mexcite.h

g729 coding ipaddressing

/**************************************************************************
*
* NAME
*				mexcite
*
* FUNCTION
*				Modify the stochastic code book excitation gain
*
* SYNOPSIS
*				subroutine mexcite1(l)
*				subroutine mexcite2(l)
*				subroutine mexcite3(cgain)
*
*	formal 
*						data	I/O
*		name			type	type	function
*		------------------------------------------------------------------
*		l				int 	i		length of error signal
*		cgain			float	i/o 	stochastic code book gain
*
*	global 
*						data	I/O
*		name			type	type	function
*		------------------------------------------------------------------
*		/ccsub/ 		see description include file
*		e0				float	i		error signal array
*		
***************************************************************************
*
* DESCRIPTION
*
*		Depending on the current system state, the stochastic code book
*		excitation is reduced to a level that is low enough to produce
*		positive perceptual effects, yet is high enough so as not to upset
*		the dynamics of the system.  The main effect of the method is that
*		during sustained voiced sounds, the excitation level is attenuated
*		and in unvoiced and transition regions the level is amplified to a
*		level slightly more than that of standard CELP.
*
*		The relative adaptive code book excitation component is
*		increased in voiced regions by decreasing the stochastic code book
*		excitation component.  The amount of decrease in the stochastic
*		component depends on the efficiency of the adaptive component.
*		More reconstruction burden is placed on the adaptive component as
*		its efficiency increases.  The efficiency is measured by the
*		closeness (in the crosscorrelation sense) of the residual signals
*		before and after pitch prediction.	When the efficiency is high
*		(e.g., > 0.9), the stochastic component is amplified slightly
*		(e.g., one quantizer level). 
*
*		The procedure for modifying the stochastic gain outside the
*		search loop is:
*		1)	Measure the efficiency of the adaptive component (ccor)
*		2)	Search the stochastic code book for the optimum codeword
*		3)	Modify the stochastic code book gain
*
*		This method is compatible with Federal Standard 1016.
*
*
*
***************************************************************************
*
* CALLED BY
*
*		csub csearch
*
* CALLS
*
***************************************************************************
* REFERENCES
*
*       Shoham, Yair, "Constrained-Stochastic Excitation Coding of Speech
*       at 4.8 kbps," in Advances in Speech Coding, ed. B. Atal, V.
*		Cuperman, and A. Gersho, submitted to Kluwer Academic Publishers.
*
*       Shoham, Yair, "Constrained-Stochastic Excitation Coding of Speech," 
*		Abstracts of the IEEE Workshop on Speech Coding for
*		Telecommunications, 1989, p. 65.
*
***************************************************************************/

#ifdef CELP_USE_CONTEXT
#define ccor	(ctx->MEXCITE_ccor)
#define e1		(ctx->MEXCITE_e1)
#define e0save	(ctx->MEXCITE_e0save)
#else
static float ccor;
static float e1, e0save[60];
#endif

static void mexcite1(int l)
{
  int i;

  /* *e1 = Euclidean norm of the first error signal 			*/
  /* (note: the error signal array e0 is reused)				*/

  e1 = 1e-6f;
  for (i = 0; i < l; i++)
  {
	e0save[i] = e0[i];
	e1 += e0[i] * e0[i];
  }
}

static void mexcite2(int l)
{
  int i;

  /* *ccor = crosscorrelation of the residual signals before	*/
  /* *and after pitch prediction								*/
  /* *(note: the error signal array e0 is reused				*/

  ccor = 1e-6f;
  for (i = 0; i < l; i++)
	ccor += e0[i] * e0save[i];

  /* *normalize the crosscorrelation							*/

  ccor = ccor / e1;
}

static void mexcite3(float *cgain)
{
  float scale;

  /* *square root crosscorrelation scaling						*/

  scale = (float) sqrt(fabs(ccor));

  /* *modify scale												*/

  if (scale < 0.2)
	scale = 0.2f;
  else if (scale > 0.9) 
	scale = (float) (scale * 1.4);

  /* *modify the stochastic component							*/

  *cgain = *cgain * scale;
}

#ifdef CELP_USE_CONTEXT
#undef ccor
#undef e1
#undef e0save
#endif