Spoken alphabets (A-Z) and digits (1-9) recognition using LSTM

Question

I want to train LSTM that can recognize spoken alphabet (A-Z) and digits (1-9). Any help will be appreciated!