Using estimate

data type of the input file

Five types of input data file can be used in estimate :

• fmt extension file (binary file of sinusoidal partials; see additive )
• format extension file (ascii file of sinusoidal partials; see additive )
• sdif extension file (SDIF file ofsinusoidal partials; see additive )
• sf or aiff extension file (signal file)

-t data-type : data-type of input : format |fmt | sf |aiff
(default: guess from input-file extension)

Note that the data type of the input file determines the spectral envelope calculation method. If the input is a file of sinusoidal partials, then the discrete cepstrum method is used. If the input is a sound file, then the lpc method is used.

data stored in the output file

The data which can be stored in the output file are the following :

• In the case of discrete cepstrum estimation, spectral envelope and/or cepstral coefficients can be stored in the output file ;
• In the case of lpc estimation, spectral envelope and/or autoregressive coefficients (a_i) and/or reflection coefficients (k_i) and/or correlation coefficients (r_i) can be stored in the output file.

-E : spectral envelope
-C : coefficients (cepstral coefficients in case of discrete cepstrum estimation ; autoregressive coefficients in case of lpc estimation)
--Cc : cepstral coefficients (in case of discrete cepstrum estimation, this option is equivalent to option -C)
--Ca : lpc autoregressive coefficients (in case of lpc estimation, this option is equivalent to option -C)
--Ck : lpc reflection coefficients
--Cr : lpc correlation coefficients
-a : output ascii (default : sdif).

Note that the last options (excepted -a option) start with TWO minus signs.

Note that the ascii output option (-a) can not be used when any -C or --C option (-C, --Cc,--Ca, --Ck, --Cr) is selected. When any -C or --C option is used, the coefficients are always stored in an SDIF file (together with the envelope, if option -E).

Remark : An sdif file MUST have sdif extension.

lpc parameters

-o order : lpc order (default: 50).

--Ws size : window size for lpc estimation (default :1024 points). Note that this option starts with TWO minus signs.

Remark : lpc estimation uses a Blackman window on the signal. The enveloppe is calculated on the frequency band [0,sampling rate/2] where the sampling rate is the one of the input file.
--Ca : lpc autoregressive coefficients in output file

--Ck : lpc reflection coefficients in output file

--Cr : lpc correlation coefficients in output file

Note that the last options start with TWO minus signs.

discrete cepstrum parameters

-M freq : estimate discrete cepstrum envelope up to freq in Hz (default: 24kHz).
freq defines the upper limit of the band [0,freq] relative to which the cepstral coefficients are calculated. Furthermore, only the partials with frequency lower than freq are considered in the estimation.

-o order : discrete cepstrum order (default: 50).

-r factor : regularization factor for discrete cepstrum estimation
(default: no regularization)(example of possible value: 0.0001).

-c : use cloud smoothing for discrete cepstrum estimation.(default : do not cloud smoothing)

--Cc : cepstral coefficients in output file.
This option is equivalent to option -C in case of discrete cepstrum estimation. Note that this option starts with TWO minus signs.

-b freq : use a logarithmic frequency scale above freq Hz for computing the envelope; for frequencies inferior to freq, a linear frequency scale is used (default: linear).

general parameters

-N number : number of points of the estimated envelope (default : 512).

-p sdif_preset_file : load the estimation parameters from the sdif_preset_file.
This command is very useful when you want to do estimation on many files with the same parameters. In order to make an sdif-preset-file, you have to apply estimate on an empty file (sound or partials). The generated sdif_preset_file contains only the estimation parameters. See example below.

Examples