Current Work:

Mobile Quality of Experience: My current work is improving quality of experience for mobile applications (Video and Web) by understanding various bottlenecks such as OS/Hardware, Poor Network and Application Complexity itself.

Previous work:

Multimedia Transport in TDMA based Wireless Networks

Picture4-page-001we proposed a generic Time Division Multiple Access (TDMA) based Medium Access Control (MAC) for transmitting multimedia traffic in Ad hoc Wireless Networks (AWNs). We studied the complete system with integrated services such as data, voice, audio and video flows by providing the Quality of Service (QoS) for real time traffic. The core idea behind the protocol is a novel slot distribution for different kind of flows (particularly video flows due to their huge bandwidth requirement) based on the dynamic requirement of bandwidth. We first present the slot distribution for different flows and then explain the algorithm to further optimize the slots usage with adaptive reservation based on the video compression analysis. Since the End to End (E-to-E) delay in a multi hop network is a critical factor in guaranteeing QoS for multimedia traffic, we considered the slots distribution most favorable to real time media transport over AWNs. The simulations were conducted for investigating the performance of the protocol such as Eto-E delay, packet loss ratio (PLR) and throughput by varying different types of services. The adaptive reservation of slots based on the analysis of the bit rate generated by video codec, highlights the work in evaluating the performance of the system.

Slot Conflicts due to Mobility in TDMA based Wireless Networks

bar10-page-001With collision free transmission at its core, Time Division Multiple Access (TDMA) is becoming increasingly popular for Mobile Ad-hoc Networks (MANETs). Each node is assigned a conflict free transmission schedule by slot assignment algorithm to avoid collisions in the transmission. As the slot assignment depends hugely on spatial reusing of slots for better bandwidth utilization, slot conflicts occur when nodes with same transmission slots move into each other’s vicinity. We propose a procedure to detect and resolve slot conflicts in MANETs.

TDMA Frame Design for Voice Traffic in MANETs

Time division multiple access (TDMA) based channel access is one of the most widely used mechanisms in Mobile Ad hoc Networks (MANET) with its efficient channel access and collision free transmissions at its core. In this paper, we propose a unique TDMA frame format designed specifically for supporting two-way voice communication in bandwidth efficient manner. The proposed algorithm uses single radio channel for both control and data message exchanges and achieves multi-hop communication. As the transmission of control messages in the same channel leads to overhead, we propose a mechanism to share the control slots between the nodes for reducing the control overhead. The TDMA frame design shows the optimized channel sharing based on the periodicity and payload size of a typical voice packet. The experimental analysis is performed on Network Simulator (NS-3) as well as on Linux kernel with Ethernet emulation of wireless network. The experimental results clearly show considerable improvement in bandwidth efficiency over the existing protocols.

Delay Sensitive TDMA Slot Assignment

tdmaframeTime slot assignments in a TDMA ad hoc wireless network (AWN) is either centrally coordinated by a root node or distributed among all the nodes in the network. In the centralized TDMA network, the root node uses the global knowledge of the network to assign slots, but becomes more challenging in case of distributed network, as each node is expected to assign a slot for itself without conflicting other nodes’ slot selection. There is plenty of literature on how slots are assigned in a centralized TDMA network but only a few on distributed. Quality of Service (QoS) is critically important in AWNs and a good slot assignment scheme prioritizes its QoS metrics during the process of slot assignments. Real-time communications require end-to-end delay and jitter within acceptable limits for better overall QoS. This paper proposes a delay sensitive approach to TDMA Slot assignment problem in distributed AWNs. The proposed approach does a balancing act between end-to-end delay and spatial reuse. The experimental results demonstrate that the proposed approach obtains quality results in terms of call acceptance rate, end-to-end delay and spatial re-usability.

Motion Estimation H.264 Video Compression

The block matching algorithm (BMA) used for motion estimation (ME) during video coding in H.264 standard takes nearly 90% of total encoding time. The proposed work is based on analyzing the video content to dynamically choose an efficient search pattern. The algorithm based on the variance of motion in the video shows efficient results by reducing the number of candidate macroblocks that need to be searched in the reference frames using different types of search patterns and also recommends an optimized motion vector search range. The algorithm shows 90% improvement in computational time over the full search algorithm and a significant improvement over other fast block matching algorithms without compromising the bitrate and quality of the video.

GPU implementation of Image Restoration Algorithm

Picture1The image processing applications involve huge amount of computational complexity as the operations are carried out on each pixel of the image. The General Purpose computations that are data independent can run on Graphics Processing Units (GPU) to enable speedup in running time due to high level of parallelism. Compute Unified Device Architecture (CUDA) and Open Computing Language (OpenCL) programming environments are well known parallel programming languages for GPU-based Single Instruction Multiple Data (SIMD) architectures. This paper presents parallel implementation of Belief Propagation (BP) algorithm for Image Restoration on GPU using OpenCL parallel programming environment. The experimental results shows that, GPU-based implementation improves the running time of BP for image restoration when compared to sequential implmentation of BP. The best and average running time of BP algorithm on GPUs with 14 multiprocessors (48 cores) is 0.81ms and 1.46ms when tested on various benchmark images with CIF and VGA resolution.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s