Academics

PROJECTS

Human Communication and Machine Learning

Worked on recognizing and predicting human behavioral patterns and emotions using visual and vocal cues for my term project.
I am using Semaine database for my project. The database contains data pertaining to various shades of human emotions (visual and vocal).

Directed Research – Prof C. -C. Jay Kuo

Working on privacy in online social networks and the behavior of online users can be modeled using their context
Behavior models can be extracted using various signals obtained from the user's activity and the network dynamics

Directed Research – Prof Keith Jenkins

Worked on multiclass object recognition using biologically inspired features
Explored state of the art algorithms like HMAX, SIFT, SURF, pyramid kernel etc. Object recognition is the holy grail of computer vision and we need to work with models which are as close to the human visual system as possible.

Multimedia Data Compression

Developed and implemented a spatially scalable video coding technique using discrete wavelet transform in MATLAB for my term paper
Designed and implemented various algorithms for multimedia data compression in C/C++. The algorithms included Shannon-Fano coding, Huffman coding, Adaptive Huffman coding, Run length coding and its variants, QM coder, Lloyd-Max scalar and vector quantizer, Tree Structured Vector Quantization, JPEG rate control and compression quality factor, Post processing of JPEG images, Analysis and rate control in MPEG-1 and MPEG-2 video encoders

Digital Image Processing

Developed and implemented a Face Recognition system in MATLAB for my term paper
Designed and implemented various algorithms for image processing in C/C++/MATLAB. The algorithms included digital watermarking, image enhancement, edge detection, morphological processing, digital halftoning, geometric modification of images, texture analysis and optical character recognition

Mathematical Pattern Recognition

Developed and implemented a pattern classification system for my term project in MATLAB. The project included classifying forensic glass database and optical digit recognition (0-9) on NIST database (10,000 images).
The algorithms included various supervised and unsupervised learning techniques like Perceptron, Widrow-Hoff method, pseudoinverse method, support vector machines, K-nearest neighbors, transforming the data into another domain to obtain better results, dimensionality reduction techniques like Principal Component Analysis, Linear and quadratic Bayes classifier etc.

At NITK Surathkal, India (Undergraduate University)

Real-time Isolated Word Recognizer (speech recognition)

Developed an isolated word recognizer which could take in real time input and recognize the words
Parallel dynamic time warping technique was used to achieve faster real time recognition

Compressive Sensing for Audio and Images (image and audio compression)

Compressive is a technique for acquiring and reconstructing a signal utilizing the prior knowledge that it is sparse or compressible by sampling the signals at less than Nyquist sampling rate
Developed algorithms for compressive sensing of images and audion using linear and non-linear optimization techniques

JPEG encoder with variable image compression rate

Developed a JPEG encoder in C using adaptive huffman coding and user controlled rate-distortion trade-off

COURSEWORK

I graduated with a GPA of 4.0 from University of Southern California (USC). The courses I have taken at USC are listed below:

In addition to above, my directed research includes:

Prof C. -C. Jay Kuo - Working on privacy in online social networks and behavior modeling of online users using context-aware information .
Prof Keith Jenkins - Worked on multiclass object recognition using biologically inspired features.