Heet Sheth

Project Title: A Method for Compression using variable entropy method.

 

Project Duration: January 2015 - April 2015 (4 months)

 

Project Type: Personal Project

 

Patent Application Number: 1893/mum/2015

 

Find Implementation of Algorithm Here (GitHub Code link): [code]

 

Mentor:

Dr. J. N. Sarvaiya, [personal profile]

Associate Professor,

Electronics and Communication Engineering Department,

SV-NIT, Surat, INDIA.

 

Colleagues:

1. Prashant Pandey

2. Dhaval Vora

3. Jignesh Sarvaiya

 

Abstract:

In this invention, we proposed a method of compression by reducing the entropy of data.A multi-level compression method is proposed by using commands.Subsequently a new method to achieve compression using the entropy by the division of data into classes & using commands is being proposed for both long & short data; the analysis for which is in the description. This method is exclusively helpful for short messages.

 

 

Earlier data/information was limited to mailing, audio but now in the era of 21st century, technological applications viz. CCTV, cloud services need a large amount of data to be stored. To store, maintain and frequently use the data we need high bandwidth. But ideally, we need to compress the data for faster transfer and less usage of bandwidth.

 

The heart of data compression lies in the communication system, and this is done at source coding level.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Huffman and Shanon-Fano are all best suited when the probability of different symbols/characters have a huge difference in them. (i.e. the symbols are not equiprobable) —Shanon-fano is better than Huffman in assigning bits sequence to symbols. —Encoder, as well as a decoder, should know one to one mapping between symbols and bit sequence, which is again an overhead.

 

—If the average number of digits used in encoding is L, then the entropy of digits is H(X)/L.

The efficiency is then defined as —Efficiency =H(X)/Llog(M)

 

—But what if data is equiprobable? —The entropy will be very high which will degrade the efficiency.

 

—All the earlier techniques used concentrate on assigning bit pattern to the symbols. —But in this approach, we decrease the Entropy of the information by using commands. —Once we decrease the Entropy then we can use any of the above bit assignment technique to do the compression.

 

So, we can say that this approach is helpful in creating a base to use existing compression techniques when data is not equiprobable.

 

—Normally the techniques which we use today, compresses data at one shot, once data is compressed the compressed data cannot be compressed again for granted.

—But by using commands, we still have equal possibility to compress again and again, it can only be limited by the types of command available.