Searching notated polyphonic music using transportation distances

Abstract

We present a method for searching databases of symbolically represented polyphonic music that exploits advantages of transportation distances such as continuity and partial matching in the pitch dimension. By segmenting queries and database documents, we also gain partial matching in the time dimension. Thus, we can find short queries in long database documents, and have a method more robust against pitch and tempo fluctuations in the queries or database documents than we would with transportation distances alone. We compare our method with three algorithms from the C-Brahms project by Lemström et al. and with PROMS by Clausen et al. and find that our method is more generally usable, retrieves a higher number of relevant documents than all three compared algorithms, and that it is faster than C-Brahms. This is the first comparative study of these algorithms involving a large database with about half a million of documents.

Compacting music signatures for efficient music retrieval

“…Content-based music retrieval has attracted much research interest recently [8,12,13,14,15,18,21,22,23,3,2]. Several content-based retrieval techniques have been proposed in the literature, and they can be classified into two categories, i.e.…”

A Measure for Evaluating Retrieval Techniques based on Partially Ordered Ground Truth Lists

“…A more complex example can be found in Figure 1, which shows the ADR for a sample result of our Earth Mover's Distance-based algorithm for measuring melodic similarity [8]. Fig.…”

“…1. Number of retrieved documents (left scale) and dynamic recall (right scale) for a distorted version of incipit 240.001.397-1 from [9] using a variant of our algorithm from [8]. The average dynamic recall here is 0.74.…”

A Query-by-Singing System for Retrieving Karaoke Music

“…The root concept, query-by-humming/singing or query-by-example, continues to motivate the development of many promising solutions for retrieving music beyond the conventional text-processing paradigm. Examples of such solutions enable users to retrieve a song by humming a catchy tune without having to name the song [1]- [17], [30]- [36], or help users find songs performed by their favorite singers [18], in their favorite genre [19] or mood [20] by playing an excerpt of the desired music as a query.…”

“…Because of this difficulty, the vast majority of current query-by-humming/singing systems [1]- [7], [15], [31], [35] work within the monophonic domain, which converts a monophonic audio query into a symbolic format to match a monophonic symbolic collection. Some studies [16], [17] focus on locating the major themes from a piece of symbolic polyphonic music in which the note information is given a priori. However, very few systems operate in the mode of monophonic audio queries applied to a polyphonic audio collection [8], [9], [13], [30], [36], or entirely polyphonic audio queries applied to a polyphonic audio collection [10]- [12].…”