FreeCalypso > hg > gsm-codec-lib
view libtwamr/qgain475.c @ 468:4104b0390fab
efrtest: new program gsmefr-dlcap-sync
author | Mychaela Falconia <falcon@freecalypso.org> |
---|---|
date | Mon, 13 May 2024 07:21:09 +0000 |
parents | 1d2b39027b70 |
children |
line wrap: on
line source
/* ******************************************************************************** * * GSM AMR-NB speech codec R98 Version 7.6.0 December 12, 2001 * R99 Version 3.3.0 * REL-4 Version 4.1.0 * ******************************************************************************** * * File : qg475.c * Purpose : Quantization of pitch and codebook gains for MR475. * ******************************************************************************** */ /* ******************************************************************************** * MODULE INCLUDE FILE AND VERSION ID ******************************************************************************** */ #include "namespace.h" #include "qgain475.h" /* ******************************************************************************** * INCLUDE FILES ******************************************************************************** */ #include "tw_amr.h" #include "typedef.h" #include "basic_op.h" #include "mac_32.h" #include "no_count.h" #include "cnst.h" #include "pow2.h" #include "log2.h" #include "qua_gain_tab.h" /* ******************************************************************************** * LOCAL VARIABLES AND TABLES ******************************************************************************** */ /* minimum allowed gain code prediction error: 102.887/4096 = 0.0251189 */ #define MIN_QUA_ENER ( -5443) /* Q10 <-> log2 (0.0251189) */ #define MIN_QUA_ENER_MR122 (-32768) /* Q10 <-> 20*log10(0.0251189) */ /* minimum allowed gain code prediction error: 32000/4096 = 7.8125 */ #define MAX_QUA_ENER ( 3037) /* Q10 <-> log2 (7.8125) */ #define MAX_QUA_ENER_MR122 ( 18284) /* Q10 <-> 20*log10(7.8125) */ /* ******************************************************************************** * PRIVATE PROGRAM CODE ******************************************************************************** */ static void MR475_quant_store_results( gc_predState *pred_st, /* i/o: gain predictor state struct */ const Word16 *p, /* i : pointer to selected quantizer table entry */ Word16 gcode0, /* i : predicted CB gain, Q(14 - exp_gcode0) */ Word16 exp_gcode0, /* i : exponent of predicted CB gain, Q0 */ Word16 *gain_pit, /* o : Pitch gain, Q14 */ Word16 *gain_cod /* o : Code gain, Q1 */ ) { Word16 g_code, exp, frac, tmp; Word32 L_tmp; Word16 qua_ener_MR122; /* o : quantized energy error, MR122 version Q10 */ Word16 qua_ener; /* o : quantized energy error, Q10 */ /* Read the quantized gains */ *gain_pit = *p++; move16 (); g_code = *p++; move16 (); /*------------------------------------------------------------------* * calculate final fixed codebook gain: * * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ * * * * gc = gc0 * g * *------------------------------------------------------------------*/ L_tmp = L_mult(g_code, gcode0); L_tmp = L_shr(L_tmp, sub(10, exp_gcode0)); *gain_cod = extract_h(L_tmp); /*------------------------------------------------------------------* * calculate predictor update values and update gain predictor: * * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ * * * * qua_ener = log2(g) * * qua_ener_MR122 = 20*log10(g) * *------------------------------------------------------------------*/ Log2 (L_deposit_l (g_code), &exp, &frac); /* Log2(x Q12) = log2(x) + 12 */ exp = sub(exp, 12); tmp = shr_r (frac, 5); qua_ener_MR122 = add (tmp, shl (exp, 10)); L_tmp = Mpy_32_16(exp, frac, 24660); /* 24660 Q12 ~= 6.0206 = 20*log10(2) */ qua_ener = round (L_shl (L_tmp, 13)); /* Q12 * Q0 = Q13 -> Q10 */ gc_pred_update(pred_st, qua_ener_MR122, qua_ener); } /* ******************************************************************************** * PUBLIC PROGRAM CODE ******************************************************************************** */ /************************************************************************* * * FUNCTION: MR475_update_unq_pred() * * PURPOSE: use optimum codebook gain and update "unquantized" * gain predictor with the (bounded) prediction error * *************************************************************************/ void MR475_update_unq_pred( gc_predState *pred_st, /* i/o: gain predictor state struct */ Word16 exp_gcode0, /* i : predicted CB gain (exponent MSW), Q0 */ Word16 frac_gcode0, /* i : predicted CB gain (exponent LSW), Q15 */ Word16 cod_gain_exp, /* i : optimum codebook gain (exponent), Q0 */ Word16 cod_gain_frac /* i : optimum codebook gain (fraction), Q15 */ ) { Word16 tmp, exp, frac; Word16 qua_ener, qua_ener_MR122; Word32 L_tmp; /* calculate prediction error factor (given optimum CB gain gcu): * * predErrFact = gcu / gcode0 * (limit to MIN_PRED_ERR_FACT <= predErrFact <= MAX_PRED_ERR_FACT * -> limit qua_ener*) * * calculate prediction error (log): * * qua_ener_MR122 = log2(predErrFact) * qua_ener = 20*log10(predErrFact) * */ if (test(), cod_gain_frac <= 0) { /* if gcu <= 0 -> predErrFact = 0 < MIN_PRED_ERR_FACT */ /* -> set qua_ener(_MR122) directly */ qua_ener = MIN_QUA_ENER; move16 (); qua_ener_MR122 = MIN_QUA_ENER_MR122; move16 (); } else { /* convert gcode0 from DPF to standard fraction/exponent format */ /* with normalized frac, i.e. 16384 <= frac <= 32767 */ /* Note: exponent correction (exp=exp-14) is done after div_s */ frac_gcode0 = extract_l (Pow2 (14, frac_gcode0)); /* make sure cod_gain_frac < frac_gcode0 for div_s */ if (test (), sub(cod_gain_frac, frac_gcode0) >= 0) { cod_gain_frac = shr (cod_gain_frac, 1); cod_gain_exp = add (cod_gain_exp, 1); } /* predErrFact = gcu / gcode0 = cod_gain_frac/frac_gcode0 * 2^(cod_gain_exp-(exp_gcode0-14)) = div_s (c_g_f, frac_gcode0)*2^-15 * 2^(c_g_e-exp_gcode0+14) = div_s * 2^(cod_gain_exp-exp_gcode0 - 1) */ frac = div_s (cod_gain_frac, frac_gcode0); tmp = sub (sub (cod_gain_exp, exp_gcode0), 1); Log2 (L_deposit_l (frac), &exp, &frac); exp = add (exp, tmp); /* calculate prediction error (log2, Q10) */ qua_ener_MR122 = shr_r (frac, 5); qua_ener_MR122 = add (qua_ener_MR122, shl (exp, 10)); if (test (), sub(qua_ener_MR122, MIN_QUA_ENER_MR122) < 0) { qua_ener = MIN_QUA_ENER; move16 (); qua_ener_MR122 = MIN_QUA_ENER_MR122; move16 (); } else if (test (), sub(qua_ener_MR122, MAX_QUA_ENER_MR122) > 0) { qua_ener = MAX_QUA_ENER; move16 (); qua_ener_MR122 = MAX_QUA_ENER_MR122; move16 (); } else { /* calculate prediction error (20*log10, Q10) */ L_tmp = Mpy_32_16(exp, frac, 24660); /* 24660 Q12 ~= 6.0206 = 20*log10(2) */ qua_ener = round (L_shl (L_tmp, 13)); /* Q12 * Q0 = Q13 -> Q26 -> Q10 */ } } /* update MA predictor memory */ gc_pred_update(pred_st, qua_ener_MR122, qua_ener); } /************************************************************************* * * FUNCTION: MR475_gain_quant() * * PURPOSE: Quantization of pitch and codebook gains for two subframes * (using predicted codebook gain) * *************************************************************************/ Word16 MR475_gain_quant( /* o : index of quantization. */ gc_predState *pred_st, /* i/o: gain predictor state struct */ /* data from subframe 0 (or 2) */ Word16 sf0_exp_gcode0, /* i : predicted CB gain (exponent), Q0 */ Word16 sf0_frac_gcode0, /* i : predicted CB gain (fraction), Q15 */ Word16 sf0_exp_coeff[], /* i : energy coeff. (5), exponent part, Q0 */ Word16 sf0_frac_coeff[], /* i : energy coeff. (5), fraction part, Q15 */ /* (frac_coeff and exp_coeff computed in */ /* calc_filt_energies()) */ Word16 sf0_exp_target_en, /* i : exponent of target energy, Q0 */ Word16 sf0_frac_target_en, /* i : fraction of target energy, Q15 */ /* data from subframe 1 (or 3) */ Word16 sf1_code_nosharp[], /* i : innovative codebook vector (L_SUBFR) */ /* (whithout pitch sharpening) */ Word16 sf1_exp_gcode0, /* i : predicted CB gain (exponent), Q0 */ Word16 sf1_frac_gcode0, /* i : predicted CB gain (fraction), Q15 */ Word16 sf1_exp_coeff[], /* i : energy coeff. (5), exponent part, Q0 */ Word16 sf1_frac_coeff[], /* i : energy coeff. (5), fraction part, Q15 */ /* (frac_coeff and exp_coeff computed in */ /* calc_filt_energies()) */ Word16 sf1_exp_target_en, /* i : exponent of target energy, Q0 */ Word16 sf1_frac_target_en, /* i : fraction of target energy, Q15 */ Word16 gp_limit, /* i : pitch gain limit */ Word16 *sf0_gain_pit, /* o : Pitch gain, Q14 */ Word16 *sf0_gain_cod, /* o : Code gain, Q1 */ Word16 *sf1_gain_pit, /* o : Pitch gain, Q14 */ Word16 *sf1_gain_cod /* o : Code gain, Q1 */ ) { const Word16 *p; Word16 i, index = 0; Word16 tmp; Word16 exp; Word16 sf0_gcode0, sf1_gcode0; Word16 g_pitch, g2_pitch, g_code, g2_code, g_pit_cod; Word16 coeff[10], coeff_lo[10], exp_max[10]; /* 0..4: sf0; 5..9: sf1 */ Word32 L_tmp, dist_min; /*-------------------------------------------------------------------* * predicted codebook gain * * ~~~~~~~~~~~~~~~~~~~~~~~ * * gc0 = 2^exp_gcode0 + 2^frac_gcode0 * * * * gcode0 (Q14) = 2^14*2^frac_gcode0 = gc0 * 2^(14-exp_gcode0) * *-------------------------------------------------------------------*/ sf0_gcode0 = extract_l(Pow2(14, sf0_frac_gcode0)); sf1_gcode0 = extract_l(Pow2(14, sf1_frac_gcode0)); /* * For each subframe, the error energy (sum) to be minimized consists * of five terms, t[0..4]. * * t[0] = gp^2 * <y1 y1> * t[1] = -2*gp * <xn y1> * t[2] = gc^2 * <y2 y2> * t[3] = -2*gc * <xn y2> * t[4] = 2*gp*gc * <y1 y2> * */ /* sf 0 */ /* determine the scaling exponent for g_code: ec = ec0 - 11 */ exp = sub(sf0_exp_gcode0, 11); /* calculate exp_max[i] = s[i]-1 */ exp_max[0] = sub(sf0_exp_coeff[0], 13); move16 (); exp_max[1] = sub(sf0_exp_coeff[1], 14); move16 (); exp_max[2] = add(sf0_exp_coeff[2], add(15, shl(exp, 1))); move16 (); exp_max[3] = add(sf0_exp_coeff[3], exp); move16 (); exp_max[4] = add(sf0_exp_coeff[4], add(1, exp)); move16 (); /* sf 1 */ /* determine the scaling exponent for g_code: ec = ec0 - 11 */ exp = sub(sf1_exp_gcode0, 11); /* calculate exp_max[i] = s[i]-1 */ exp_max[5] = sub(sf1_exp_coeff[0], 13); move16 (); exp_max[6] = sub(sf1_exp_coeff[1], 14); move16 (); exp_max[7] = add(sf1_exp_coeff[2], add(15, shl(exp, 1))); move16 (); exp_max[8] = add(sf1_exp_coeff[3], exp); move16 (); exp_max[9] = add(sf1_exp_coeff[4], add(1, exp)); move16 (); /*-------------------------------------------------------------------* * Gain search equalisation: * * ~~~~~~~~~~~~~~~~~~~~~~~~~ * * The MSE for the two subframes is weighted differently if there * * is a big difference in the corresponding target energies * *-------------------------------------------------------------------*/ /* make the target energy exponents the same by de-normalizing the fraction of the smaller one. This is necessary to be able to compare them */ exp = sf0_exp_target_en - sf1_exp_target_en; test (); if (exp > 0) { sf1_frac_target_en = shr (sf1_frac_target_en, exp); } else { sf0_frac_target_en = shl (sf0_frac_target_en, exp); } /* assume no change of exponents */ exp = 0; move16 (); /* test for target energy difference; set exp to +1 or -1 to scale * up/down coefficients for sf 1 */ tmp = shr_r (sf1_frac_target_en, 1); /* tmp = ceil(0.5*en(sf1)) */ test (); if (sub (tmp, sf0_frac_target_en) > 0) /* tmp > en(sf0)? */ { /* * target_energy(sf1) > 2*target_energy(sf0) * -> scale up MSE(sf0) by 2 by adding 1 to exponents 0..4 */ exp = 1; move16 (); } else { tmp = shr (add (sf0_frac_target_en, 3), 2); /* tmp=ceil(0.25*en(sf0)) */ test(); if (sub (tmp, sf1_frac_target_en) > 0) /* tmp > en(sf1)? */ { /* * target_energy(sf1) < 0.25*target_energy(sf0) * -> scale down MSE(sf0) by 0.5 by subtracting 1 from * coefficients 0..4 */ exp = -1; move16 (); } } for (i = 0; i < 5; i++) { exp_max[i] = add (exp_max[i], exp); move16 (); } /*-------------------------------------------------------------------* * Find maximum exponent: * * ~~~~~~~~~~~~~~~~~~~~~~ * * * * For the sum operation, all terms must have the same scaling; * * that scaling should be low enough to prevent overflow. There- * * fore, the maximum scale is determined and all coefficients are * * re-scaled: * * * * exp = max(exp_max[i]) + 1; * * e = exp_max[i]-exp; e <= 0! * * c[i] = c[i]*2^e * *-------------------------------------------------------------------*/ exp = exp_max[0]; move16 (); for (i = 1; i < 10; i++) { move16(); test(); if (sub(exp_max[i], exp) > 0) { exp = exp_max[i]; move16 (); } } exp = add(exp, 1); /* To avoid overflow */ p = &sf0_frac_coeff[0]; move16 (); for (i = 0; i < 5; i++) { tmp = sub(exp, exp_max[i]); L_tmp = L_deposit_h(*p++); L_tmp = L_shr(L_tmp, tmp); L_Extract(L_tmp, &coeff[i], &coeff_lo[i]); } p = &sf1_frac_coeff[0]; move16 (); for (; i < 10; i++) { tmp = sub(exp, exp_max[i]); L_tmp = L_deposit_h(*p++); L_tmp = L_shr(L_tmp, tmp); L_Extract(L_tmp, &coeff[i], &coeff_lo[i]); } /*-------------------------------------------------------------------* * Codebook search: * * ~~~~~~~~~~~~~~~~ * * * * For each pair (g_pitch, g_fac) in the table calculate the * * terms t[0..4] and sum them up; the result is the mean squared * * error for the quantized gains from the table. The index for the * * minimum MSE is stored and finally used to retrieve the quantized * * gains * *-------------------------------------------------------------------*/ /* start with "infinite" MSE */ dist_min = MAX_32; move32(); p = &table_gain_MR475[0]; move16 (); for (i = 0; i < MR475_VQ_SIZE; i++) { /* subframe 0 (and 2) calculations */ g_pitch = *p++; move16 (); g_code = *p++; move16 (); g_code = mult(g_code, sf0_gcode0); g2_pitch = mult(g_pitch, g_pitch); g2_code = mult(g_code, g_code); g_pit_cod = mult(g_code, g_pitch); L_tmp = Mpy_32_16( coeff[0], coeff_lo[0], g2_pitch); L_tmp = Mac_32_16(L_tmp, coeff[1], coeff_lo[1], g_pitch); L_tmp = Mac_32_16(L_tmp, coeff[2], coeff_lo[2], g2_code); L_tmp = Mac_32_16(L_tmp, coeff[3], coeff_lo[3], g_code); L_tmp = Mac_32_16(L_tmp, coeff[4], coeff_lo[4], g_pit_cod); tmp = sub (g_pitch, gp_limit); /* subframe 1 (and 3) calculations */ g_pitch = *p++; move16 (); g_code = *p++; move16 (); test (); test (); test (); if (tmp <= 0 && sub(g_pitch, gp_limit) <= 0) { g_code = mult(g_code, sf1_gcode0); g2_pitch = mult(g_pitch, g_pitch); g2_code = mult(g_code, g_code); g_pit_cod = mult(g_code, g_pitch); L_tmp = Mac_32_16(L_tmp, coeff[5], coeff_lo[5], g2_pitch); L_tmp = Mac_32_16(L_tmp, coeff[6], coeff_lo[6], g_pitch); L_tmp = Mac_32_16(L_tmp, coeff[7], coeff_lo[7], g2_code); L_tmp = Mac_32_16(L_tmp, coeff[8], coeff_lo[8], g_code); L_tmp = Mac_32_16(L_tmp, coeff[9], coeff_lo[9], g_pit_cod); /* store table index if MSE for this index is lower than the minimum MSE seen so far */ test (); if (L_sub(L_tmp, dist_min) < (Word32) 0) { dist_min = L_tmp; move32 (); index = i; move16 (); } } } /*------------------------------------------------------------------* * read quantized gains and update MA predictor memories * * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ * *------------------------------------------------------------------*/ /* for subframe 0, the pre-calculated gcode0/exp_gcode0 are the same as those calculated from the "real" predictor using quantized gains */ tmp = shl(index, 2); MR475_quant_store_results(pred_st, &table_gain_MR475[tmp], sf0_gcode0, sf0_exp_gcode0, sf0_gain_pit, sf0_gain_cod); /* calculate new predicted gain for subframe 1 (this time using the real, quantized gains) */ gc_pred(pred_st, MR475, sf1_code_nosharp, &sf1_exp_gcode0, &sf1_frac_gcode0, &sf0_exp_gcode0, &sf0_gcode0); /* last two args are dummy */ sf1_gcode0 = extract_l(Pow2(14, sf1_frac_gcode0)); tmp = add (tmp, 2); MR475_quant_store_results(pred_st, &table_gain_MR475[tmp], sf1_gcode0, sf1_exp_gcode0, sf1_gain_pit, sf1_gain_cod); return index; }