Step 2: Literature survey (Understand current status)

While deciding the most suitable research topic (in blog 1), we already have gone through hundreds of paper titles and many abstracts. Now, literature survey is to be taken more seriously. We usually feel that, anyways I am going to do new research / innovative research and WHY should I read what others have done. In this blog, I will explain why it is important and why it should take at least 10 to 15% of your overall research time frame.

Considering ongoing courses of 9am to 4pm everyday and considering the possibility of thousands of papers in the chosen research area, the question is how to manage in 15% time. For example, for extracting the dicrotic notch of pulse signal coming out of Nadi Tarangini, or for extracting the PR interval from electrocardiogram of MAC600 GE ECG machine, there will be numerous approaches. Most importantly the approaches of ECG are also suitable for pulse and vice versa.

 Evaluate your sources: [approximately 4 hours] First important task is to tabulate the sources (library / online contents / journals / conference proceedings etc.) from which you might gather your knowledge about the research topic. Discuss with your guide and seniors about the authenticity and creditability. Nowadays, reviews are usually available about almost everything, so make use of it.

We should read at least 50 complete papers / articles of the research area. One way of choosing the papers is dynamic as:

An important point to make here is – a PhD student typically spends 1 year in the literature survey and reproducing the data-methods-results of earlier authors to really understand the problem domain. You may now realize how important this step of comprehensive and critical analysis of literature survey is before finalizing the objectives.

When you read the above papers, note that every author is biased about something. E.g. when I personally write a paper, I am biased towards wavelet transform than Fourier transform. Both the approaches have their academic and practical pros and cons. But, as I am more comfortable in writing the codes for wavelet transform, in my paper, I will show that how wavelet transform is better than Fourier transform. Some author might write the other way. So do not get confused. Typically, choose the latest developed method – it is usually better.

In most such papers, you will observe that:

• A certain paper sets some objectives and achieves them up to certain extent
• The second paper (may be by the same author) brings out the limitations of the first paper and sets new objectives
• The third paper brings out the limitations of second paper as well as sometimes of first paper
• And so on

Make a compilation of them and present to your guide and seniors with these 5 slides of different objectives, different input data, different methods, different results and different limitations. Also make one big slide which connects all dots in your review as:

You may observe that your institute typically asks for a paper, 2 months after starting your research journey. These 5 slides will generate your review paper. But write the paper after next step of objectives. Presenting the paper and the review process by an expert helps in understanding whether the literature survey is completed correctly or not.

Finally, if possible financially and time wise, visit an actual place of problem domain say authentic Panchakarma center in Kerala, commercial Panchakarma center in Goa, Signal processing lab in Tata motors, Big data cloud server in Hyderabad, etc to get the complete feel. This quantitative knowledge will surely motivate you in your further steps of research.

In next blog, we will set the objectives of the problem domain. Find below an image copied from “pinterest”, where an author has very beautifully covered what is literature survey.