ADPAR1
Contents
ADPAR1
The spectral weighting (if AW=1) and the frequency range truncation (FMIN .. FMAX) is applied to all input spectra. The further processing of the spectra depends on the selected TYPE.
TYPE=0
The RMS energy (in dB) of the spectrum (band) is computed and stored in the output vector Y.
TYPE=1
|2
The pre-processed spectra are converted to dB and copied to the output array SPG (1 row per spectrum).
TYPE=2
After collecting all input spectra, the following descriptors are computed.
Y[i,0] = rms - RMS energy of the spectrum (dB)
Y[i,1] = am - mean amplitude of the spectrum (dB)
Y[i,2] = ad - standard deviation of the spectrum (dB)
Y[i,3] = sl - spectral slope (dB / kHz)
Y[i,4] = acno - mean amplitude of the normalized spectrum (dB)
Y[i,5] = fcno - spectral centroid frequency (Hz)
Y[i,6] = spno - spectral spread
Y[i,7] = fx - spectral flux (1st order, correlation of spectrum i-1 and i)
Y[i,8] = skno - spectral skewness
Note: all descriptors with the subscript no are computed from the normalized spectrum. The spectra are normalized as follows:
offset = max(
SPG[*,*]) -
ARANGEif A[i,j]>offset
, then Ano[i,j]=
A[i,j]-offset
otherwise, Ano[i,j] = 0
with: i
= frameindex, j
= frequency bin indexThis SPAtom was developed for the NOIDESc project in 2006.
Usage:
ADPAR1 TYPE N A AW AREF ARANGE DF FMIN FMAX
Parameters:
- TYPE
- The following types are supported:
0
= RMS track
1
= spectrogram
2
= descriptors
- N
- The number of frames (>
1
).
- A
- The linear amplitude spectrum.
- AW
- Apply amplitude weighting:
NONE
|0
- no spectrum weighting
A
|1
- spectrum weighting in dBA.
- AREF
- The reference amplitude (linear).
- ARANGE
- The amplitude normalization range for descriptors (TYPE=
2
) in dB.
- DF
- The FFT frequency resolution in Hz.
- FMIN, FMAX
- The analysis frequency range.
Outputs:
- Y
- If TYPE==
0
, the rms track (rms[0..N-1]).
- If TYPE==
1
, the spectrogram array (one spectrum per row).
- SPG
- The logarithmic amplitude spectrogram.
- If TYPE=
0
- this is not initialized.
- If TYPE=
1
|2
- N xnf
(= the number of frequency bins in the selected range).