My goal is to make an approach to recognize modulation type of unknown signals. I have energy detector which monitors spectrum and it provides information about center frequency and bandwidth of detected signal. After shifting desired signal to 0 frequency I make a resampling that new frequency sampling equals detected signal bandwidth. Problem is that I don't know exact sampling rate at which desired signal was generated at the transmitter. While analyzing the constellation plot after resampling there is no chance that convolution network will recognize modulation type. Do you think that problem is unsolvable and this is waste of time? Are there any techniques to overcome sampling frequency offset?
Any guidance will be appreciated.
Thanks in advance,
Marcin

My goal is to make an approach to recognize modulation type of unknown signals. 

I have energy detector which monitors spectrum and it provides information about center frequency and bandwidth of detected signal. After shifting desired signal to 0 frequency I make a resampling that new frequency sampling equals detected signal bandwidth. Problem is that I don't know exact sampling rate at which desired signal was generated at the transmitter. While analyzing the constellation plot after resampling there is no chance that convolution network will recognize modulation type. 

Do you think that problem is unsolvable and this is waste of time? Are there any techniques to overcome sampling frequency offset?