20
Internet of Things (IoT)
The Google Glass headset provides a display and sensors to show the information found
on a smart watch or smartphone. Additionally, a camera in the device can be used not
only for taking pictures but also for computer vision tasks such as object recognition or, in
combination with the display, augmented reality where computer-generated graphics are
displayed as an overlay to highlight objects in the real world. Other headsets, such as the
Microsoft HoloLens, are expected to provide a similar augmented reality experience, serving as a way to improve both creation and consumption of multimedia. The HoloLens also
promises to improve interactions with real-world objects, such as providing step-by-step
instructions on repairing a light switch.
Headbands such as Muse or Emotiv provide brain–computer interfaces (BCI). The Muse
headband uses four EEG sensors, two on the forehead and two behind the ears, and three
additional reference sensors. Muse can also record blinking and jaw clenching. These sensors are used to detect and measure electrical activity in the brain. Currently, the Muse
headband is targeted as a mindfulness training device, helping users calm their minds.
However, developers have used the device as a BCI to control robots and perform research.
Similarly, the Emotiv makes use of even more EEG channels, accelerometer sensors, magnetometers, and gyroscope sensors to detect a wide variety of facial expressions, emotional
states, and mental commands, such as push, pull, left, or right.
2.2.4 e-Textiles: Smart Clothing, Smart Textiles, Smart Fabrics
e-Textiles comprise various clothing and fabrics with integrated electronics that allow them
to communicate, measure, and transform. Smart clothing, such as the smart shirts, developed by OMsignal and Hexoskin, provides in-depth biometric measurements, including
heart rate, breathing rate, step counts, calorie counts, and more. The smart clothing is used
by athletes to improve training and athletic performance, and by researchers to research
sleep patterns, stress levels, respiratory ability, and air pollution.
Smart fabrics such as Smart Skin, while currently is not wearable by humans, is used for
products on packaging lines to measure various factors such as pressure and orientation to
provide higher levels of quality assurance.
2.3 Review of Supervised Learning
In this chapter, we compare the results of many classification and clustering algorithms.
A brief overview of the algorithms used in this study is provided below.
2.3.1 Decision Tree
The objective of a decision tree algorithm is to determine the rules that can be used to
classify an instance based on the value of the instance’s attributes [2]. The possible combination of rules includes all partitions that can be obtained from the process of recursively
splitting the data [10], which may include multiple splits on the same attribute [10].
To determine the optimal attribute for splitting and the corresponding cut-point value,
impurity reduction is used. The impurity in each node is calculated with entropy measures, such as the Gini Index or the Shannon Entropy [10]. Impurity reduction is measured
as the difference between the impurity value for the parent node and the average impurity
value for the child nodes.
Internet of Things (IoT)
The Google Glass headset provides a display and sensors to show the information found
on a smart watch or smartphone. Additionally, a camera in the device can be used not
only for taking pictures but also for computer vision tasks such as object recognition or, in
combination with the display, augmented reality where computer-generated graphics are
displayed as an overlay to highlight objects in the real world. Other headsets, such as the
Microsoft HoloLens, are expected to provide a similar augmented reality experience, serving as a way to improve both creation and consumption of multimedia. The HoloLens also
promises to improve interactions with real-world objects, such as providing step-by-step
instructions on repairing a light switch.
Headbands such as Muse or Emotiv provide brain–computer interfaces (BCI). The Muse
headband uses four EEG sensors, two on the forehead and two behind the ears, and three
additional reference sensors. Muse can also record blinking and jaw clenching. These sensors are used to detect and measure electrical activity in the brain. Currently, the Muse
headband is targeted as a mindfulness training device, helping users calm their minds.
However, developers have used the device as a BCI to control robots and perform research.
Similarly, the Emotiv makes use of even more EEG channels, accelerometer sensors, magnetometers, and gyroscope sensors to detect a wide variety of facial expressions, emotional
states, and mental commands, such as push, pull, left, or right.
2.2.4 e-Textiles: Smart Clothing, Smart Textiles, Smart Fabrics
e-Textiles comprise various clothing and fabrics with integrated electronics that allow them
to communicate, measure, and transform. Smart clothing, such as the smart shirts, developed by OMsignal and Hexoskin, provides in-depth biometric measurements, including
heart rate, breathing rate, step counts, calorie counts, and more. The smart clothing is used
by athletes to improve training and athletic performance, and by researchers to research
sleep patterns, stress levels, respiratory ability, and air pollution.
Smart fabrics such as Smart Skin, while currently is not wearable by humans, is used for
products on packaging lines to measure various factors such as pressure and orientation to
provide higher levels of quality assurance.
2.3 Review of Supervised Learning
In this chapter, we compare the results of many classification and clustering algorithms.
A brief overview of the algorithms used in this study is provided below.
2.3.1 Decision Tree
The objective of a decision tree algorithm is to determine the rules that can be used to
classify an instance based on the value of the instance’s attributes [2]. The possible combination of rules includes all partitions that can be obtained from the process of recursively
splitting the data [10], which may include multiple splits on the same attribute [10].
To determine the optimal attribute for splitting and the corresponding cut-point value,
impurity reduction is used. The impurity in each node is calculated with entropy measures, such as the Gini Index or the Shannon Entropy [10]. Impurity reduction is measured
as the difference between the impurity value for the parent node and the average impurity
value for the child nodes.