Speech Database in Sylheti and Speech Recognition using Convolutional Neural Network

Speech dataset is the primary and core element for a speech/speaker recognition system specific to a language. Sylheti, a language of Indo-Aryan family, is a member of under-resourced language group which is spoken in Sylhet district of Bangladesh and Barak valley part of Assam, India. It does not have any electronic and web data resources. So, there is a requirement to design a speech database in Sylheti. As the speech database construction is an ongoing process, the preliminary task of developing a small vocabulary new speech database in Sylheti is presented in this paper. Thereafter, implementation of two Automatic Speech Recognition systems for the Sylheti language by using convolutional neural network are reported here.