How to recognize words / phrases (recorded earlier)?

Question

Delphi

Good day.
There was such an idea. The user writes down some word or phrase.
It is necessary to determine when he will say it / her later. I read about the Fourier transform, figured it out in principle, drew a pronunciation into a microphone on a real-time canvas.
Where should I dig next to get the functionality I need?
Thank you in advance.