Speech recognition by Turkish word
Onur Peker · 2016
In this study, voice verification process for the Turkish words is introduced by using Linde-Buzo-Gray algorithm and artificial neural networks. Speech recognition process which is carried out, is independent from the speaker but dependent on the text. For the voice signals belonging to each speaker, attribute vectors cluster is formed by subtracting mel cepstral factors. These vectors have been modeled with Linde-Buzo-Gray algorithms and artificial neural networks, the codebook for Linde-Buzo-Gray and artificial neural networks for neural networks have been constituted and new sounds have been clustered by finding the closest code vector in the existing book or the neural network. In this way, voice verification is performed.