Papers by shankar venkatesan
Quantifiable fitness tracking using wearable devices
2015 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2015
Monitoring health and fitness is emerging as an important benefit that smartphone users could exp... more Monitoring health and fitness is emerging as an important benefit that smartphone users could expect from their mobile devices today. Rule of thumb calorie tracking and recommendation based on selective activity monitoring is widely available today, as both on-device and server based solutions. What is surprisingly not available to the users is a simple application geared towards quantitative fitness tracking. Such an application potentially can be a direct indicator of one's cardio-vascular performance and associated long term health risks. Since wearable devices with various inbuilt sensors like accelerometer, gyroscope, SPO2 and heart rate are increasingly becoming available, it is vital that the enormous data coming from these sensors be used to perform analytics to uncover hidden health and fitness associated facts. A continuous estimation of fitness level employing these wearable devices can potentially help users in setting personalized short and long-term exercise goals leading to positive impact on one's overall health. The present work describes a step in this direction. This work involves an unobtrusive method to track an individual's physical activity seamlessly, estimate calorie consumption during a day by mapping the activity to the calories spent and assess fitness level using heart rate data from wearable sensors. We employ a heart rate based parameter called Endurance to quantitatively estimate cardio-respiratory fitness of a person. This opens up avenues for personalization and adaptiveness by dynamically using individual's personal fitness data towards building robust modeling based on analytical principles.
We propose two novel Tensor Voting (TV) based supervised binary classification algorithms for N-D... more We propose two novel Tensor Voting (TV) based supervised binary classification algorithms for N-Dimensional (N-D) data points. (a) The first one finds an approximation to a separating hyper-surface that best separates the given two classes in N-D: this is done by finding a set of candidate decision-surface points (using the training data) and then modeling the decision surface by local planes using N-D TV; test points are then classified based on local plane equations. (b) The second algorithm defines a class similarity measure for a given test point t, which is the maximum over all inner products of the vector from t (to training point p) and the tangent at p (computed with TV): t is then assigned the class with the best similarity measure. Our approach is fast, local in nature and is equally valid for different kinds of decisions: we performed several experiments on real and synthetic data to validate our approach, and compared our approaches with standard classifiers such as kNN and Decision Trees.
Design and Development of Data Distribution Management Environment
SIMULATION, 2001
This paper describes the design and development of the DEVS/GDDM environment, a layered simulatio... more This paper describes the design and development of the DEVS/GDDM environment, a layered simulation environment that supports data dis tribution management and allows us to study space-based quantization schemes. These schemes aim to achieve effective reduction of data commu nication in distributed simulation. After a brief review of the space-based quantization scheme and an HLA-Interface environment, we discuss the design issues of the DEVS/GDDM environ ment. We analyze system performance and scalability of the space-based quantization scheme, especially with predictive and multiplex ing extensions, and empirical results for a ballis tic missiles simulation executing on the DEVS/ GDDM environment on NT networking plat forms. The results indicate the DEVS/GDDM environment is very effective and scalable due to reduced local computation demands and ex tremely favorable communication data reduction.
Approximation results for the two-layer constrained-via-minimization problem
Computer-Aided Design, 1989
ABSTRACT
Parameterized transform domain computation of the Hilbert Transform applied to separation of channels in Doppler spectra
2013 3rd IEEE International Advance Computing Conference (IACC), 2013
ABSTRACT
Video Stabilization, which is important for better analysis and user experience, is typically don... more Video Stabilization, which is important for better analysis and user experience, is typically done through Global Motion Estimation (GME) and Compensation. GME can be done in image domain using many techniques or in Transform domain using the well-known Phase Correlation methods which relate motion to phase shift in the spectrum. While image domain methods are generally slower (due to dense vector field computations), they can do global as well as local motion estimation. Transform domain methods cannot normally do local motion, but are faster and more accurate on homogeneous images, and are resilient to even rapid illumination changes and large motion. However both these approaches can become very time consuming if one needs more accuracy and smoothness because of the nature of the tradeoff. We show here that wavelet transforms can be used in a novel way to achieve a very smooth stabilization along with a significant speedup in this Fourier domain computation without sacrificing ac...
The University of Arizona definition and implementation of the DEVS framework is well known in th... more The University of Arizona definition and implementation of the DEVS framework is well known in the community of researchers that work on DEVS [1,2,6]. Not only does it provide an Object-Oriented implementation in C++ (and Java), but it also has a tight HLA connectivity (which was replaced by the HLA Interface developed at Lockheed Martin [4,5]). We present a faster and more efficient implementation of the University of Arizona DEVS here, which improves and clarifies many features of their implementation.
Low Dimensional Deep Features for facial landmark alignment
We propose a Low-Dimensional Deep Feature based Face Alignment (LDFFA) method to address the prob... more We propose a Low-Dimensional Deep Feature based Face Alignment (LDFFA) method to address the problem of face alignment “in-the-wild”. Recently, Deep Bottleneck Features (DBF) has been proposed as an effective channel to represent input with compact, low-dimensional descriptors. The locations of fiducial landmarks of human faces could be effectively represented using low dimensional features due to the large correlation between them. In this paper, we propose a novel deep CNN with a bottleneck layer which learns to extract a low-dimensional representation (DBF) of the fiducial landmarks from images of human faces. We pre-train the CNN with a large dataset of synthetically annotated data so that the extracted DBFs are robust across variations in pose, occlusions, and illumination. Our experiments show that the proposed approach demonstrates near real-time performance and higher accuracy when compared with state-of-the-art results on numerous benchmarks.
Multi-task learning (MTL) involves the simultaneous training of two or more related tasks over sh... more Multi-task learning (MTL) involves the simultaneous training of two or more related tasks over shared representations. In this work, we apply MTL to audio-visual automatic speech recognition(AV-ASR). Our primary task is to learn a mapping between audio-visual fused features and frame labels obtained from acoustic GMM/HMM model. This is combined with an auxiliary task which maps visual features to frame labels obtained from a separate visual GMM/HMM model. The MTL model is tested at various levels of babble noise and the results are compared with a base-line hybrid DNN-HMM AV-ASR model. Our results indicate that MTL is especially useful at higher level of noise. Compared to base-line, upto 7\% relative improvement in WER is reported at -3 SNR dB
Interspeech 2018, Sep 2, 2018
This paper demonstrates two novel methods to estimate the global SNR of speech signals. In both m... more This paper demonstrates two novel methods to estimate the global SNR of speech signals. In both methods, Deep Neural Network-Hidden Markov Model (DNN-HMM) acoustic model used in speech recognition systems is leveraged for the additional task of SNR estimation. In the first method, the entropy of the DNN-HMM output is computed. Recent work on bayesian deep learning has shown that a DNN-HMM trained with dropout can be used to estimate model uncertainty by approximating it as a deep Gaussian process. In the second method, this approximation is used to obtain model uncertainty estimates. Noise specific regressors are used to predict the SNR from the entropy and model uncertainty. The DNN-HMM is trained on GRID corpus and tested on different noise profiles from the DEMAND noise database at SNR levels ranging from-10 dB to 30 dB.
Interspeech 2019
Visual speech recognition or lipreading suffers from high word error rate (WER) as lipreading is ... more Visual speech recognition or lipreading suffers from high word error rate (WER) as lipreading is based solely on articulators that are visible to the camera. Recent works mitigated this problem using complex architectures of deep neural networks. Ivector based speaker adaptation is a well known technique in ASR systems used to reduce WER on unseen speakers. In this work, we explore speaker adaptation of lipreading models using latent identity vectors (visual i-vectors) obtained by factor analysis on visual features. In order to estimate the visual i-vectors, we employ two ways to collect sufficient statistics: first using GMM based universal background model (UBM) and second using RNN-HMM based UBM. The speaker-specific visual i-vector is given as an additional input to the hidden layers of the lipreading model during train and test phases. On GRID corpus, use of visual i-vectors results in 15% and 10% relative improvements over current state of the art lipreading architectures on unseen speakers using RNN-HMM and GMM based methods respectively. Furthermore, we explore the variation of WER with dimension of visual i-vectors, and with the amount of unseen speaker data required for visual i-vector estimation. We also report the results on Korean visual corpus that we created.
A Note on the Contractibility of Edges in 4-connected Maximal Planar Graphs
System and Method for the Detection of Abnormalities in a Biological Sample
Partition of Planar Flow Networks (Preliminary Version)
Focs, 1983
Some results on partitioning a planar graph into two halves
On epsilon partitioning of a planar graph
Fibonacci numbers are not context-free
Fibonacci Quarterly
ABSTRACT
Approximization Results for the 2-layer Constrained-via-minimization Problem
Contractible edges in 4-connected maximal planar graphs
Ars Combinatoria -Waterloo then Winnipeg-
ABSTRACT
Uploads
Papers by shankar venkatesan