I have been searching for robust algorithm that can be implemented for avoiding the ISI (Inter-Symbol interference), I came across two types of channel equalizations are there. One is data Aided (training based), other is non-data aided (blind). LMS for Data aided requires large training sequence to guarantee the convergence which makes it not suitable. CMA is an example for non-data aided, but it good for PSKs, and 4-QAM because all symbols have the same amplitude, so it will not be suitable for higher order QAMs which are not encoded deferentially.
Finally I came to know, there are equalizers known as Fractionally Spaced Equalizers in which the sapling rate is greater than symbol rate (at least twice). Each symbol is sampled M times, then M sub-channel equalizers of length greater than or equal to the channel impulse response length have to be computed. On combining the all equalizers output, we will get the equalized symbols.
There are many papers on the fractionally spaced equalizers, but no one is clearly explaining how to estimate (or compute) and update the M sub-channel equalizers and how to combine those outputs to get the equalized symbols.
So please could you recommend any links or papers or books or any information that you know how to estimate and update the Fractionally Spaced Equalizer and how to get the equalized symbols from that equalizer.