Özet:

The security problems arising with the developing technology have put secure systems in quest of being developed with new techniques. The voice processing technology among these techniques has gained great importance. This situation has directed the reliability ensuring the man-machine interaction together with the voice processing applications such as speech recognition, speaker recognition, converting audio to text or converting text to audio to the design of high and correct systems. In this study, a digit speech recognition system consisting of the audio examples of digits from 0 to 9 in Kurdish spoken in Turkey was realized. For this purpose, audio recordings were obtained from 102 adult speakers who speak Kurdish. The feature vectors of the audio data acquired from these recordings were obtained with the Mel-Frequency Cepstral Coefficient by using different windowing methods, and the success rates were compared by conducting the recognition process with Dynamic Time Warping according to the feature vectors obtained.

Anahtar Kelimeler: