This lecture highlights role of Machine Learning in Modern Signal Processing Applications such as Driver-less Cars, Robotics, Smart Environment Monitoring, Healthcare etc.
Processing & Properties of Floor and Wall Tiles.pptx
Modern signal processing is dead without machine learning! 5th july 2020
1. Modern Signal Processing Application is Dead without Machine Learning!
G R Sinha, PhD
IEEE Senior Member, ACM DistinguishedSpeaker, IEEE DistinguishedSpeaker
Adjunct Professor IIIT Bangalore & Professor, Myanmar Institute of InformationTechnology Mandalay
Email: drgrsinha@ieee.org, ganeshsinha@acm.org, gr_sinha@miit.edu.mm
2. Motivation
Signal Processing
Role of Machine Learning
Signal Processing Applications
Concluding Remarks and Challenges
2Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
Lecture Outline
Source: https://etu.ru/en/study/masters-degree/information-systems-and-technologies
4. 4
Source: EEG/ERPAnalysis by Kamel Nidal & Aamir Malik, CRC, 2017.
Human Brain Structure
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
8. 8
A Signal is a function or an entity that carries some information about a phenomenon.
In Electronic Communication: voltage, current and EM wave are signals.
Audio, video, speech, image , sonar, Radar signals etc. are examples of signals.
Source:: Google Images
Signal
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
9. 9
Signal processing deals with analyzing, modifying, and synthesizing the signals and can be used
to improve transmission, storage efficiency and subjective quality. This can also be used to detect
components of interest in a signal.
Source: https://en.wikipedia.org/wiki/Signal_processing#/media/File:Signal_processing_system.png
Signal Processing
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
11. 11
Role of Machine Learning
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
12. 12
Machine learning is an application of artificial intelligence (AI) that empowers computers with the
ability to automatically learn and improve from experience.
The Learning Process begins with observations or data, such as examples, direct experience, or
instructions.
The primary aim is to allow the computers learn automatically without human intervention and it
can mimic human.
Source:
Google Image
Machine Learning (ML)
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
13. 13
Machine learning is also a data analytics technique use to teach computers as humans and
animals.
ML algorithms use computational methods to “learn” information directly from data. The
algorithms adaptively improve their performance as the number of samples available for
learning increases.
Deep Learning is most emerging form of machine learning.
ML has become an important tool for solving problems in areas, such as:
1. Image processing and computer vision
2. Computational biology, for tumor detection, drug discovery, and DNA sequencing
3. Energy production and load management
4. Automotive, aerospace, and manufacturing
5. Natural language processing, and many more
Why ML Matters
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
14. 14
Supervised learning—This trains a model on known input and output data so that it can predict future outputs.
Unsupervised learning---This finds hidden patterns or intrinsic structures in input data.
Source:
https://www.mathworks.com/discov
ery/machine-
learning/_jcr_content/mainParsys3/
discoverysubsection_1965078453/m
ainParsys/image_2128876021_cop.a
dapt.full.high.svg/1586543170363.sv
g
Types of ML
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
15. 15
This builds a model that makes predictions based on evidence in the presence of uncertainty,
by taking a known set of input data and known responses to the data; and training a model to
generate reasonable predictions for the response to new data.
Used when we have known data for the output, we are trying to predict.
Classification and regression are used to develop predictive models.
Classification Technique: Classification models classify input data into categories, in medical
imaging, speech recognition etc. For instance, if an email is genuine or spam, or whether a
tumor is cancerous/benign.
This is used if the data can be tagged, categorized, or separated into specific groups or classes.
Examples: SVM, Decision Trees, k-nearest neighbor, Naïve Bayes, Discriminant analysis, Neural
Networks etc.
Supervised ML
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
16. 16
Regression Technique—This predicts continuous responses, for example, changes in
temperature or fluctuations in power demand, like electricity load forecasting application.
Used when we are working with a data range or if the nature of the response is a real number,
such as temperature or the time until failure for a piece of equipment.
Commonly used methods are: Linear model, Nonlinear model, Regularization, Decision Trees,
Neural Network, Nero-fuzzy method.
A practical example: Clinicians want to predict whether someone will have a heart attack within
a year. They have data on previous patients, including age, weight, height, and blood pressure.
They know whether the previous patients had heart attacks within a year. Combining the
existing data into a model that can predict whether a new person will have a heart attack
within a year.
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
(contd..)
17. 17
This uses hidden patterns or intrinsic structures in data and is used to draw inferences from
datasets consisting of input data without labeled responses.
Clustering is the most common unsupervised learning technique, which is used for
exploratory data analysis to find hidden patterns or groupings in data.
Examples, gene sequence analysis, market research, object recognition etc.
Common techniques: k-means, hierarchical clustering, Gaussian mixture model, HMM, Self
organizing map, Fuzzy c-means clustering etc.
A practical example: If a cell phone company wants optimize the locations to deploy phone
towers then machine learning can help to estimate the number of clusters of people relying on
their towers. The algorithm helps to design the best placement of cell towers to optimize
signal reception for groups, or clusters, of their customers.
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
Unsupervised ML
18. 18
There is no best method which fits all requirements and finding the right algorithm is partly
just trial and error.
The algorithm selection also depends on the size and type of data we are working with; and
also on the insights we want to get from the data.
There is no robust method for all applications.
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
Choosing ML
19. 19
ML is used in daily life-such as Google Maps, Google assistant, Alexa, etc.
Source:
https://static.javatpoint.com/tutorial/mac
hine-learning/images/applications-of-
machine-learning.png
Applications of ML
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
Source:
https://www.excellentwebwor
ld.com/amazon-alexa-skill-
development/
20. 20
Image Recognition: One of the most common applications, used to identify objects, persons,
places, digital images, etc. Example, Automatic friend tagging in Facebook: It is based on the
Facebook project named "Deep Face," which is responsible for face recognition and person
identification in the picture.
Speech Recognition: Process of converting voice instructions into text. Example, Google
assistant, Alexa are examples
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
(contd..)
21. 21
Deep learning is a key technology behind driverless cars, in which a computer model learns to
perform classification tasks directly from images, text, or sound.
Models are trained by using a large set of data and neural network architectures that contain
many layers.
Deep Learning
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
Source: https://mc.ai/its-deep-learning-times-a-new-
frontier-of-data/
22. 22
Deep learning achieves recognition accuracy at higher levels than ever before, especially in
safety-critical applications like driverless cars.
Deep learning requires large amounts of labeled data. For example, driverless car development
requires millions of images and thousands of hours of video.
Deep learning requires substantial computing power. High-performance GPUs have a parallel
architecture that is efficient for deep learning, which is to reduce training time weeks to hours or
less.
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
(contd..)
Source:
https://foreignpolicy.com/20
19/07/04/who-will-win-self-
driving-future-book-review-
autonomous-vehicles/
23. 23
Automated Driving
Aerospace and Defense: To identify objects from satellites that locate areas of interest, and identify
safe or unsafe zones for troops.
Medical Research-Cancer research
Industrial Automation: To improve worker safety around heavy machinery by automatically
detecting when people or objects are within an unsafe distance of machines.
Electronics: Automated hearing and speech translation. For example, home assistance devices that
respond to voice.
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
Deep Learning Applications
28. 28
Mobile Robots &Vyom Mitra
Source: https://www.google.co.in/search?q=mobile+robots+in+isro+for+space+mission&tbm=isch&ved=2ahUKEwif_qjG3sjnAhX1XHwKHbNuDs8Q2-
cCegQIABAA&oq=mobile+robots+in+isro+for+space+mission&gs_l=img.3...37305.38793..40133...1.0..0.95.455.5......0....1..gws-wiz-img.3sPQn-DNb8w&ei=2zdCXt-
kA_W58QOz3bn4DA&bih=576&biw=1366#imgrc=oEa8yUEUaN00RM
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
31. 31
In Peripheral nervous system (PNS), sensory neurons detect stimuli from ears and
other sensors, and motor neurons carry out effector functions such as muscle contraction.
Interneurons carry information to and from the Central nervous system (CNS),
where integration of the information takes place.
Source: http://bio1152.nicerweb.com/Locked/media/ch48/information_processing.html
Nervous System
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
32. 32
The cell body contains a nucleus and numerous dendrites detect stimuli or receive signals from
other neurons and send the signals toward the cell body.
A single axon (joined to the cell body by the axon hillock) transmits signals away from the cell
body to other cells (neurons or effector cells). Small molecules
called neurotransmitters communicate the signals between neurons.
Neuron
Source: http://bio1152.nicerweb.com/Locked/media/ch48/information_processing.html
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
33. 33
BCI is essentially a communication pathway between human brain and an external device.
There are two ways that your brain can connect with an external device, invasively and non-
invasively.
The brain emits brainwaves when we perform an action because the brain cells (neurons)
interact with each other by sending and receiving small electrical signals, and we see those
electrical signals as brainwaves.
The brainwaves that the brain emits when performing an action are interpreted by the
computer.
Brain Waves
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
34. 34
Source: EEG/ERP Analysis by Kamel Nidal & Aamir Malik, CRC, 2017.
(contd..)
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
35. 35
EEG recording can be noninvasive or invasive. The noninvasive procedures use surface
electrodes and are safe & painless.
EEG can be measured using special electrodes with a typical diameter of 0.4 cm to 1.0 cm.
The electrodes are placed on the scalp with a paste (wet or dry), depending on the design of
electrodes.
EEG Recording
Source: https://www.wetalkuav.com/brain-implant-allows-users-to-control-drones-with-their-mind/eeg-recording/
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
36. 36
Concluding Remarks and Challenges
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020
37. 37
Signal Processing plays significant role in all modern applications of Science, Technology,
and Engineering.
Smart Signal Processing has tremendous scope of Research in the areas of Signal
Acquisition, Processing, Analysis, Classification and Soft Computing Tools.
Machine Learning plays most Important role in Modern Signal Processing applications.
Robustness is all time Challenge.
Concluding Remarks and Challenges
Modern Signal Processing Application is Dead without Machine Learning! @STTP-TEQIP-IIST G R Sinha 5th July, 2020