An architecture for effective music information retrieval

Abstract

We have explored methods for music information retrieval for polyphonic music stored in the MIDI format. These methods use a query, expressed as a series of notes that are intended to represent a melody or theme, to identify similar pieces. Our work has shown that a three-phase architecture is appropriate for this task, in which the first phase is melody extraction, the second is standardisation, and the third is query-to-melody matching. We have investigated and systematically compared algorithms for each of these phases. To ensure that our results are robust, we have applied methodologies that are derived from text information retrieval: we developed test collections and compared different ways of acquiring test queries and relevance judgements. In this paper we review this program of work, compare to other approaches to music information retrieval, and identify outstanding issues.

When "Everything" is Information: Irish Traditional Music and Information Retrieval

“…In the actual information seeking and retrieval process, the limitations of the physical paradigm may result in a number of problems for TheSession.org users when determining relevance of music information. Some of these problems derive from the differences between the written physical object and the actual musical performance of the written object (Uitdenbogerd & Zobel, 2004). Others are based upon the various levels of knowledge of the users creating the objects and the users seeking those objects.…”

“…Duggan's work in MIR and Irish traditional music will likely benefit those working with other traditional musics, as issues of ornamentation, rubato, transposition, breathing on wind instruments, and octave normalization are not unique to Irish music. Other traditional musics employ non-Western common practice scales, tuning systems, and unique methods of improvisation and will require additional MIR tools as well (Uitdenbogerd & Zobel, 2004;Tzanetakis et. al, 2007).…”

A Mid-Level Representation for Melody-Based Retrieval in Audio Collections

“…Melody is an important descriptor of music [5] and is therefore a very natural descriptor to be used for measuring music similarity. In the symbolic domain, melodic similarity has been extensively studied [6][7][8], but the work cannot be applied to audio signals directly, while methods for transcribing polyphonic audio signals to symbolic representations are not yet accurate enough [9,10]. Shwartz et al [11] presented a probabilistic model for query by melody that uses symbolic melodies to query audio collections.…”

Music structure based vector space retrieval

“…Many MIR systems have been reported in the survey articles [18] [24]. MIR research community initially focused on developing text based systems where both database and the query are in the MIDI format and the information is retrieved by matching the melody of query with the database [5] [6][10] [12][17] [19] [25]. Since the melody information of both query and song database are text based (MIDI), the research has been devoted to database organization of the music information (monophonic or/and polyphonic nature) and to textbased retrieval models.…”

“…Since the melody information of both query and song database are text based (MIDI), the research has been devoted to database organization of the music information (monophonic or/and polyphonic nature) and to textbased retrieval models. The retrieval models in those systems includes dynamic programming (DP) [15][22] [25], n-gram-based matching [5][6] [25] and vector space model [17].…”