Ge’ez Reading Level Classification By Audio Feature Extraction Using CNN

The Ge’ez language is an ancient Semitic language of Ethiopian.This language is the origin of the area surrounded by the Ethiopian and Eritrean Orthodox Church of Tewahedo.The language is given as a course in the Orthodox Tewahedo Church spiritual school.Nowadays the number of Ethiopian traditional school scholars is highly decreasing.Furthermore,due to different problems,the students are not getting suitable conditions to attend spiritual school.Attending every day in the school may be difficult for those with different work.Therefore,this research is aimed to apply machine learning techniques to build a model that can assist in Ge’ez reading level classification.On the basis of analyzing the spectral characteristics of the Ge’ez language,comparing the classification results of various models and some optimization methods,an ideal recognition system is designed.This study employs Python anaconda programming language and classification techniques such as Convolutional Neural Network to build Ge’ez reading level classification models using sample data obtained from Ethiopia Orthodox Tewahido Church.The performances of the models were evaluated using sensitivity,specificity,precision,recall,and Accuracy.The proposed model has five components: data acquisition,preprocessing,segmentation,feature extraction,and classification.In Audio signal processing,we collected a dataset from a spiritual scholar with audio files and different spiritual websites.We convert raw data into a useful and usable format during data pre-processing.Segmentation is the stage used to split the audio signal before changing into the spectrogram with equal time intervals.In feature extraction,we used to apply a Gabor filter on the input spectrogram image for texture feature extraction.Finally,the proposed model classifies the input spectrogram image using the convolutional neural network approach for grading into a specific class(Ge’ez,Wurd,and Kume).Experimental results show that among Convolutional neural network models,Alex Net,VGGNet,Google Net,and GRC_Model model with all attributes shows classification accuracy result of 99.34% and 99.12%,98.40% and 99.26% respectively.Our new outcome shows that Alex Net Model provides the best classification accuracy or performance in analyzing the Ge’ez reading level compared to other Convolutional neural network machine learning algorithms.The prevalence of the total data set which is 1701 with 3 classes was used.