# What are the ranges of values for each audio feature/descriptor?

I need to normalize some audio feature measurements I've done with the software Wolfram Mathematica.

Coming from a non signal processing background, I have some difficulties understanding what several audio features are actually measuring and thus the range of values in which each measurement is expressed.

I can't seem to find any documentation regarding the "physical" range of these descriptors, both in the Wolfram software and the MPEG-7 standard.

This is the list of features analyzed, using the names given to them by Mathematica:

{"Max", "MaxAbs", "Min", "MinAbs", "MinMax", "MinMaxAbs", "Mean",
"Median", "StandardDeviation", "Total", "Power", "RMSAmplitude",
"Loudness", "CrestFactor", "Entropy", "LPC",
"PeakToAveragePowerRatio", "TemporalCentroid", "ZeroCrossingRate",
"ZeroCrossings", "FundamentalFrequency", "Formants",
"HighFrequencyContent", "MFCC", "SpectralCentroid", "SpectralCrest",
"SpectralFlatness", "SpectralKurtosis", "SpectralRollOff",
"ComplexDomainDistance", "ModifiedKullbackLeibler", "Novelty",
"PhaseDeviation", "SpectralFlux"}


I can suppose some of the ranges, e.g. the one of Spectral Centroid, are similar to the range of the audible spectrum, but I have no clue about the more exotic ones.

Is there any trusted source I can rely on?

• Hi, Mumen Raida, and welcome to DSP.SE! May 1 '19 at 12:42
• See also this question and its answers - especially this site... May 1 '19 at 12:43