Kurt's Works

Community detection methods from complex network theory are applied to a subset of the Myspace artist network to identify groups of similar artists. Methods based on the greedy optimization of modularity and random walks are used. In a second iteration, inter-artist audio-based similarity scores are used as input to enhance these community detection methods. The resulting community structures are evaluated using a collection of artist-assigned genre tags.

A sample of the Myspace artist network is examined to investigate the relationship between social connectivity and audio-based similarity.  Audio data from the Myspace artist pages is analyzed using well-established signal-based music information retrieval techniques.  In addition to showing that the Myspace artist network exhibits many of the properties common to social networks, we show there is an ambiguous relationship between audio-based similarity and the social network topology.

In large audio collections, it is common to store audio content with perceptual encoding. However, encoding
parameters may vary from collection to collection or even within a collection - using different bit rates,
sample rates, codecs, etc. We evaluate the effect of various audio encodings on the onset detection task.
We show that audio-based onset detection methods are surprisingly robust in the presence of MP3 encoded
audio. Statistically significant changes in onset detection accuracy only occur at bit-rates lower than 32kbps.

A sample of the Myspace social network is examined. Using
methods from complex network theory, we show empirically that artists tend to form on-line social connections with artists of the same genre. This motivates the use of on-line social networks as data resources for musicology and music information retrieval.

In large audio collections, it is common to store audio content using perceptual encoding. However, encoding
parameters may vary from collection to collection or even within a collection - using different bit rates, sample
rates, codecs, etc. We evaluate the effect of various lossy audio encodings on the application of audio spectrum
projection features to the automatic genre classification tasks. We show that decreases in mean classification
accuracy, while small, are statistically significant for bit-rates of 96kbps or lower. Also, a heterogeneous

[10/2007] Paper for the 123rd AES convention in which we match drum loops against a database of unheard music signals to automatically apply a text label describing the rhythmic style of the music signal PDF

[10/2006] ISMIR 2006 paper describing what is basically the same audio-based music similarity system as described in my U Miami thesis PDF

my master's thesis describing an audio-based music similarity system PDF