Speech recognition based Radon-Discrete Cosine Transforms by Delta Neural Network learning rule

Abstract

The recognition of seven recorded words in this paper is proposed by using a Radon and Discrete Cosine Transforms. The Radon Transform is used to reorder the data input with a new shape so that each voice maintain the same number of samples approximately, while the second frequency transform is used to minimize the dimensions of each audio signal to a small number of samples. The goal of this method is to rise the number of recognized audio signals and consequently increasing the database. The learning rule of Delta Neural Network is used for recognition with the assistance of multi neurons of a single layer such that the number of audio signals (recorded words) are coincide the number of neurons. The results will be compared based on the learning speed. The proposed system also examined by a test audio signal. Here, seven different words are recorded. Eight different persons (men and women) recorded these different words, so that there are 56 audio signal. Each eight signals belongs to the selfsame word; consequently, the outputs of these eight audio signals are the same.