Research Projects

Mining of Personal Multimedia

Amount of multimedia data accumulated by users grows exponentially fast. In this project we address a problem of how to improve search, browsing and other interaction with this media data. Recently we developed a content-based image retreival framework that targets specifically personal foto collections (SPIE 2006, ACM MIR 2006). See recent demo at Intel R&D Day.

SoC design for Multimedia and Communication Applications

We apply a platform-based design methodology and transaction level simulation approach to analyze end-to-end system performance with accurate modeling of SW an HW components and Power consumption.

Distributed Digital Signal Processing (DDSP) Project

The goal of the project is to transform a network of heterogeneous, distributed computing platforms into an array of audio/video sensors and actuators capable of performing complex DSP tasks such as distributed beamforming, audio rendering, audio/visual tracking, camera array processing. In particular, this requires putting distributed heterogeneous computing platforms with audio-visual sensors into a common time and space coordinate system. Towards that goal ad-hoc synchronization, I/O calibration, sensor/actuator localization, joint audio-video processing, distributed source coding as well as distributed signal processing in general are investigated from a system hardware-software angle. The DDSP Project also explores the vast dimension and opportunities offered by the rich media in multiple streams, from multiple channels of audio to audio combined with visual data, and/or video and synthetic data.

Scalable Video Multicast of Real-Time Video over Wireless LANs

We developed a framework for distributing video over heterogeneous packet wireless channels to multiple users. MPEG4 Fine Granularity Salability (FGS) was combined with Forward Error Control (FEC) coding scheme. The proposed solution achieves minimax-fair scalable transmission of video to wireless and/or wired heterogeneous receivers. Together with a team of researchers in Intel Labs, Russia, I developed software for multicast of HDTV-resolution streams of up to 24 MB/s to multiple clients using application-level FEC.

Efficient Point-to-Point Real-Time Media Streaming over Wireless LANs

We proposed a new hybrid Automatic Repeat-ReQuest (ARQ) scheme for improving media streaming over IEEE 802.11b networks. The algorithm involved packet erasure correcting FECs and improved the effective data throughput by 25% compared to TCP-based transmission. The performance of our scheme allows transmitting DVD-quality videos in real-time over conventional IEEE802.11b networks.

Scalable Adaptive Video Streaming over Bluetooth LAN

We proposed a low-overhead standard-compatible scheme for streaming MPEG4 FGS video over Bluetooth isochronous channel. The proposed method allows implementing scalable channel-adaptive video streaming in the existing hardware.

Joint Source-Channel Coding for Image Communication (Ph.D. Thesis Work)

I developed novel statistical models for images and wireless fading channels. I proposed new multiresolution and hybrid analog-digital schemes for optimal image communications over time-varying wireless channels. One patent application filed.

Graphical Models for Image Processing and Communications (Ph.D. Thesis Work).

I have introduced a factor graph framework for statistical modeling of wavelet image coefficients and joint source-channel coding. The framework was extended to multimedia (video) retrieval based on iterative learning algorithms. I have developed a factor-graph MATLAB toolbox for low-density parity check (LDPC) codes available online (Google's Coding Theory section).