=Paper=
{{Paper
|id=Vol-1177/CLEF2011wn-MusiCLEF-Di_BuccioEt2011
|storemode=property
|title=University of Padua at MusiCLEF 2011: Music Identification Task
|pdfUrl=https://ceur-ws.org/Vol-1177/CLEF2011wn-MusiCLEF-Di_BuccioEt2011.pdf
|volume=Vol-1177
|dblpUrl=https://dblp.org/rec/conf/clef/BuccioMO11
}}
==University of Padua at MusiCLEF 2011: Music Identification Task==
https://ceur-ws.org/Vol-1177/CLEF2011wn-MusiCLEF-Di_BuccioEt2011.pdf
University of Padua at MusiCLEF 2011: Music
Identification Task
Emanuele Di Buccio, Nicola Montecchio and Nicola Orio
Department of Information Engineering, University of Padua, Italy
{emanuele.dibuccio,nicola.montecchio,nicola.orio}@dei.unipd.it
Abstract. This paper reports on the participation of the Information
Management System Research Group of the University of Padua to the
Music Identification task of the MusiCLEF Laboratory in CLEF 2011.
The system under evaluation is FALCON, an open-source engine for
content-based cover song identification written in Java that applies clas-
sic techniques derived from textual Information Retrieval to music iden-
tification. The obtained results show how such approach yields satisfying
results using little computational resources.
1 Introduction
Automatic identification of music documents has become an essential component
of many popular web services. Audio fingerprinting techniques are widely used
in order to identify copies of a recording, which can differ from the original
because of data compression, A/D conversion or environmental noises [1]. In most
major video sharing websites, the background music of user generated videos is
automatically identified in order to either remove the video (for copyright issues)
or suggest where to purchase the original music (advertising)1 .
Audio fingerprint approaches aim at the identification of a particular per-
formance/recording rather than work : even alternate takes of a composition are
considered different items instead of being regarded as different instances of
the same piece. This assumption imposes significant restrictions to the range of
possible differences between copies of the same recording, allowing audio finger-
printing techniques to be particularly efficient.
Aside from audio fingerprinting techniques, past research on music identifica-
tion has focused, for obvious commercial reasons, on popular music; the research
field is commonly referred to as cover song identification, a more general name
for this research field being version identification [2]. The problem however is
also of interest for other genres: in classical music there is especially a vast num-
ber of interpretations of the same work and version identification technology can
be beneficial for many music libraries and archives that aim at the preservation
and dissemination of classical music. For a comprehensive review of previous ap-
proaches to the version identification problem the reader can refer to [2], where
1
See, for instance, http://www.youtube.com/t/copyright_my_video
the author provides a review based on a functional block decomposition of pre-
viously proposed systems.
The experience gained in the development of text search engines shows that
in most cases simple and efficient techniques can be generally employed in various
retrieval tasks, with results that are often comparable with more complex and
less efficient approaches. Following this idea, we developed a music identification
engine named FALCON2 . The fact that our software is implemented on top of
Apache Lucene3 , an open source text search engine, substantiates the claim that
the problem of music identification can be modelled as a more general retrieval
task. In this paper we study its adaptability to classical music, making use of the
dataset prepared for the Music Identification task of the MusiCLEF 2011 bench-
marking activity. The collection is constituted by circa 7 thousand recordings for
which the MusiCLEF Laboratory organizers provided precomputed descriptors.
Testing FALCON on this test collection has allowed us to investigate the scala-
bility of our proposed approach, thus extending previous investigations carried
out on popular music [3].
2 Methodology
Our approach is based on a two-level bag-of-features hierarchy; the input de-
scriptors (chroma features) are transformed into hashes which are subsequently
grouped into a “set of hash sets” representation for each recording. The cardi-
nality of set intersection is adopted as a similarity measure. The methodology,
described in detail in [3,4], can be summarized as follows.
audio content analysis - this step consists in: (i) chroma feature extraction
from audio waveforms, (ii) key-finding — the most probable keys of a record-
ing are estimated in order to preserve transposition invariance — and (iii)
hashing of the transposed chroma features into a sequence of integer values
which form the output of this phase.
indexing - each sequence of hashes is segmented into a set of possible over-
lapping segments. Hashes are interpreted as terms in a textual document
and segments as passages constituted by sets of terms. This representation
can be easily stored in an inverted index: the set of all the distinct hashes
appearing in the sequences obtained from the recordings in the collection
constitutes the index vocabulary; each item in the posting list is associated
to an hash and retains information about the frequency of occurrence of the
hash in a specific segment.
querying - the similarity between a query Q and a recording D is computed
as v
uY
u n X hf(t, d) hf(t, q) o
S(Q, D) = |Q|t max min , (1)
d∈D |d| |q|
q∈Q t∈q∩d
2
http://ims.dei.unipd.it/falcon
3
http://lucene.apache.org/
where q and d denote two segments of a query and a collection recording,
respectively of length (number of hashes) |q|, |d|; similarly, |Q|, |D| denote
the number of distinct segments for Q, D and hf(t, d) denotes the frequency
of the hash t in the segment d. Formula 1 can be interpreted as follows: the
first step (inner summation) computes local similarity between segments as
the (normalized) number of terms they have in common; the second step
aggregates the contributions of all the query segments, by computing the
geometric mean of the best local similarities.
The procedure is repeated multiple times according to the most probable
keys detected in the audio analysis step; for each recording in the collection,
the highest similarity value is preserved.
3 Experiments
3.1 Test Collection and Setup
The test collection comprises 6679 recordings of classical music works, which add
up to more than 572 hours of music. Of these recordings, 2,671 are associated
to works that are represented at least twice in the data base, forming 945 cover
sets 4 . The audio descriptors were extracted using the MIRToolbox package [5].
More details on the collection, and in general on the MusiClef campaign, are
available in [6].
All experiments were repeated twice, one time using a key-finding algorithm
to preserve independency to transposition, and one time without such strategy.
The other parameters were set according to the values reported in [3], except
the segment overlap which was set to 50% the length of each segment (15s).
3.2 Identification Accuracy and Computational Load
In FALCON, the trade-off between accuracy and speed of retrieval privileges
the latter, as the architecture was designed with large collections in mind. It
is therefore important to measure the computational resources required by the
software. All experiments were run on a machine with a 3.4 GHz dual-core
processor (4 logical cores), 24 GB of RAM and a 7200 RPM hard disk.
Table 1 shows the results of our experimentation. As can be seen, the software
can index about 3.5 hours of music per second and is able to perform a typical
query in under 3 seconds. The average query time is different because the key-
finding algorithm repeats a query multiple times (in this case 3) in parallel; each
thread is completely independent of the others.
4
The term “cover set” is commonly used to define a set of recordings of the same
piece of music.
Table 1: Accuracy and timing for 667 queries over 6679 recordings.
no transposition with transposition
total indexing time 163s 166s
average query time 1.79s 2.32s
MAP 0.6754 0.6826
MRR 0.7617 0.7771
4 Concluding Remarks
This paper reports our contribution to the Music Identification task of the
MusiCLEF Laboratory in CLEF2011. Our experimental results show how a sim-
ple approach based on text retrieval techniques can yield satisfying results using
little computational resources.
These results suggest that the methodology implemented in FALCON can be
adopted as the first of a two-step methodology. The identification performed by
FALCON would aim at providing candidate versions of the query recording at
the top k rank positions. Then identification could be refined by means of more
sophisticated but also more resource consuming music alignment techniques,
that when exploited in isolation and on the entire collection are hardly scalable.
References
1. Cano, P., Batlle, E., Kalker, T., Haitsma, J.: A review of audio fingerprinting.
Journal of VLSI Signal Processing Systems 41 (November 2005) 271–284
2. Serrà, J.: Identification of versions of the same musical composition by processing
audio descriptions. PhD thesis, Universitat Pompeu Fabra, Barcelona (2011)
3. Di Buccio, E., Montecchio, N., Orio, N.: A scalable cover identification engine. In:
Proceedings of the international conference on Multimedia. MM ’10, New York, NY,
USA, ACM (2010) 1143–1146
4. Miotto, R., Orio, N.: A music identification system based on chroma indexing and
statistical modeling. In: Proceedings of the 9th International Conference on Music
Information Retrieval. ISMIR (2008) 301–306
5. Lartillot, O., Toiviainen, P.: A matlab toolbox for musical feature extraction from
audio. In: Proceedings of the 10th International Conference on Digital Audio Effects,
Bordeaux, France. (2007)
6. Orio, N., Rizo, D., Lartillot, O., Miotto, R., Montecchio, N., Schedl, M.: Musiclef:
A benchmark activity in multimodal music information retrieval. In: Proceedings
of the 12th International Conference on Music Information Retrieval (to appear).
ISMIR (2011)