Content-Based Streaming and Real-Time Regional Language Captioning of E-Learning Video Data

Introduction

In India, a majority of population dwells in non-urban setup where the educational infrastructure and resources are usually meager and scanty. Trained teaching faculty at all education levels – technical, secondary and primary is lacking, or teachers are unmotivated. The students that come from such background are less likely to excel than those who are exposed to the best education in the world. E-learning techniques like video lectures can be very beneficial for students in rural educational institutions as they can provide the students the exposure to the best education in the world. Many institutes have opened their web servers for free lecture-on-demand for several courses. However, there are the following hurdles to accomplish the above mentioned goal:

  1. Students are usually economically backward and cannot afford personal computers and the present infrastructure lacks an easy access to computing facilities.

  2. Even if the end user facility is provided, the Bandwidth in rural areas (or even towns like Roorkee or Meerut) is extremely scarce and it is a usual experience in small Indian cities and towns that a file of even a few MBs takes substantial time to download. Typically, the video lectures available publicly run into a few MBs and require an efficient adaptive streaming strategy.

  3. The students usually communicate in their local languages and can understand bits or pieces of English and cannot completely catch the flow of the video lectures that are being downloaded.

Solutions

This project proposes to provide some solutions for Problems 2 and 3. Traditionally, problem 2 is solved using adaptive streaming where the resolution or the quality is compromised to meet with the bandwidth constraint. However, we propose a new streaming technique that can adapt to changes in real available bandwidth.

The second part of the project involves captioning in regional language. Typically, the lectures in major universities are delivered in English. The optimal use of these lectures is possible if the audience, who may only know English partially, get some additional input in their own language. A regional language captioning system is being developed for complementing the end user’s comprehension of the lectures delivered in English.

The product is primarily for making E-Learning widespread and it is likely to be employed by several government and private educational institutes.

Chief Investigator & Primary contact person :

Dr. Ankush mittal 
Associate Professor,
Dept. of Electronics & Computer Engg.,
IIT Roorkee.