Digital Signal Processing Reference
In-Depth Information
Appendix
openSMILE Standardised Feature Sets
All's well that ends well.
—William Shakespeare.
In Table
A.1
the LLDs and functionals and their frequency across the four
openSMILE standard feature sets as were mentioned in this topic are given.
LLDs are processed by simple moving average (SMA) low-pass filtering.
Delta regression coefficients are added per LLD. The total number of features
is—in principle—obtained by multiplying the number of LLD times two times the
number of functionals. However, for the two larger feature sets exceptions hold
from this strict brute-forcing rule as are indicated. This prevents creation of non-
sense features.
