Analysis - ana




Analysis - ana

Ana does harmonic additive sinusoidal and spectral envelope analysis. It analyses a file either at equidistant steps, or pitch synchronous.

Ana reads the files with the following suffixes appended to the stem of the sound file name (generated by PetC):

_T<step>.ta
Analysis points in samples (int)

i_T<step>.pitch
Fundamental frequency (float)

i_<mode -C>_T<step>.coupe
Cut frequency between voiced and unvoiced band (float)

i_<mode -C>_T<step>.coupe_cod
If coding is enabled (-q, attention to choose an appropriate number of bits with -C), cut frequency between voiced and unvoiced band (float)

_T<step>.rien
Detection of absence of (speech) signal (int, 1 if nothing, 0 if signal present)

It generates the following files (the _ps output files are written only in pitch synchronous mode):

_T<step>.ta_ps
Analysis times (float)

_T<step>.phase[_ps]
Exact phases: analysis time (float), number of frame (int), number of phases (int), phases (float)

_T<step>.ampl[_ps]
Amplitudes: analysis time (float), number of frame (int), number of amplitudes (int), amplitudes (float)

i_<mode -C>_T<step>.coupe
Cut frequency between voiced and unvoiced band (float)

_T<step>.en[_ps]
Energy (float)

_T<step>.rien_ps
Absence of (speech) signal (int, 1 if nothing, 0 if signal present)

_T<step>.trans[_ps]
Transitions: head: number of transitions nt (int), f0 of transitions (float), then per frame: analysis time (float), detection (int, 0 if no transition, number of partials np if it is), np amplitudes of partials (float), np phases of partials (float), nt delay indices (n_j) for enhanced harmonic coding with frequency domain transition modeling (int), nt gain values (float)

_T<step>.env<method>[_b|_d]_<order>[_ps]
Envelope parameters. The file with _b contains the residual noise envelope, the file without contains the envelope of the sinusoidal part, which is denoised, if _b is present (flag -d). Order is specified by -o/-b for sinusoidal or noise, respectively.

Possible methods (options -E/-B):

Envelope file format
One frame consists of: analysis time (float), number of frame (int), absence of signal (int 1 or 0), number of coefficients (int, 0 when no signal present), coefficients (float), which are either cepstral coefficients (for cepstral methods Cx) or the reflection coefficients k_i (for AR methods ARx).



Usage

ana :  
	 ana est un programme d'Analyse  
 
 description:  
	 le programme necessite le nom du fichier original (sans extension):  
USAGE:  
 ana -I originalfile pour une analyse a pas constant 
 ana -O originalfile pour une analyse synchronis'ee sur le pitch 
 
 If originalfile starts with '-', native short samples are read from stdin, 
 otherwise the given file is read in raw SGI short format. 
 
 Output files are created in a directory whose name is the stem of the 
 filename part of the soundfile.  The directory must exist. 
 
OPTIONS:  
 NO SPACE between flag and value. 
 The default is given in (parenthesis) or marked by an asterisk*. 
 
 pour avoir de l'aide -h  
 -F freq_ech(Hz) (8000)  
 -T temps entre deux instants d'analyse (ms) (20ms)  
 -E type de calcul pour l'enveloppe (1*:DC 2:AR 3:CPLE 4:ARPLE)  
 -o ordre de la parametrisation (44 pour le cepstre et 10 en AR) 
 -p ordre de la parametrisation cepstrale intermediaire (20 par defaut) 
 -C Flag_coupe ( 0:Fc=plus grande frequence voisee, 1 et 2=nbre de bits utilises au choix  
 -c + valeur de la Fc (fixee)  
 -d + dsp de bruit (debruite et prend le resultat du VAD au lieu de calculer RIEN)  
 -P Flag_pics (0*: TFD et 1: msq  2: wmsq 3:msql 4:wmsql 
 -B enveloppe de bruit (0: lpc et 1: cepstre optimise)  
 -b ordre de l'enveloppe de bruit 
 -L valeur du facteur de regularisation en cas d'optimisation (1e-3) 
 -v verbose  
 -q avec options de codage



Source Code

src/main/ana.c ( 22.1KB, 29-Apr-1999)
Main source file

Top Back to top - Diemo Schwarz - last update Tue May 04 21:28:29 1999