Table 1.
Related Work.
Fig 1.
Library General Diagram.
Fig 2.
pyAudioAnalysis provides easy-to-use and high-level Python wrappers for several audio analysis tasks.
Table 2.
Audio Features.
Fig 3.
Local maxima detection for beat extraction.
An example of local maxima detection on each of the adopted short-term features. The time distances between successive local maxima are used in the beat extraction process.
Fig 4.
An aggregated histogram of time distances between successive feature local maxima. The histogram’s maximum position is used to estimate the BPM rate.
Fig 5.
Supervised segmentation results and statistics for a radio recording. A binary speech vs music classifier is used to classify each fix-sized segment.
Table 3.
HMM joint segmentation classification performance.
Fig 6.
An example of applying the silence removal method on an audio recording. Upper subfigure represents the audio signal, while the second subfigure shows the SVM probabilistic sequence.
Table 4.
HMM joint segmentation classification performance.
Fig 7.
Example of a self-similarity matrix for the song “Charmless Man” by Blur. The detected diagonal segment defines the two thumbnails, i.e. segments (115.0sec–135.0sec) and (156.0sec–176.0sec).
Fig 8.
Chordial Content Visualization Example.
Different colors of the edges and nodes (recordings) represent different categories (artists in our case).
Table 5.
Realtime ratios for some basic functionalities and different devices.