Scribd
Upload
32 views20 pages

BCSE401L Module 7 Slides

Uploaded by

fy2443246
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
32 views20 pages

BCSE401L Module 7 Slides

Uploaded by

fy2443246
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 20

BCSE401L: Internet of Things

Module-7: Applications of IoT


Dr. Raju Patel
Associate Professor
School of Electronics Engineering
[email protected]
Introduction
• In a smart environment, heterogeneous things send data to the gateway, fog node, or
cloud using wireless protocols.
• Encryption of data is carried out in a secured smart environment. Further, things data
analytic is used for automated decisions and making our life easier.
• The chapter discusses two use cases of smart environments, namely, smart healthcare,
and smart city.
• In particular, human activity recognition using wearable sensors and channel state
information in a smart healthcare system is presented.
• The channel state information provides a device-free solution and hence, better than
wearable sensors and image/video-based.
• The image and video-based approach breach the privacy of the user and require a large
transmit bandwidth.
• The chapter also presents smart parking, smart farming, and smart air pollution
monitoring for smart city applications.
• The smart system facilitates improved services to the users without any human
intervention.
Smart Healthcare
• Three applications are presented herein for a smart healthcare system: human
activity recognition using wearable sensors, Channel State Information (CSI)
based human activity recognition, and human health monitoring.

Figure 1: Human activity recognition using wearable sensors


Human Activity Recognition using Wearable Sensors

• The applications of HAR include remote patient monitoring, elderly people


monitoring, and transportation amongst others. The wearable biomedical sensors
are the following:
• Electromyography (EMG): To assess the health of muscles.
• Electroencephalograph (EEG): To detect abnormalities in the brain waves.
• Electrocardiogram (ECG): To record electrical activity of the heart.
• Accelerometer (ACC): To measure acceleration of the body.
• Electrodermal Activity (EDA): To measure changes in the electrical properties
of the skin.
• Respiration (RESP): To measures breathing rate.
• Force (FSR): To detect numbness of the body.
• Temperature (TEMP): To measure body temperature.
Human Activity Recognition using Wearable Sensors
• The other biomedical sensors are blood pressure sensor, Galvanic skin response sensor, airflow
sensor, sound generator, body position sensor, snore sensor, alert patient button, spirometer,
glucometer, and SPO2 pulse oximeter.
• The data from the things are sent using wireless technology such as Bluetooth,
• Wi-Fi, or LoRaWAN to the gateway, fog node or cloud for computational purpose securely. The
activities that can be detected are classified into several categories. Those can be (a) sitting, (b)
lying on a bed, (c) standing, (d) walking, (e) jogging, and (f) running to name a few.
• Wearable sensors based HAR is shown in Figure 1. The wearable sensors data change uniquely
according to the activity.
• For example, the heart rate and body temperature increase from rest to intense activity conditions.
• One can choose a biomedical kit depending on the requirements of specific biomedical signals,
sampling rate, computation and storage facilities, ease of use, and cost amongst others.
• We utilize data analytic technique for classification of different activities using wearable sensor
data.
Human Activity Recognition using Wearable Sensors
• If we use machine learning algorithm, we can extract features of wearable biomedical data for
different activities.
• For instance, we compute mean, variance, standard deviation, skewness, kurtosis, minimum and
maximum value, and many more as features.
• Notably, the deep learning algorithm automatically extracts features.
• In order to have multiple instances of a specific biomedical signal, the concept of windowing is
used.
• The original time-series data is windowed to get different overlapping segments of a single time-
series data.
• The feature is computed for each windowed signal.
• Subsequently, the features are fed to a classifier like kNN, naive Bayes, or support vector machine
(SVM) algorithm.
• Finally, the activities are recognized based on the wearable sensor data.
• Notably, we can use fewer sensors also, however, multiple sensors provide complementary
information and generally increases the activity recognition accuracy at the cost of computational
complexity.
Human Activity Recognition using WiFi Channel State Information

• HAR algorithms may be classified into three categories: wearable sensor based,
image/video-based and wireless link based.
• In the last section, we discussed wearable sensor-based activity recognition.
• However, wearable sensors may cause discomfort to humans.
• Image/video-based HAR consumes a lot of bandwidth during transmission and
breaches the privacy of users.
• Therefore, we discuss wireless link-based HAR in this section.
FIGURE 2: Human activity recognition using differential CSI
Human Activity Recognition using WiFi Channel State Information

• We use CSI between transmitter and receiver to recognize the activity of a person as shown in
Figure 2.
• Wi-Fi access points present in the environment act as a transmitter and smart device as a
receiver.
• The medium between atransmitter and a receiver is called a channel. The CSI is a complex
number and hence has both magnitude and phase.
• Generally, the CSI contains the carrier frequency offset and sampling frequency offset as per
literature.
• If we take the difference of two CSI for the same activity, the offset can be easily canceled.
• This soft computing approach enhances the activity recognition performance.
• The differential CSI is further fed to things data analytic for classification of different
activities. The different classified activities are sitting, lying on the bed, walk, pick up, stand
up, run and fall.
• One application of the sensed activities can be in a pacemaker for maintaining the required
heart rate using a controller.
Human Health Monitoring
• We detect anomalies in ECG and EEG data of a smart healthcare system.
• In particular, arrhythmia and seizure are detected using ECG and EEG signals, respectively.
• There are three main components of an ECG signal: P wave, QRS complex, and T wave. The
P wave, QRS complex, and R peak represent atrial depolarization, ventricular depolarization,
and ventricular repolarization, respectively.
• As we know QRS complex gives information about cardiac arrhythmias.
• The R peak is detected using the Pan-Tompkins algorithm.
• Subsequently, the classification of normal, ventricular, and super-ventricular beats are carried
out. The challenge is that normal beat and superventricular beats are similar.
• Therefore, 2-stage data analytic is used as shown in Figure 3. In the first stage, ventricular
beat is separated from the normal plus super-ventricular beats.
• In the next stage, the normal and super-ventricular beats are separated. There is a class
imbalance problem because of the abundance of normal beats and scarcity of abnormal beats.
• In order to balance the number of samples of both classes, upsampling of abnormal class data
is done to increase the number of samples of this class.
FIGURE 3: Arrhythmia detection using ECG signal
Smart City- Smart Parking Application
• In a smart parking system, the driver gets live information of vacant parking slots on
his/her smartphone application.
• The occupied or free slot in a parking area is detected using pressure or infrared sensors.
• The cost of an infrared sensor is low and transmits information using IoT protocols.
• When a car arrives at the entry, it is auto navigated to the nearest free parking slot as
shown in Figure 4.
• The location tracking of the vehicle is carried out using RSS based localization algorithm.
• Note that, in underground or multi-story complex buildings, GPS does not work well.
• Therefore, we need to resort to the low cost and ubiquitous RSS measurements for the
localization task.
• The location information helps in navigating the driver in a large underground parking
area.
• On exit, the number of free slots is incremented by one.
• If the blockchain technology is adopted, the system negotiates the parking fee and payment
is done securely without any human intervention.
FIGURE 4: Smart parking system
Smart Farming
• We use Meshlium 4G 868/900 AP based smart agriculture IoT vertical kit for data collection, data processing,
and drawing inferences. This helps in enhancing crop productivity, measuring climate change, and animal health
monitoring in real-time. Smart Agriculture PRO 868/900 PRO 5 DBi is a plug and sense kit and uses 6600 ma-h
rechargeable battery and an external solar panel. The smart agriculture kit measures the following:
• Temperature
• Humidity
• Pressure
• Soil temperature
• Soil moisture up to 1.5 m and 4.5 m
• Leaf wetness
• Wind speed using anemometer, wind vane, and pluviometer
• Solar radiation
• Similar to other smart systems, sensors measure soil and water (moisture) health, temperature, humidity, and
light. We can also make use of a drone for data gathering from a large agriculture field. The collected data is
further sent to the gateway, fog node or cloud using IoT protocols for processing as show in Figure 5. Things
data analytic is used for decision-making purposes. The location of animals grazing in the field is tracked using a
localization algorithm. The overall cost reduces and the productivity increases in smart farming.
FIGURE 5: Smart farming system
Smart Air Pollution Monitoring System
• Air pollution is caused by urbanization, transportation, industrial emission, dust,
wildfire, and burning of fossil fuels. The solid and liquid particles of gases are
suspended in air pollution.
• We use a smart pollution monitoring system for pollution control measures. We use air
quality index as shown in Figure 9.6 like
• Ozone (O3)
• Particulate Matter (PM) 2.5 and PM10
• Nitric dioxide (NO2)
• Sulphur dioxide (SO2) and
• Carbon monoxide (CO)
• We use Meshlium 4G 802.15.4 AP-based smart cities IoT vertical kit which is simply a
plug and sense kit. The power option is provided by 6600 mah rechargeable battery
and external solar panel. The air quality indices are estimated on hourly, day wise and
week wise basis. The data are sent to the gateway, fog node or cloud node using IoT
protocols.
Smart Air Pollution Monitoring System
• Subsequently, the following values are computed and mapped are
• 1. Excellent, range: 0 to 50
• 2. Fine, range: 51 to 100
• 3. Moderate, range: 101 to 150
• 4. Poor, range: 151 to 200
• 5. Very poor, range: 201 to 300 and
• 6. Hazardous, range: 301 to 500.
• Hence, we designate our environment with one of these classes and know
whether it is safe to live or not.
FIGURE 6: Smart air pollution monitoring system
FIGURE 7: IoT Architecture
FIGURE 7: IoT Architecture
FIGURE 8: Air Pollution Monitoring

You might also like