/*====================================================================================
EVS Codec 3GPP TS26.443 Jun 30, 2015. Version CR 26.443-0006
====================================================================================*/
#include <stdlib.h>
#include <math.h>
#include "options.h"
#include "cnst.h"
#include "prot.h"
/*---------------------------------------------------------------------*
* Local constants
*---------------------------------------------------------------------*/
#define K_COR_ENC 2.857f
#define C_COR_ENC -1.286f
#define K_EE_ENC 0.04167f
#define C_EE_ENC 0.0f
#define K_ZC_ENC -0.04f
#define C_ZC_ENC 2.4f
#define K_RELE_ENC 0.05f
#define C_RELE_ENC 0.45f
#define K_PC_ENC -0.07143f
#define C_PC_ENC 1.857f
#define K_SNR_ENC 0.1111f
#define C_SNR_ENC -0.3333f
/*-------------------------------------------------------------------*
* signal_clas()
*
* Classification state machine for FEC
* Coder type modification
*-------------------------------------------------------------------*/
short signal_clas( /* o : classification for current frames */
Encoder_State *st, /* i/o: encoder state structure */
short *coder_type, /* i/o: coder type */
const float voicing[3], /* i : normalized correlation for 3 half-frames */
const float *speech, /* i : pointer to speech signal for E computation */
const short localVAD, /* i : vad without hangover */
const short pit[3], /* i : open loop pitch values for 3 half-frames */
const float *ee, /* i : lf/hf E ration for 2 half-frames */
const float relE, /* i : frame relative E to the long term average */
const short L_look, /* i : look-ahead */
short *uc_clas /* o : flag for unvoiced class used in sp/mus classifier */
)
{
float mean_voi2, mean_ee2, tmp;
float een, corn, zcn, relEn, pcn, fmerit1;
short i, clas, pc, zc;
short unmod_coder_type;
/*----------------------------------------------------------------*
* Calculate average voicing
* Calculate average spectral tilt
* Calculate zero-crossing rate
* Calculate pitch stability
*----------------------------------------------------------------*/
/* average voicing on second half-frame and look-ahead */
mean_voi2 = 0.5f * (voicing[1] + voicing[2]);
/* average spectral tilt in dB */
tmp = ee[0] * ee[1];
if( tmp < 1.0f )
{
tmp = 1.0f;
}
mean_ee2 = 0.5f * 20.0f * (float)log10( tmp );
/* compute zero crossing rate */
zc = 0;
for( i = L_look; i < L_FRAME + L_look; i++ )
{
if( speech[i] <= 0.0f && speech[i-1] > 0.0f )
{
zc++;
}
}
/* compute pitch stability */
pc = (short)(abs(pit[1] - pit[0]) + abs(pit[2] - pit[1]));
/*-----------------------------------------------------------------*
* Transform parameters to the range <0:1>
* Compute the merit function
*-----------------------------------------------------------------*/
een = K_EE_ENC * mean_ee2 + C_EE_ENC;
if( een > 1.0f )
{
een = 1.0f;
}
else if( een < 0.0f )
{
een = 0.0f;
}
corn = K_COR_ENC * mean_voi2 + C_COR_ENC;
if( corn > 1.0f )
{
corn = 1.0f;
}
else if( corn < 0.0f )
{
corn = 0.0f;
}
zcn = K_ZC_ENC * zc + C_ZC_ENC;
if( zcn > 1.0f )
{
zcn = 1.0f;
}
else if( zcn < 0.0f )
{
zcn = 0.0f;
}
relEn = K_RELE_ENC * relE + C_RELE_ENC;
if( relEn > 1.0f )
{
relEn = 1.0f;
}
else if( relEn < 0.5f )
{
relEn = 0.5f;
}
pcn = K_PC_ENC * pc + C_PC_ENC;
if( pcn > 1.0f )
{
pcn = 1.0f;
}
else if( pcn < 0.0f )
{
pcn = 0.0f;
}
fmerit1 = (1.0f/6.0f) * (een + 2.0f*corn + zcn + relEn + pcn);
/*-----------------------------------------------------------------*
* FEC classification
* Onset classification
*-----------------------------------------------------------------*/
/* FEC classification */
if( localVAD == 0 || *coder_type == UNVOICED || relE < -6.0f )
{
clas = UNVOICED_CLAS;
}
else
{
switch( st->last_clas )
{
case VOICED_CLAS:
case ONSET:
case VOICED_TRANSITION:
if( fmerit1 < 0.49f )
{
clas = UNVOICED_CLAS;
}
else if( fmerit1 < 0.66f )
{
clas = VOICED_TRANSITION;
}
else
{
clas = VOICED_CLAS;
}
break;
case UNVOICED_CLAS:
case UNVOICED_TRANSITION:
if( fmerit1 > 0.63f )
{
clas = ONSET;
}
else if( fmerit1 > 0.585f )
{
clas = UNVOICED_TRANSITION;
}
else
{
clas = UNVOICED_CLAS;
}
break;
default:
clas = UNVOICED_CLAS;
break;
}
}
/* set flag for unvoiced class, it will be used in sp/mus classifier */
*uc_clas = clas;
if( ( ( *coder_type == UNVOICED ) ||
( st->input_bwidth != NB && fmerit1 < 0.41f && st->mold_corr > 0.65f ) || /* WB case */
( st->input_bwidth == NB && fmerit1 * 0.88f < 0.41f && st->mold_corr > 0.55f ) ) && /* NB case */
relE > -15.0f && st->lt_dec_thres < 1.5f )
{
*uc_clas = UNVOICED_CLAS;
}
/* Onset classification */
/* tc_cnt == -1: frame after TC frame in continuous block of GC/VC frames */
/* tc_cnt == 0: UC frame */
/* tc_cnt == 1: onset/transition frame, coded by GC mode */
/* tc_cnt == 2: frame after onset/transition frame, coded by TC mode */
if( clas == UNVOICED_CLAS )
{
st->tc_cnt = 0;
}
if( clas >= VOICED_TRANSITION && st->tc_cnt >= 0 )
{
st->tc_cnt += 1;
}
if( st->tc_cnt > 2 )
{
st->tc_cnt = -1;
}
if ( st->codec_mode == MODE1 )
{
/*---------------------------------------------------------------------*
* Coder type modification
*
* Prevent UC mode in certain conditions
* Prevent VC mode in certain conditions
* Select TC mode in appropriate frames
*---------------------------------------------------------------------*/
/* At higher rates, use GC coding instead of UC coding to improve quality */
if( st->total_brate > ACELP_9k60 && *coder_type == UNVOICED )
{
*coder_type = GENERIC;
}
/* Prevent UC coding on mixed content at 9.6 kb/s */
if( st->total_brate == ACELP_9k60 && *coder_type == UNVOICED && st->audio_frame_cnt != 0 )
{
*coder_type = GENERIC;
}
unmod_coder_type = *coder_type;
/* Enforce GC mode on inactive signal (this can be later overwritten to INACTIVE) */
if( localVAD == 0 && ( (*coder_type == UNVOICED && (!st->Opt_SC_VBR ||
( st->Opt_SC_VBR && st->vbr_generic_ho == 0 && st->last_coder_type > UNVOICED )) )
|| *coder_type == TRANSITION || *coder_type == VOICED )
)
{
*coder_type = GENERIC;
}
if( *coder_type == GENERIC && unmod_coder_type == UNVOICED && st->Opt_SC_VBR )
{
st->vbr_generic_ho = 1;
}
if ( *coder_type > UNVOICED && st->Opt_SC_VBR )
{
st->vbr_generic_ho = 0;
}
if( localVAD == 0 && *coder_type == UNVOICED )
{
st->last_7k2_coder_type = GENERIC;
}
else
{
st->last_7k2_coder_type = *coder_type;
}
/* Select TC mode for appropriate frames which is in general MODE1_VOICED_TRANSITION, VOICED_CLAS or MODE1_ONSET frames following UNVOICED_CLAS frames */
if( localVAD != 0 && st->tc_cnt >= 1 )
{
if ( st->tc_cnt == 1 )
{
/* onset/transition frame is always coded using GC mode */
*coder_type = GENERIC;
}
else
{
/* frame after onset/transition frame is coded by TC mode */
*coder_type = TRANSITION;
}
}
/* At higher rates and with 16kHz core, allow only GC and TC mode */
if( st->total_brate >= ACELP_24k40 && *coder_type != GENERIC && *coder_type != TRANSITION )
{
*coder_type = GENERIC;
}
/* Patch for certain low-level signals for which the gain quantizer sometimes goes out of its dynamic range */
if( *coder_type == VOICED && st->input_bwidth == NB && relE < -10.0f && st->total_brate <= ACELP_8k00 )
{
*coder_type = GENERIC;
}
}
return clas;
}