I have designed one acoustic 8 psk modem working on acoustic signal.It functionally works.I have written the code on Matlab and tested with Speaker and Mic.But there are several improvements needed like reducing the training symbols needed for convergence of DFE.I have also integrated two DPLL s needed for Carrier Frequency Offset(CFO) and Sampling Clock Offset(SCO).It has been tested in hardware.A lot of performence improvements have to be done.This works on standard 3 KHz band width.The same modem can be used for HF channels.I want to do reseacrch by first gaining practical experience through making my algorithms to work on hardware.After gaining considrable pracical experience I want to publish research papers in reputed journals like IEEE.I want to work on both under water as well as HF channels.The entire code development will be done in Matlab platform and using AI tools.I am also improving my theoretical knowledge by revising my mathematical skills needed for publication.If any one who is well versed in Matlab and Signal Processing/Communication tool boxes is interested to  do research jointly  may contact me.I am living in India.The research co-operation does not involve any financial help to either side.It will have NO commercial interest.It is only for research and publication.Students may use the work for there thesis work.

Due to sever ISI the acoustic modem takes thousands of symbols for training.To reduce training symbols I estimated the channel with`known preamble with less symbols.I trained the DFE with locally generated DUMMY symbols fed to the Channel Estimated.The DFE converged.But when I switched to data mode with actual data from hardware to equalize DFE does not work.The Symbol Error Probability looks it is random.Any suggestions to fix the problem.I got this idea of using dummy symbol from Widrow's Adaptive Control book.He had used for plant.That technique I adapted for 8 psk modem.Any suggestion welcome

V.Muralidharan

Mannai_Murali@hotmail.com