Applications of Chroma and Scale Transform for Music Retrieval
Date Issued
2012
Date
2012
Author(s)
Shih, Wen-Chun
Abstract
Music is full of our life. With the changes of the time, the amount of the music piece is growing incredibly fast. Therefore, we need some methods to organize or discover new music. Nowadays, the music recommendation system or the music website such as KKBOX, Last.fm, etc achieves purposes according the similarity between music. Therefore, how to analyze music similarity is an important issue.
This thesis was mainly divided in two topics. The first topic was the chroma. The chroma was often used to synthesize or recognize songs in the music signal processing. My proposed chroma enhanced the fundamental frequency and decreased the odd harmonics. It was good to use on recognition. Consequently, I applied the chroma to detect cover songs. From the objectivity, the cover song identification system provides a direct and objective method to evaluate the similarity between music. The second topic was the scale transform. The scale transform is a special case of the Mellin transform. It has the property of the scale-invariant. In the music analysis, take the ballroom dance music for example, the same class music has the similar rhythm. Therefore, it can be classified by using the property of the scale-invariant. Besides, the scale transform also can be applied to the speech recognition. For the same vowel, different speakers generate different spectrum, but there is a scaling relationship between them. The scale-invariant property can reduce the difference of the feature between the same vowels. The scale transform used in this thesis was different from other methods and its performance in music and speech was better than the original method.
Subjects
Chroma
Scale transform
Music retrieval
Music similarity
Type
thesis
File(s)![Thumbnail Image]()
Loading...
Name
ntu-101-R99942112-1.pdf
Size
23.32 KB
Format
Adobe PDF
Checksum
(MD5):f86a6bb25bfee568ed5b759a1e900ee6