Correlation and clusterisation of traditional Malay musical instrument sound using the I-KAZTM statistical signal analysis
The best feature scheme is vital in musical instrument sound clustering and classification, as it is an input and feed towards the pattern recognition technique. This paper studies the relationship of every traditional Malay musical instrument acoustic sounds by implementing a correlation and cluste...
Saved in:
| Published in: | Journal of Mechanical Engineering and Sciences Vol. 14; no. 1; pp. 2552 - 2566 |
|---|---|
| Main Authors: | , , , , |
| Format: | Journal Article |
| Language: | English |
| Published: |
Pekan
Universiti Malaysia Pahang
01-03-2017
Universiti Malaysia Pahang Publishing |
| Subjects: | |
| Online Access: | Get full text |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The best feature scheme is vital in musical instrument sound clustering and classification, as it is an input and feed towards the pattern recognition technique. This paper studies the relationship of every traditional Malay musical instrument acoustic sounds by implementing a correlation and clustering method through the selected features. Two types of musical instruments are proposed, namely flutes involving key C and key G classes and caklempong consisting of gereteh and saua. Each of them is represented with a set of music notes. The acoustic music recording process is conducted using a developed design experiment that consists of a microphone, power module and data acquisition system. An alternative statistical analysis method, namely the Integrated Kurtosis-based Algorithm for Z-notch Filter (I-kazTM), denoted by the I-kaz coefficient, Z∞, has been applied and the standard deviation is calculated from the recorded music notes signal to investigate and extract the signal’s features. Correlation and clustering is done by interpreting the data through Z∞ and the standard deviation in the regression analysis and data mining. The results revealed that a difference wave pattern is formed for a difference instrument on the time-frequency domain but remains unclear, thus correlation and clusterisation are needed to classify them. The correlation of determination, R2 ranging from 0.9291 to 0.9831, thus shows a high dependency and strong statistical relationship between them. The classification of flute and caklempong through mapping and clustering is successfully built with each of them separated with their own region area without overlapping, with statistical coefficients ranging from (2.79 x 10-10, 0.002932) to (1.64 x 10-8, 0.013957) for caklempong, while the flute measured from (2.45 x 10-9, 0.013143) to (1.92 x 10-6, 0.322713) in the x and y axis. |
|---|---|
| ISSN: | 2289-4659 2231-8380 |
| DOI: | 10.15282/jmes.11.1.2017.13.0234 |