Programmer Guide/Command Reference/EVAL/formants: Difference between revisions
From STX Wiki
< Programmer Guide | Command Reference | EVAL
Jump to navigationJump to search
No edit summary |
No edit summary |
||
Line 12: | Line 12: | ||
::*if ''f0'' is supplied, the formant tracking is only applied to voiced regions | ::*if ''f0'' is supplied, the formant tracking is only applied to voiced regions | ||
:;<var>mflag</var>selects the tracking method | :;<var>mflag</var>selects the tracking method | ||
::*currently only one method (''mflag''=0) is implemented | ::*currently only one method (''mflag''=0) is implemented (see [[Programmer_Guide/SPU_Reference/FORMANTS|sp-atom FORMANTS]] for more details) | ||
::*a second method using a formant modell is under development | ::*a second method using a formant modell is under development | ||
:;<var>bias</var>:tracking bias in Hz; This value is used to check if two candidates can be connected (<code>|cf[i1,j1]-cf[i2,j2]) / (cf[i1,j1]+cf[i2,j2]+2*bias)</code> must be lower than '''0.1'''). | :;<var>bias</var>:tracking bias in Hz; This value is used to check if two candidates can be connected (<code>|cf[i1,j1]-cf[i2,j2]) / (cf[i1,j1]+cf[i2,j2]+2*bias)</code> must be lower than '''0.1'''). | ||
:;<var>lmin</var>:the minimum length of a track in frames | :;<var>lmin</var>:the minimum length of a track in frames | ||
;Result: The f0 vector. The vector contains the f0 values in Hz. The vector length depends mainly on the length of the signal vector ''x'' and the hopsize ''dt''. For unvoiced frames the f0 values are set to zero. | ;Result: The f0 vector. The vector contains the f0 values in Hz. The vector length depends mainly on the length of the signal vector ''x'' and the hopsize ''dt''. For unvoiced frames the f0 values are set to zero. | ||
;See also: [[../f0ac|f0ac]], [[../lpc|lpc]], [[../cepstrum|cepstrum]], [[../ipeaks|ipeaks]], [[Programmer_Guide/SPU_Reference/FORMANTS]] | ;See also: [[../f0ac|f0ac]], [[../lpc|lpc]], [[../cepstrum|cepstrum]], [[../ipeaks|ipeaks]], [[Programmer_Guide/SPU_Reference/FORMANTS|FORMANTS]] | ||
[[../#Functions|<function list>]] | [[../#Functions|<function list>]] |
Revision as of 14:51, 20 April 2011
Format tracking algorithm.
- Usage
formants(cf, ca, f0, mflag, bias, lmin)
- cf
- a matrix containing the frequencies of the formant candidates in Hz
- ca
- a matrix containing the amplitudes of formant candidates in dB
- each row of cf/ca contains the candidates of one frame
- number of frames: N=
nrow(cf)
- maximum number of candidates M=
ncol(cf)
- ca must be a matrix with the same dimensions as cf or a scalar (if amplitudes should not be used for formant tracking)
- f0
- f0 vector
- f0 must be a vector with length N containing the f0 values or a scalar (if f0 should not be used for formant tracking)
- if f0 is supplied, the formant tracking is only applied to voiced regions
- mflagselects the tracking method
-
- currently only one method (mflag=0) is implemented (see sp-atom FORMANTS for more details)
- a second method using a formant modell is under development
- bias
- tracking bias in Hz; This value is used to check if two candidates can be connected (
|cf[i1,j1]-cf[i2,j2]) / (cf[i1,j1]+cf[i2,j2]+2*bias)
must be lower than 0.1). - lmin
- the minimum length of a track in frames