This research focuses on the investigation of a unified methodology for the onset detection in Sri Lankan radio broadcast context with the approach of classification of the broadcast context. Various audio patterns in the broadcast context were observed and a supervised learning approach was employed in the classification process. Different audio features were examined with respect to the broadcast context. Identified audio semantics in the broadcast context were used in refining the output gained in supervised learning models. Onsets were predicted using the classification results. The evaluation method was carried out with ground truth data obtained from a Sri Lankan FM broadcast recording. The proposed approach provided the accuracies of 41% for news, 76% for radio commercials, 75% for songs and 59% for other voice related segment classification. The onset detection model was successful in predicting the onsets with an error rate of (+/-) 2.5s with approximately 82% of accuracy level.
This paper describes our attempt of assessing the capability of music melodies in isolation in order to classify music files into different emotional categories in the context of Sri Lankan music. In our approach, Melodies (predominant pitch sequences) are extracted from songs and the feature vectors are created from them which are ultimately subjected to supervised learning approaches with different classifier algorithms and also with classifier accuracy enhancing algorithms. The models we trained didn’t perform well enough to classify songs into different emotions, but they always showed that the melody is an important factor for the classification. Further experiments with melody features along with some non-melody features showed us that those feature combinations perform much better, hence brought us to the conclusion that, even though, the melody plays a major role in differentiating the emotions into different categories, it needs the support of other features too for a proper classification.