Abstract :
Voice is often used to send important information in the field of defense and
others. A system is needed to keep this information accurate. Speech recognition is
a process performed by a device to recognize and understand spoken words.
Meanwhile, Euclidean Distance is the calculation of the distance from 2 points in
Euclidean space. This study designed a speech recognition system that can verify
speech based on what has been said with FFT (Fast Fourier Transform) feature
extraction and Euclidean distance pattern matching.
The speech recognition system is designed starting from the process of
reference data and test data, both processes have the same process starting from
voice recording, voice cutting, preprocessing, feature extraction. This system uses
the FFT (Fast Fourier Transform) method for feature extraction. The results of
feature extraction from the two data will be compared with the euclidean distance
as a pattern match. Calculations using euclidean distance that produce the smallest
distance or value will be determined as the output.
The system test results show that the system can verify numeric speech with
an average recognition accuracy of 53%. The results were obtained from 10 tests of
the reference data of each speaker, which was carried out by 3 speakers. Each reciter
recites the sound of the assigned number one, two, three, four, five, six, seven, eight,
nine, ten.