PowerPointPrsentation

1
Music Information Retrieval
Thomas Lidy, Andreas Rauber
audio files wav/mp3
extract features
Rhythm Pattern
search similar songs
classify into genres
music landscapes
2


Feature Extraction
S6 Sone (specific loudness sensation)
conversion to mono segmentation (5.9 sec)
S1 spectrogram (FFT)
R1 modulation amplitude spectrum (FFT)
S2 Bark scale (24 critical bands)
R2 fluctuation strength model
S3 spectral masking
R3 gradient filter Gaussian smoothing
S4 decibelSPL
summarization median over segments
S5 Phon (equal loudness)

Rhythm Pattern loudness modulation on all
frequency bands