Research On Speech Recognition Model Of Arabic Numbers

Speech recognition technology is part of the broad field of AI technology.In the past decade,due to the development of computer technology,the technology in this field has also made amazing breakthroughs.More and more industrial products have come out,which has brought great changes to people's lives.This paper describes the application processes of speech recognition technology in Arabic,and mainly studies the pronunciation characteristics of numbers in Arabic.At the same time,based on the Meier frequency cepstral coefficient(MFCC),the corresponding model is established to predict the new speech data and determine which number the connotation of its expression belongs to.Because the traditional matching model and the artificial neural network model(RNN)have their own shortcomings,the author proposes the matching-RNN model.Experiments show that the new model performs better in training set and test set than the traditional model.