ETHIOPIAN ORTHODOX TEWAHIDO CHURCH DRUM SOUNDS CLASSIFICATION USING MACHINE LEARNING APPROACH

KASSAHUN, ASIBO

dc.contributor.author	KASSAHUN, ASIBO
dc.date.accessioned	2021-08-13T06:45:53Z
dc.date.available	2021-08-13T06:45:53Z
dc.date.issued	2021-03
dc.identifier.uri	http://ir.bdu.edu.et/handle/123456789/12391
dc.description.abstract	Musical instrument classification is the process of building a system that listens to the specific sounds of the instrument and classifies which instrument is being played. Various researches have been done to automate classifications of musical instruments based on their sound for music information retrieval (MIR) system, music data analysis and music transcription purpose. When the number of music data increases, indexing and retrieving becomes more challenging. Previous studies in this area had been concentrated on the distinguishing of western musical instruments for modern music services. Automatic classification of spiritual drum sounds has not been given due attention so far. In order to contribute to this limitation, a model was developed to classify the sounds of drum in the Ethiopian Orthodox Tewahido Church (EOTC). Drum is one of the most popular spiritual song instruments in the Ethiopian Orthodox Tewahido Church (EOTC) used for devotional services and accompanying St.Yared chants. Drum has produced different sounds based on hitting styles and spiritual chants such as Meregd, Nusi-Meregd, Tsifat, and Chebchebo. The sounds of the drum were recorded from EOTC at North Wollo, Dawunt, Chet St.Gebriel and kurba medhanialem using a smartphone. Convolutional Neural Network (CNN) and Gray Level Co-occurrence Matrix (GLCM) were used to extract the deep and handcrafted texture features from the Mel-spectrogram image respectively. These features were fed to softmax and Support Vector Machine (SVM) independently and with combination to classify the drum sounds. Then, the grid search algorithm was used in the SVM parameter value selection for selecting the best estimators because there is no theoretical technique to determine or select the efficient hyperparameter values. Accordingly, the proposed model achieved 93% accuracy on CNN with softmax and 95%, 74%, and 97% accuracy on CNN, GLCM, and combined features with SVM respectively.	en_US
dc.language.iso	en_US	en_US
dc.subject	computer science	en_US
dc.title	ETHIOPIAN ORTHODOX TEWAHIDO CHURCH DRUM SOUNDS CLASSIFICATION USING MACHINE LEARNING APPROACH	en_US
dc.type	Thesis	en_US