I did this a few years back for modulation recognition of OFDM (and non-OFDM) waveforms (using MATLAB as it happens). Essentially you use correlation techniques to obtain the periodicity of an OFDM signal. This periodicity appears due to the guard intervals at the start of each symbol and the (usually) regular pattern of pilot tones within each symbol (e.g. scattered pilots in DVB-T or positioning reference signals in LTE).

Without going in to too much detail (yet!) I evaluated cyclostationary processing first to determine the symbol length using an adapted form of eqn. 23 in the publication listed below. I used an FFT to iterate that equation quickly. Then I calculated the normalised cyclo-stationary function (NCF) below (apologies for lack of formatted maths) over a number of symbols to determine the guard length. Again I used an FFT to determine a periodicity in the NCF. The periodicity is related to the guard length.

NCF(n) = sum(from k=n to k-n+NCP-1) {x(k)*conj(x(k+N)} / (1/NCP* sum(from k=n+1 to k=n+NCP) abs(x(k))^2)

where x is your time-domain signal, N is the assumed FFT length (that was used to generate the symbol).

You can confirm you have the correct symbol length by doing a spectral analysis of each symbol, and average to view the pilot tones or you can do a spectral analysis of the FFT of each symbol (cepstral analysis) to spot the periodicity in any regular pilot tones that may be present.

https://www.google.co.uk/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved=0ahUKEwj11KbgrJ_MAhWBHsAKHQFXAd4QFgggMAA&url=http%3A%2F%2Fperso.telecom-paristech.fr%2F~ciblat%2Fdocs_recherche%2Farticles%2Fsigpro2010.pdf&usg=AFQjCNGkJcshwydpJf4P-PATrZlvRAZFVg&sig2=eSRxMa7r1PVNhu-w-ERXiA

Thank you so much for this valuable response, but Sir my worries are always towards the mathematical laws which I do not control

Really appreciate your answer Miss Aparna