physically implemented in
the same device; therefore, they can share part of the hardware
resources, thus reducing the cost.
It should be noted, that the operation of expert systems and
classifiers commonly requires some " training data " to operate
properly. Furthermore, they are always subject to possible errors
and misunderstanding, especially if something changes
inside the environment: new people enter home, furniture or
sensors move from their original position, or there is just a significant
change in behavior or habit of the occupants. These
possible sources of errors can be negligible in most applications,
such as air conditioning or home appliance automation,
but they can become extremely important in applications
which can interfere with the well-being and the health of people
[16], such as telemedicine applications [17].
Sensing Devices Commonly Employed
in Smart Home and Healthcare
Applications
Understanding what is happening inside a home and the behaviors
of its occupants is not an easy task. This is particularly
true when more than one person is living inside the home. In
such a situation, the system not only has to correctly interpret
the data acquired by the sensors deployed in the home, it
has to determine the current behavior. It has also to recognize
who is actually performing the actions. In order to properly
accomplish this task, it is fundamental to acquire the most suitable
parameters from the environment and from the home's
occupants.
Therefore, sensors are one of most important devices employed
in such applications. Three basic types of sensors can
be deployed in smart homes: environmental sensors, wearable
sensors and infrastructure sensors. Fig. 2 shows an example of
smart home implementation, including the placement of different
kinds of sensors in the environment.
The most common sensor types are surely the environmental
ones. Such sensors are directly deployed in the
environment or attached to objects present in the home. They
IEEE Instrumentation & Measurement Magazine
39
Instrumentation & Measurement Magazine 24-6

Table of Contents for the Digital Edition of Instrumentation & Measurement Magazine 24-6
