Music analysis and retrieval systems for audio signals

Abstract

The constantly increasing amount of audio available in digital form necessitates the development of software systems for analyzing and retrieving digital audio. In this work, we describe our efforts in developing such systems. More specifically, we describe the design philosophy behind our approach, the specific problems we try to solve, and how we evaluate the performance of our algorithms. Automatic music analysis and retrieval of non-speech digital audio is a relatively new field, and the existing techniques are far from perfect. To improve the performance of the developed techniques, two main techniques are used: (1) integration of information from multiple analysis and retrieval algorithms and (2) the use of graphical user interfaces that enable the user to provide feedback to the design, development, and evaluation of the algorithms. All the developed algorithms and user interfaces are integrated under MARSYAS, a software framework for research in computer audition.

The State of the Art Ten Years After a State of the Art: Future Research in Music Information Retrieval

“…Half of these report results from a single split [8,10,12,27,45,52]; but the other half reports a mean of many trials, e.g., 5 [43], 30 [97], and 100 trials [65,81,92,93]. Seven papers [6,21,38,56,59,69,87] do not coherently describe their design parameters.…”

“…GTZAN is composed of 1,000 half-minute music audio excerpts singly labeled in ten categories [92,93]; and though its use is so widespread, it has always been missing metadata identifying its contents. In fact, GTZAN was not expressly created for MGR, 2 but its availability has made it a benchmark dataset, and thus a measuring stick for comparing MGR systems, e.g., [102].…”

Are the Emotions Expressed in Music Genre-specific? An Audio-based Evaluation of Datasets Spanning Classical, Film, Pop and Mixed Genres

“…All features were extracted with the MIRtoolbox 2 ) using a framebased approach (Tzanetakis & Cook, 2002a Harte, Sandler, & Gasser, 2006), the centroid of an uncollapsed Chromagram], the frames were 2 s long, with an overlap of 50%, while for structural features (Repetition), frame length was 100 ms and overlap also at 50%. The results from the frames were then summarized by either the mean, the standard deviation, or the slope function.…”

Deploying music characteristics for an affective music player

“…This is summarized in Table 1, which provides concrete design guidelines for the use of music characteristics in an AMP. Songs in music databases can be tagged with music characteristics by employing one of the algorithms proposed by [19,26,31,36]. Then, using the guidelines provided, an AMP can be implemented directly.…”

“…It is generally accepted that music characteristics can influence affect [13,18,27,33]. Consequently, several algorithms were already introduced for content analysis of music [19,25,31,36]. They are also employed in music systems like last.fm [17]; although, for other purposes than to direct affect.…”