First the system is trained in a noisy environment with strong pronunciation of each word and letter then during test and
recognition, user must be close enough to the microphone and pronouns the words strongly. Fig 11 below is shown that the
system recognized “GO” word, as you can see recorded signal is similar to the library signal, there are some changes of
noise, pitch and utterance but Matlab by using FFT method is successful in finding the correct word.