We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.SD

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Sound

Title: Kinit Classification in Ethiopian Chants, Azmaris and Modern Music: A New Dataset and CNN Benchmark

Abstract: In this paper, we create EMIR, the first-ever Music Information Retrieval dataset for Ethiopian music. EMIR is freely available for research purposes and contains 600 sample recordings of Orthodox Tewahedo chants, traditional Azmari songs and contemporary Ethiopian secular music. Each sample is classified by five expert judges into one of four well-known Ethiopian Kinits, Tizita, Bati, Ambassel and Anchihoye. Each Kinit uses its own pentatonic scale and also has its own stylistic characteristics. Thus, Kinit classification needs to combine scale identification with genre recognition. After describing the dataset, we present the Ethio Kinits Model (EKM), based on VGG, for classifying the EMIR clips. In Experiment 1, we investigated whether Filterbank, Mel-spectrogram, Chroma, or Mel-frequency Cepstral coefficient (MFCC) features work best for Kinit classification using EKM. MFCC was found to be superior and was therefore adopted for Experiment 2, where the performance of EKM models using MFCC was compared using three different audio sample lengths. 3s length gave the best results. In Experiment 3, EKM and four existing models were compared on the EMIR dataset: AlexNet, ResNet50, VGG16 and LSTM. EKM was found to have the best accuracy (95.00%) as well as the fastest training time. We hope this work will encourage others to explore Ethiopian music and to experiment with other models for Kinit classification.
Comments: 11 pages, 4 tables, 3 figures
Subjects: Sound (cs.SD); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS)
Cite as: arXiv:2201.08448 [cs.SD]
  (or arXiv:2201.08448v1 [cs.SD] for this version)

Submission history

From: Richard Sutcliffe [view email]
[v1] Thu, 20 Jan 2022 20:48:07 GMT (425kb,D)

Link back to: arXiv, form interface, contact.