Archive for the ‘Invited Talks’ Category

Talk: Semi-supervised Learning for Amazon Alexa

Thursday, November 21st, 2019
The IEEE Signal Processing Society, Bangalore Chapter and Department of Electrical Engineering, Indian Institute of Science, welcomes you to following talk.
Location and Date : MMCR, EE, Thursday Nov. 21, 4 pm (Coffee at 345pm).
Speakers : Dr. Sivaram Garimella and Kishore Nandury


Title: Semi-supervised Learning for Amazon Alexa.
State-of-the-art Acoustic Models (AM) are large, complex deep neural networks that typically comprise millions of model parameters. Deep neural networks can express highly complex input-output relationships and transformations, but the key to getting the best performance out of them is the availability of large amounts of matched acoustic data – matched to the desired dialect, language, environmental/channel condition, microphone characteristic, speaking style, and so on. Since it is both time consuming and expensive to transcribe large amounts of matched acoustic data for every desired condition, we leverage Teacher/Student based Semi-Supervised Learning technology for improving the AM. Our training leverages vast amount of un-transcribed data in addition to multi-dialect transcribed data yielding up to 7% relative word error rate reduction over the baseline model, which has not seen any unlabelled data.
Sri Garimella is a Senior Manager heading the Alexa Machine Learning/Speech Recognition group in Amazon, India. He has been associated with Amazon for more than 7 years. He obtained PhD from the Department of Electrical and Computer Engineering, Center for Language and Speech Processing at the Johns Hopkins University, Baltimore, USA in 2012. And Master of Engineering in Signal Processing from the Indian Institute of Science, Bangalore, India in 2006.
Kishore Nandury is an Applied scientist in Alexa ASR team in Amazon Bangalore. Prior to Amazon, he has worked in Intel, Sling media & NVIDIA graphics. He has obtained Masters degree in Signal Processing from Indian Institute of Science in 2005.
Host Faculty:  Sriram Ganapathy

Talk: Bayesian approaches for target tracking in radar applications

Friday, November 8th, 2019

CNRS, France, Indian Institute of Science, Bangalore, and IEEE Signal Processing Society Bangalore Chapter invite you to the following talk

Title: Some contributions to Bayesian approaches for target tracking in radar applications

Speaker: Prof. Eric Grivel, IMS Lab, Bordeaux, France

Venue: Golden Jubilee Seminar Hall, ECE Department, Indian Institute of Science Bangalore

Date and time: November 13, 2019; 4 PM – 5 PM.

Abstract: Detecting and tracking maritime or ground targets is one of the application fields for surveillance by airborne radar systems. More particularly, the purpose is to estimate the trajectories of one or more moving objects over time based on noisy radar measurements. When dealing with one target, one approach consists in using a Kalman filter by making an assumption on the type of the target motion and the parameters of the motion model. However, these assumptions are not necessarily well-suited to the situation. In addition, when dealing with maneuvering targets, the motion model may often change over time. Finally, false detections may appear and have a bad influence on the estimation of the target position. To address these issues, multiple-model algorithms, joint tracking and classification approaches and Bernoulli filtering can be considered. The purpose of this talk is to present some variants based on these concepts.

It should be noted that the methods that will be presented were developed with PhD students and a French colleague.

Biography of the speaker: Eric Grivel received his PhD in signal processing in 2000 in Bordeaux (France). He joined Bordeaux Institute of Technology (Bordeaux INP), in 2001 as an assistant professor  and then as a professor in 2011.  For more than 20 years, he has been with the Signal & Image research group at IMS lab (which is a joint research unit for the French National Center for Scientific Research CNRS, University of Bordeaux and Bordeaux INP). His research activities deal with statistical signal processing with applications in speech and audio processing, mobile communication systems, radar processing, GPS navigation and biomedical.

Talk on Latent Dirichlet Allocation

Tuesday, October 29th, 2019

The IEEE Signal Processing Society, Bangalore Chapter and Department of Electrical Engineering, Indian Institute of Science, welcomes you to following talk.

Location and Date : MMCR, EE, Thursday Oct. 31, 4-5 pm

Speaker : Dr. Hemant Misra (Vice President, Head of Applied Research, Swiggy)  

Abstract : Topic models such as Latent Dirichlet Allocation (LDA: have been used extensively in the last decade for tasks such as information retrieval, topic discovery, dimensionality reduction etc. In the current presentation, the application of LDA for the task of text-segmentation ( has been explained. Results on multiple datasets are shown to demonstrate the performance of the proposed LDA based system vis-a-vis other standard methods. The talk also uncovers challenge faced by the dynamic programming (DP) algorithm used in proposed LDA based segmentation and how it was overcome.

Talk will also cover some of the exciting things we are doing at Swiggy in the Applied Research team in the areas of speech, computer vision (CV) and natural language processing (NLP).

Speaker Bio : Dr. Hemant Misra is an active researcher in the areas of text and signal processing, speech/speaker recognition, machine learning, healthcare applications and education. He did his MS (1999) from IIT, Madras, and PhD from EPFL (2006). Then he held post-doc positions at Telecom ParisTech, University of Glasgow, and Xerox Research Centre Europe.  After having successful stints at Philips (Healthcare) Research, IBM’s India Research Lab and Citicorp Services India, currently Hemant is ‘VP – Head of Applied Research’ at Swiggy.

Host Faculty : Sriram Ganapathy (EE)

Talk: Large Scale Data Analytics for Airborne Imagery

Sunday, September 8th, 2019

Dept. of Electrical Communication Engineering and


IEEE Signal Processing Society Bangalore Chapter


invite you to the following seminar:


Title: Large Scale Data Analytics for Airborne Imagery

Speaker: Prof. Gaurav Sharma, Univ. of Rochester

Time and Date: 11 AM, September 10, 2019

Venue: Golden Jubilee Hall, ECE




The widespread availability of high resolution aerial imagery covering wide geographical areas is spurring a revolution in large scale visual data analytics. Specifically, modern aerial wide area motion imagery (WAMI) platforms capture large high resolution at rates of 1-3 frames per second. The sequences of images, which individually span several square miles of ground area, represent rich spatio-termporal datasets that are key enablers for new applications. The effectiveness of such analytics can be enhanced by combining WAMI with alternative sources of rich geo-spatial information such as road maps or prior georegistered images. We present results from our recent research in this area covering three topics. First, we describe a novel method for pixel accurate, real-time registration of vector roadmaps to WAMI imagery based on moving vehicles in the scene. Next, we present a framework for tracking WAMI vehicles across multiple frames by using the registered roadmap and a new probabilistic framework that allows us to better estimate associations across multiple frames in a computationally tractable algorithm. Finally, in the third part, we highlight, how we can combine structure from motion and our proposed registration approach to obtain 3D georegistration for use in application such as change detection. We present results on multiple WAMI datasets, including nighttime infrared WAMI imagery, highlighting the effectiveness of the proposed methods through both visual and numerical comparisons.


Speaker Biography


Gaurav Sharma is a professor in the Electrical and Computer Engineering Department and a Distinguished Researcher in Center of Excellence in Data Science (CoE) at the Goergen Institute for Data Science at the University of Rochester. He received the PhD degree in Electrical and Computer engineering from North Carolina State University, Raleigh in 1996. From 1993 through 2003, he was with the Xerox Innovation group in Webster, NY, most recently in the position of Principal Scientist and Project Leader. His research interests include data analytics, cyber physical systems, signal and image processing, computer vision, and media security; areas in which he has 52 patents and has authored over 200 journal and conference publications. He currently serves as the Editor-in-Chief for the IEEE Transactions on Image Processing. From 2011 through 2015, he served as the Editor-in-Chief for the Journal of Electronic Imaging and, in the past, has served as an associate editor for the Journal of Electronic Imaging, the IEEE Transactions on Image Processing, and for the IEEE Transactions on Information Forensics and Security. He is a member of the IEEE Publications, Products, and Services Board (PSPB) and chaired the IEEE Conference Publications Committee in 2017-18. He is the editor of the Digital Color Imaging Handbook published by CRC press in 2003. Dr. Sharma is a fellow of the IEEE, a fellow of SPIE, a fellow of the Society for Imaging Science and Technology (IS&T) and has been elected to Sigma Xi, Phi Kappa Phi, and Pi Mu Epsilon. In recognition of his research contributions, he received an IEEE Region I technical innovation award in 2008.

Special Lecture: Brain Inspired Automated Concept and Object Learning

Tuesday, September 3rd, 2019


Department of ECE, Indian Institute of Science

IEEE Bangalore Section

IEEE Signal Processing Society Bangalore Chapter

Welcome you to a 

Special Lecture


Title:  Brain Inspired Automated Concept and Object Learning: Vision, Text, and Beyond

Speakers: Vwani Roychowdhury (UCLA) and 

Thomas Kailath (Stanford) 

Venue: ECE Golden Jubilee Seminar Hall

             Department of  ECE, IISc

Day/ Date: Friday, 6 September 2019

Time: 3-5 pm

High Tea at 5pm


Brains are endowed with innate models that can learn effective informational and reasoning prototypes of the various objects and concepts in the real world around us. A distinctive hallmark of the brain, for example, is its ability to automatically discover and model objects, at multi-scale resolutions, from repeated exposures to unlabeled contextual data and then to be able to robustly detect the learned objects under various non-ideal circumstances, such as partial occlusion and different view angles. Replication of such capabilities in a machine would require three key ingredients: (i) access to large-scale perceptual data of the kind that humans experience, (ii) flexible representations of objects, and (iii) an efficient unsupervised learning algorithm. The Internet fortunately provides unprecedented access to vast amounts of visual data. The first part of this work will focus on our recent work that leverages the availability of such data to develop a scalable framework for unsupervised learning of object prototypes—brain-inspired flexible, scale, and shift invariant representations of deformable objects (e.g., humans, motorcycles, cars, airplanes) composed of parts, their different configurations and views, and their spatial relationships. We apply our framework to various datasets and show that our approach is computationally scalable and can construct accurate and operational part-aware object models much more efficiently than in much of the recent computer vision literature. We also present efficient algorithms for detection and localization in new scenes of objects and their partial views. The second part of this work will focus on processing large scale textual data, wherein our algorithms can create semantic concept-level maps from unstructured data sets. Finally we will conclude with the outlines of a general framework of contextual unsupervised learning that can remove many of the scalability and robustness limitations of existing supervised frameworks that require large amounts of labeled training sets and mostly act as impressive memorization engines. 

Short Bio:

Vwani Roychowdhury is a Professor of Electrical and Computer Engineering at UCLA and received his BTech and PhD degrees in Electrical Engineering from IIT Kanpur and Stanford University, respectively. Prof. Roychowdhury’s expertise lies in combining tools from a number of disciplines, including computer science, engineering, information theory, mathematics, and physics, and solving fundamental problems in multiple disciplines. His research interests have spanned a diverse set of topics related to combinatorics and theoretical computer science, Artificial Neural Networks, nanoelectronics and device modeling, quantum computing, quantum information and cryptography, the physics of information processing and computation, Bioinformatics, and more recently, brain inspired machine learning and brain modeling. He has published more than 250 journal and conference papers, and coauthored several books. He has mentored more than 25 Ph.D. students and 20 post-doctoral fellows and is always seeking collaborations with problem solvers and seekers. He also cofounded four silicon valley startups; one of these,, was founded in Jan. 2017, pioneered the unsupervised distillation of Concept Graphs from billions of documents, raised upwards of 45M US dollars in investment and was acquired in February 2017. 

Short Bio of Prof. Kailath:

Thomas Kailath received a B.E. (Telecom) degree in 1956 from the College of Engineering, Pune, India, and S.M. (1959) and Sc.D. (1961) degrees in electrical engineering from the Massachusetts Institute of Technology. He then worked at the Jet Propulsion Labs in Pasadena, CA, before being appointed to Stanford University as Associate Professor of Electrical Engineering in 1963. He was promoted to Professor in 1968, and appointed as the first holder of the Hitachi America Professorship in Engineering in1988. He assumed emeritus status in 2001, but remains active with his research and writing activities. He also held shorter-term appointments at several institutions around the world. 

His research and teaching have ranged over several fields of engineering and mathematics: information theory, communications, linear systems, estimation and control, signal processing, semiconductor manufacturing, probability and statistics, and matrix and operator theory. He has also co-founded and served as a director of several high-technology companies. He has mentored an outstanding array of over a hundred doctoral and postdoctoral scholars. Their joint efforts have led to over 300 journal papers, a dozen patents and several books and monographs, including the major textbooks: Linear Systems (1980) and Linear Estimation (2000). 

He received the IEEE Medal of Honor in 2007 for “exceptional contributions to the development of powerful algorithms for communications, control, computing and signal processing.” Among other major honors are the Shannon Award of the IEEE Information Theory Society; the IEEE Education Medal and the IEEE Signal Processing Medal; the 2009 BBVA Foundation Prize for Information and Communication Technologies; the Padma Bhushan, India’s third highest civilian award; election to the U.S. National Academy of Engineering, the U.S. National Academy of Sciences, and the American Academy of Arts and Sciences; foreign membership of the Royal Society of London, the Royal Spanish Academy of Engineering, the Indian National Academy of Engineering, the Indian National Science Academy, the National Academy of Sciences,India, the Indian Academy of Sciences, and TWAS (The World Academy of Sciences). 

In November 2014, he received the 2012 US National Medal of Science from President Obama “for transformative contributions to the fields of information and system science, for distinctive and sustained mentoring of young scholars, and for translation of scientific ideas into entrepreneurial ventures that have had a significant impact on industry.”

IEEE SP chapter talk: 14 June 2019 – Physics-Based Vision and Learning

Thursday, June 13th, 2019


Department of Electrical Engineering (EE), IISc and IEEE SPS Bangalore Chapter invite you to the following seminar:

Title: “Physics-Based Vision and Learning ”
(Joint Work with Yunhao Ba and Guangyuan Zhao)

Speaker: Dr. Achuta Kadambi

Achuta Kadambi

Host faculty : Dr. Chandra Sekhar Seelamantula

Time & Venue : 14th June, 2019, Friday on 4:00 PM at MMCR first floor, Electrical Engineering Department, IISc.

Abstract: Today, deep learning is the de facto approach to solving many computer vision problems. However, in adopting deep learning, one may overlook a subtlety: the physics of how light interacts with matter. By exploiting these previously overlooked subtleties, we will describe how we can rethink the longstanding problem of 3D reconstruction. Using the lessons learned from this prior work, we will then discuss the future symbiosis between physics and machine learning, and how this fusion can transform many application areas in imaging.

Biography: Achuta Kadambi is an Assistant Professor of Electrical and Computer Engineering at UCLA, where he directs the Visual Machines Group. The group blends the physics of light with artificial intelligence to give the gift of sight to robots. Achuta received his BS from UC Berkeley and his PhD from MIT, completing an interdepartmental doctorate between the MIT Media Lab and MIT EECS. Please see his group web page for research specifics:

Talk: August 27, 2015: Effective exploitation of long-term correlations for audio coding and networking

Wednesday, August 26th, 2015

The Department of Electrical Engineering
IEEE Signal Processing Society, Bangalore Chapter

cordially invite you to a lecture on

Title: Effective exploitation of long-term correlations for audio coding
and networking

Speaker: Dr. Tejaswi Nanjundaswamy, Postdoc fellow, UC Santa Barbara

Venue: Multimedia Classroom (MMCR), EE Department

Date and Time: August 27, 2015; 4.30 PM to 5.30 PM.

Abstract: A wide range of multimedia applications such as internet radio
and television, online media streaming, gaming, and high fidelity
teleconferencing heavily rely on efficient transmission of audio signals
over networks. The two main challenges for such transmission is delay
constrained compression, and dealing with loss of content due to noisy
channels. Constraints on delay means that the algorithms can only
operate on small block sizes (or frame lengths). Thus the key to
addressing these challenges is efficiently exploiting inter-frame
redundancies due to long term correlations. While well known audio
coders are effective in eliminating redundancies within a block of data,
and the only known inter-frame redundancy removal technique of employing
a long term prediction (LTP) filter is too simplistic, as it is
suboptimal for the commonly occurring polyphonic audio signals, which
contain a mixture of several periodic components, and also suboptimal
for speech and vocal content, which is quasi-periodic with small
variations in pitch period. Moreover the typically employed parameter
estimation technique is mismatched to the ultimate perceptual distortion
criteria of audio coding. Similarly even in loss concealment, none of
the existing techniques are designed to overcome the main challenge due
to the polyphonic nature of most music signals. This talk covers
contributions towards addressing all these shortcomings by employing
novel sophisticated filter structures suitable for a wide variety of
audio signals, with parameter estimation which takes into account the
perceptual distortion criteria for audio compression, and utilizes all
the available information for loss concealment.

Biography of the speaker: Tejaswi Nanjundaswamy received his B.E degree in electronics and
communications engineering from the National Institute of Technology
Karnataka, India, in 2004, the M.S. and the PhD. degree in electrical
and computer engineering from UCSB, in 2009 and 2013, respectively. He
is currently a post-doctoral researcher at the Signal Compression Lab in
UCSB, where he focuses on audio/video compression, processing and
related technologies. He worked at Ittiam Systems, Bangalore, India from
2004 to 2008 as Senior Engineer in the Audio group. He won the Student
Technical Paper Award at the Audio Engineering Society’s 129th
Convention. He was also a Best Paper Award Finalist at IEEE Workshop on
Applications of Signal Processing to Audio and Acoustics (WASPAA) 2011
and co-author of a Top 10% Award winning paper at IEEE International
Conference on Image Processing (ICIP) 2015.

Talk: May 29, 2015: Bayesian Models for Computational Rhythm Analysis in Indian Art Music

Saturday, May 16th, 2015

Department of Electrical Communication Engineering and
Department of Electrical Engineering,
Indian Institute of Science, Bangalore

invite you to a talk by
Ajay Srinivasamurthy
Music Technology Group, Universitat Pompeu Fabra
Barcelona, Spain

Bayesian Models for Computational Rhythm Analysis in Indian Art Music

Time & Date: 11 AM, May 29, 2015.
Venue: Golden Jubilee Hall, Department of Electrical Communication Engineering (ECE)

Rhythm in Indian art music (Carnatic and Hindustani music) is organized in the framework of tāla (or tāl). A tala consists of time cycles that provide a broad structure for repetition of music phrases, motifs and improvisations. Detecting different events such as the beats and the sama (downbeats) within a tāla cycle and tracking them through a piece of audio recording, referred to as meter inference, is an important computational rhythm analysis task in Indian Art Music. Useful in itself, it additionally provides a basis for further computational analyses such as structural analysis and extraction of rhythmic patterns. The talk mainly focuses on my recent work with Bayesian models for meter inference in Carnatic music. Efficient approximate inference in these models are presented to overcome the limitations of exact inference. Further,I will discuss extensions of these models that generalize to other music styles and different metrical structures.

Biography of the Speaker
Ajay is a PhD student at the Music Technology Group, Universitat Pompeu Fabra, Barcelona, Spain. He is a part of the CompMusic project led by Prof. Xavier Serra, where he works on rhythm related research problems in Indian Art Music and Beijing Opera. He is currently working towards developing signal processing and machine learning approaches for the analysis and characterization of rhythmic structures and patterns from audio recordings. Prior to joining UPF, he was a research assistant at the Georgia Tech Center for Music Technology, Atlanta, USA and worked at the mobile music startup Smule Inc. He has a masters in Signal Processing from the Indian Institute of Science, Bangalore, India and a B.Tech in Electronics and Communication Engineering from National Institute of Technology Karnataka, Surathkal, India.

Co-sponsors of the event:
IEEE Bangalore section
IEEE SP Society Bangalore Chapter

Special Lecture: 10 March 2015: Our recent work related to the design of filter banks with uncertainty considerations

Monday, March 9th, 2015







Dear All:

Speaker: Prof. V. M. Gadre, IIT Bombay, Mumbai, India
Title: Our recent work related to the design of filter banks with uncertainty considerations

Date/time: Tuesday, 10th March 2015, 3pm (coffee/tea at 2.45pm)
Venue: ECE GJH

Abstract: Two band filter banks have typically been designed with passband and stopband considerations and there is scant work on designing them with the uncertainty criterion. In this talk, we present some of the work that we have done in designing the two band filter bank with the uncertainty principle as a criterion.

Bio: Prof. Vikram Gadre received his Ph. D in Electrical Engineering from Indian Institute of Technology (IIT) Delhi in 1994, and thereafter has been working at IIT Bombay. He is currently the head of Centre for Distance Engineering Education. His research areas are communication and signal processing. He has taught 16 courses like signals and systems, control systems. He received ‘Excellence in Teaching’ award from IIT Bombay in 1999, 2004, and 2009. He has guided sponsored research projects for organizations like Tata Infotech, Texas Instruments, and 9 PhD, 65 M.Tech and Dual Degree, and 45 B.Tech projects. He has authored two books and has published 60 conference proceedings, 23 journal papers, and chapters in edited monographs.


Talks: 18 December 2014

Sunday, February 15th, 2015

IEEE Signal Processing Society, Bangalore Chapter, IEEE Bangalore Section,
and Supercomputer Education Research Centre, Indian Institute of Science

Invite you to the following talks:

1) Title: Beyond Mindless Labeling: *Really* Leveraging Humans to Build
Intelligent Machines

by Dr. Devi Parikh, Assistant Prof., Virginia Tech.


Human ability to understand natural images far exceeds machines today. One
reason for this gap is an artificially restrictive learning set up –
humans today teach machines via Morse code (e.g. providing binary labels
on images, such as “this is a horse” or “this is not”), and machines are
typically silent. These systems have the potential to be significantly
more accurate if they tap into the vast common-sense knowledge humans have
about the visual world. I will talk about our work on enriching the
communication between humans and machines by exploiting mid-level visual
properties or attributes. I will also talk about the more difficult
problem of directly learning common-sense knowledge simply by observing
the structure of the visual world around us. Unfortunately, this requires
automatic and accurate detection of objects, their attributes, poses, etc.
in images, leading to a chicken-and-egg problem. I will argue that the
solution is to give up on photorealism. Specifically, I will talk about
our work on exploiting human-generated abstract visual scenes to learn
common-sense knowledge and study high-level vision problems.


2) Title: Hedging Against Uncertainty in Machine Perception via Multiple
Diverse Predictions

by Dr. Duruv Bhatra, Assistant Prof., Virginia Tech.


What does a young child or a high-school student with no knowledge of
probability do when faced with a problem whose answer they are uncertain
of? They make guesses.

Modern machine perception algorithms (for object detection, pose
estimation, or semantic scene understanding), despite dealing with
tremendous amounts of ambiguity, do not.

In this talk, I will describe a line of work in my lab where we have been
developing machine perception models that output not just a single-best
solution, rather a /diverse/ set of plausible guesses. I will discuss
inference in graphical models, connections to submodular maximization over
a “doubly-exponential” space, and how/why this achieves state-of-art
performance on challenging Pascal VOC 2012 segmentation dataset.