Skip to main content

Table 1 Frame-level descriptors chosen by the feature-selection process on our dataset.

From: Ecological Acoustics Perspective for Content-Based Retrieval of Environmental Sounds

High frequency content
Instantaneous confidence of pitch detector (yinFFT)
Spectral contrast coefficients
Silence rate (−20 dB, −30 dB and −60 dB)
Spectral centroid
Spectral complexity
Spectral crest
Spectral spread
Shape-based spectral contrast
Ratio of energy per band (20–150 Hz, 150–800 Hz, 800–4 k Hz, 4 k–20 kHz)
Zero crossing rate
Inharmonicity
Tristimulus of harmonic peaks