Description of acoustic parameters extracted from songs (see Figure 3).

*For all parameters except 1–3, mean was calculated for notes within a song.
†Power values in short-time spectra and frequency bands that compose the spectrogram are summed to generate aggregate power envelopes and spectra, resulting in a power versus time envelope and power versus frequency spectrum, respectively.
The aggregates are normalized and treated as probability density functions with time or frequency being the variate, and density the fraction of the total signal energy. From the distribution function, various measures of central tendency and dispersion are then used to characterize the signal energy distribution in time and frequency. (See [39]).