Methods - Soundscape Ecology: Principles, Patterns, Methods and Applications - page 240

The spectral analysis allows in some cases the automated identification of the
species, but this is possible only in favorable context conditions (no wind, no rain,
no anthropogenic noise, no choruses, etc.). The difficulties in analyzing aurally
acoustic data can be in part overcome with a hybrid processing that consists of the
identification by experts of some key features of the acoustic data bank and then the
use of this key information to match the data processed automatically.
Here we describe a few selected metrics that are specifically devoted to the
identification of the (ecological) characters of a soundscape independently from the
source of the sounds.

9.7.2.1 Acoustic Entropy Index (H)

This index, based on the Shannon theory, which measures the information present
in a sonogram, was first developed by Sueur et al. ( 2008 ) and is composed of two
subindices:
The temporal entropy index H t and the spectral entropy index H f ,
where H ¼ H t H f with 0 H 1.

X

n

ðÞ

1

H t ¼ At ðÞ log 2 At ð ðÞ Þ log 2 n

t¼ 1

with H t ∈ [0,1]

X

N

ðÞ

1

H f ¼ Sf ðÞ log 2 Sf ð ðÞ Þ log 2 N

f ¼ 1

with H f ∈ [0,1]

where n ¼ length of the signal in number of digitized points

A ( t ) ¼ probability mass function of the amplitude envelope

probability mass function of the mean spectrum calculated using a short-
time Fourier transform (STFT), with a non-overlapping Hanning window of
N

S ( f ) ¼

512 points.
This index has been tested with success in a tropical forest, but when applied
to temperate conditions where the diversity of species is lower and the back-
ground noise is high, some biases result in the H f component as reported by
these authors.

¼

9.7.2.2 Median of amplitude envelope (M)

This index, proposed by Depraetere et al. ( 2012 ), is calculated as

2 1 depth

ð Þ with 0

M ¼ median At ðÞ M 1