Dr. Anirban Lekharu

Two client-side Rate Adaptation models are proposed for the DASH (Dynamic Adaptive Streaming over HTTP) framework in maximizing the Quality of Experience (QoE) of the clients.

In the first approach Client-side Scalable Video Coding (SVC) based bitrate adaptation framework called Video Quality Adaptation Framework (VQAF) with two distinct phases called the Download and Smooth-out phases is proposed.

The Download Phase fetches new segments at selected quality levels until buffer size reaches a stipulated safety Threshold.

The Smooth-out Phase repeatedly upgrades a subset of the already downloaded segments by one enhancement level, until buffer size reduces below the safety Threshold.

In the next approach a Machine Learning based Rate Prediction model called the Video Bit-Rate Prediction Model (VBPM) using Long Short Term Memory (LSTM) for Adaptive Streaming over HTTP is proposed.

The Proposed Framework evaluates the robustness of the video watermarking schemes against content adaptation attacks as well as general attacks.

Designed and Integrated the various modules of our Framework.

Tested the Watermark Framework on different types of videos.

Implemented a model on Image Captioning using Attention Networks.

The model was able to generate text descriptions of images which not only captures the objects contained in an image, but it also must express how these objects relate to each other.

Presented a model for Human Activity Recognition using Sensor data

The model comprised of a PIR, PING and LIGHT sensors which gave us different sensor values of the people activities in performed in the Robotics Lab, Dept. of CSE, IIT Guwahati.

Implemented some widely used algorithms to recognize the human activities in the Lab.

Surveyed and presented a comparative study of various research papers, which found flaws in security mechanism used by ZigBee that works on IoT devices for smart home applications.