# Compressing A Range Of Decibel Values to One “Representative” Value

Don't have any formal dsp experience - sophomore in college.

Working on constructing a 25 band bark spectrogram for psychoacoustic analysis of audio FFT data. To do this, I'm grouping all FFT frequency bins into their relevant bark band and then averaging (summing them and dividing by the total number) the db values in each band to get the representative value for each band.

Is this the correct way to represent a bunch of db values as a single value?

-
"dB values" are really just a way of representing values on a logarithmic scale. Arithmetic-averaging values on a logarithmic scale is equivalent to geometrically averaging the original values (in linear scale). This probably isn't what you want. You probably would like a measure of total power in each band, which you would get by just summing the power spectrum bins across each band. You could divide by the number of bins, but if the bands aren't equal-width, that will add some bias to the result. –  Jason R Feb 19 '13 at 13:04
+1, Could be promoted to an answer! –  pichenettes Feb 19 '13 at 13:56
How do I promote to answer? Thanks Jason. –  Newmu Feb 21 '13 at 19:49

3dB + 3dB != 6dB