Implementation of audio signal processing for automatic indonesian musical genre classification
Musical genres can be defined as categorical description of music based on its behavior. Indonesia as a multicultural nation which has a wide diversity in the area of music does not have a categorical description for its local and traditional music. This study focuses on finding the pattern of Indonesian music with experimenting several genre-sets with the proposed feature-sets: musical surface features, MFCC, and rhythm features. Various experiments relating to feature extraction are conducted which resulted in utilizing all feature-sets producing the best classification performance. Separate experiments of local and traditional Indonesian genres are conducted, showing the classification performance of each genre-set, with traditional genre-set delivers the highest classification accuracy. Classification tasks are implemented in real-time and non real-time, using decision trees and Naive Bayes classifier. The results are evaluated and compared to validate the strength of each feature-set and to synchronize the human musical perception with the machine musical perception.
B00524 | (wh) | Available |
No other version available