| Title | : | Segmentation of Musical Items: A Computational Perspective |
| Speaker | : | Sridharan Sankaran (IITM) |
| Details | : | Thu, 27 Oct, 2016 4:00 PM @ BSB 361 |
| Abstract: | : | Segmentation of a musical item from the audio by locating repetition of melodic phrase is a well researched problem in Western music. Various techniques have been used for segmentation such as those using machine learning and those that use other methods. While these techniques have been attempted for Western music where the repetitions have more or less static time-frequency melodic content, finding repetitions in improvisational music such as Indian classical music is a difficult task.
A Carnatic music item has multiple segments. In this work, we identify the boundary between alapana and composition using difference in timbre due to absence of percussion in alapana segment. This done by evaluating Information theoretic KL2 distance between adjacent samples and thereby locating the boundary of change. The boundary is verified using GMM. Next , we segment a composition into its constituent parts using Pallavi (or a part of it) of the composition as the query template. The repeated line of a Pallavi is seen as a trajectory in the time-frequency plane. A sliding window approach is used to determine the locations of the query in the composition. The locations at which the correlation is maximum corresponds to matches with the query. In the above, we use Cent filterbank based features which are tonic independent and hence comparable across musicians and concerts where variation in tonic is possible. |
