Turkish Regional Dialect Recognition Using Acoustic Features of Voiced Segments
Baran Uslu1 and Hakan Tora2
1. Atilim University, Electrical and Electronics Engineering, Ankara, Turkey
2. Atilim University, Electrical and Electronics Engineering / Avionics, Ankara, Turkey
2. Atilim University, Electrical and Electronics Engineering / Avionics, Ankara, Turkey
Abstract—In this study, we aimed to classify Turkish dialects by machine learning methods. The Black Sea, Aegean and Eastern dialects were examined in the scope of this paper. First of all, the speeches belonging to the relevant regions were obtained by recording. Then the speech records were examined in detail by Praat and Matlab software. The attributes that characterize these three regional dialects were selected as pitch, jitter and shimmer. These features were then used for training an artificial neural network classifier. Classification rates of 98.9% for training set and 95.8% for test set were achieved. Also, this study bridges between the two disciplines of Linguistics and Speech Processing for Turkish dialect recognition.
Index Terms—turkish dialect, regional accent identification, jitter, shimmer, neural network
Cite: Baran Uslu and Hakan Tora, "Turkish Regional Dialect Recognition Using Acoustic Features of Voiced Segments," International Journal of Signal Processing Systems, Vol. 6, No. 2, pp. 17-21, June 2018. doi: 10.18178/ijsps.6.2.17-21
Index Terms—turkish dialect, regional accent identification, jitter, shimmer, neural network
Cite: Baran Uslu and Hakan Tora, "Turkish Regional Dialect Recognition Using Acoustic Features of Voiced Segments," International Journal of Signal Processing Systems, Vol. 6, No. 2, pp. 17-21, June 2018. doi: 10.18178/ijsps.6.2.17-21