FreeCalypso > hg > gsmhr-codec-ref
view vad.c @ 4:a2529fec5442
add README
author | Mychaela Falconia <falcon@freecalypso.org> |
---|---|
date | Sun, 21 Jul 2024 22:40:24 +0000 |
parents | 9008dbc8ca74 |
children |
line wrap: on
line source
/**************************************************************************** * * TITLE: Half-Rate GSM Voice Activity Detector (VAD) Modules * * VERSION: 1.2 * * REFERENCE: Recommendation GSM 06.42 * ***************************************************************************/ /*_________________________________________________________________________ | | | Include Files | |_________________________________________________________________________| */ #include "typedefs.h" #include "mathhalf.h" #include "mathdp31.h" #include "vad.h" /*_________________________________________________________________________ | | | Local Defines | |_________________________________________________________________________| */ /*** Floating point representations of constants pth, plev and margin ***/ #define M_PTH 26250 #define E_PTH 18 #define M_PLEV 17500 #define E_PLEV 20 #define M_MARGIN 27343 #define E_MARGIN 27 /*_________________________________________________________________________ | | | Static Variables | |_________________________________________________________________________| */ static Shortword pswRvad[9], swNormRvad, swPt_sacf, swPt_sav0, swE_thvad, swM_thvad, swAdaptCount, swBurstCount, swHangCount, swOldLagCount, swVeryOldLagCount, swOldLag; static Longword pL_sacf[27], pL_sav0[36], L_lastdm; /**************************************************************************** * * FUNCTION: vad_reset * * VERSION: 1.2 * * PURPOSE: Resets VAD static variables to their initial value. * ***************************************************************************/ void vad_reset(void) { /*_________________________________________________________________________ | | | Automatic Variables | |_________________________________________________________________________| */ int i; /*_________________________________________________________________________ | | | Executable Code | |_________________________________________________________________________| */ pswRvad[0] = 24576; swNormRvad = 7; swPt_sacf = 0; swPt_sav0 = 0; L_lastdm = 0; swE_thvad = 21; swM_thvad = 21875; swAdaptCount = 0; swBurstCount = 0; swHangCount = -1; swOldLagCount = 0; swVeryOldLagCount = 0; swOldLag = 21; for (i = 1; i < 9; i++) pswRvad[i] = 0; for (i = 0; i < 27; i++) pL_sacf[i] = 0; for (i = 0; i < 36; i++) pL_sav0[i] = 0; } /**************************************************************************** * * FUNCTION: vad_algorithm * * VERSION: 1.2 * * PURPOSE: Returns a decision as to whether the current frame being * processed by the speech encoder contains speech or not. * * INPUTS: pL_acf[0..8] autocorrelation of input signal frame * swScaleAcf L_acf scaling factor * pswRc[0..3] speech encoder reflection coefficients * swPtch flag to indicate a periodic signal component * * OUTPUTS: pswVadFlag vad decision * ***************************************************************************/ void vad_algorithm(Longword pL_acf[9], Shortword swScaleAcf, Shortword pswRc[4], Shortword swPtch, Shortword *pswVadFlag) { /*_________________________________________________________________________ | | | Automatic Variables | |_________________________________________________________________________| */ Longword pL_av0[9], pL_av1[9]; Shortword swM_acf0, swE_acf0, pswRav1[9], swNormRav1, swM_pvad, swE_pvad, swStat, swTone, swVvad; /*_________________________________________________________________________ | | | Executable Code | |_________________________________________________________________________| */ energy_computation ( pL_acf, swScaleAcf, pswRvad, swNormRvad, &swM_pvad, &swE_pvad, &swM_acf0, &swE_acf0 ); average_acf ( pL_acf, swScaleAcf, pL_av0, pL_av1 ); predictor_values ( pL_av1, pswRav1, &swNormRav1 ); spectral_comparison ( pswRav1, swNormRav1, pL_av0, &swStat ); tone_detection ( pswRc, &swTone ); threshold_adaptation ( swStat, swPtch, swTone, pswRav1, swNormRav1, swM_pvad, swE_pvad, swM_acf0, swE_acf0, pswRvad, &swNormRvad, &swM_thvad, &swE_thvad ); vad_decision ( swM_pvad, swE_pvad, swM_thvad, swE_thvad, &swVvad ); vad_hangover ( swVvad, pswVadFlag ); } /**************************************************************************** * * FUNCTION: energy_computation * * VERSION: 1.2 * * PURPOSE: Computes the input and residual energies of the adaptive * filter in a floating point representation. * * INPUTS: pL_acf[0..8] autocorrelation of input signal frame * swScaleAcf L_acf scaling factor * pswRvad[0..8] autocorrelated adaptive filter coefficients * swNormRvad rvad scaling factor * * OUTPUTS: pswM_pvad mantissa of filtered signal energy * pswE_pvad exponent of filtered signal energy * pswM_acf0 mantissa of signal frame energy * pswE_acf0 exponent of signal frame energy * ***************************************************************************/ void energy_computation(Longword pL_acf[], Shortword swScaleAcf, Shortword pswRvad[], Shortword swNormRvad, Shortword *pswM_pvad, Shortword *pswE_pvad, Shortword *pswM_acf0, Shortword *pswE_acf0) { /*_________________________________________________________________________ | | | Automatic Variables | |_________________________________________________________________________| */ Longword L_temp; Shortword pswSacf[9], swNormAcf, swNormProd, swShift; int i; /*_________________________________________________________________________ | | | Executable Code | |_________________________________________________________________________| */ /*** Test if acf[0] is zero ***/ if (pL_acf[0] == 0) { *pswE_pvad = -0x8000; *pswM_pvad = 0; *pswE_acf0 = -0x8000; *pswM_acf0 = 0; return; } /*** Re-normalisation of L_acf[0..8] ***/ swNormAcf = norm_l(pL_acf[0]); swShift = sub(swNormAcf, 3); for (i = 0; i <= 8; i++) pswSacf[i] = extract_h(L_shl(pL_acf[i], swShift)); /*** Computation of e_acf0 and m_acf0 ***/ *pswE_acf0 = add(32, shl(swScaleAcf, 1)); *pswE_acf0 = sub(*pswE_acf0, swNormAcf); *pswM_acf0 = shl(pswSacf[0], 3); /*** Computation of e_pvad and m_pvad ***/ *pswE_pvad = add(*pswE_acf0, 14); *pswE_pvad = sub(*pswE_pvad, swNormRvad); L_temp = 0; for (i = 1; i <= 8; i++) L_temp = L_mac(L_temp, pswSacf[i], pswRvad[i]); L_temp = L_add(L_temp, L_shr(L_mult(pswSacf[0], pswRvad[0]), 1)); if (L_temp <= 0) L_temp = 1; swNormProd = norm_l(L_temp); *pswE_pvad = sub(*pswE_pvad, swNormProd); *pswM_pvad = extract_h(L_shl(L_temp, swNormProd)); } /**************************************************************************** * * FUNCTION: average_acf * * VERSION: 1.2 * * PURPOSE: Computes the arrays L_av0 [0..8] and L_av1 [0..8]. * * INPUTS: pL_acf[0..8] autocorrelation of input signal frame * swScaleAcf L_acf scaling factor * * OUTPUTS: pL_av0[0..8] ACF averaged over last four frames * pL_av1[0..8] ACF averaged over previous four frames * ***************************************************************************/ void average_acf(Longword pL_acf[], Shortword swScaleAcf, Longword pL_av0[], Longword pL_av1[]) { /*_________________________________________________________________________ | | | Automatic Variables | |_________________________________________________________________________| */ Longword L_temp; Shortword swScale; int i; /*_________________________________________________________________________ | | | Executable Code | |_________________________________________________________________________| */ /*** computation of the scaleing factor ***/ swScale = sub(10, shl(swScaleAcf, 1)); /*** Computation of the arrays L_av0 and L_av1 ***/ for (i = 0; i <= 8; i++) { L_temp = L_shr(pL_acf[i], swScale); pL_av0[i] = L_add(pL_sacf[i], L_temp); pL_av0[i] = L_add(pL_sacf[i + 9], pL_av0[i]); pL_av0[i] = L_add(pL_sacf[i + 18], pL_av0[i]); pL_sacf[swPt_sacf + i] = L_temp; pL_av1[i] = pL_sav0[swPt_sav0 + i]; pL_sav0[swPt_sav0 + i] = pL_av0[i]; } /*** Update the array pointers ***/ if (swPt_sacf == 18) swPt_sacf = 0; else swPt_sacf = add(swPt_sacf, 9); if (swPt_sav0 == 27) swPt_sav0 = 0; else swPt_sav0 = add(swPt_sav0, 9); } /**************************************************************************** * * FUNCTION: predictor_values * * VERSION: 1.2 * * PURPOSE: Computes the array rav [0..8] needed for the spectral * comparison and the threshold adaptation. * * INPUTS: pL_av1 [0..8] ACF averaged over previous four frames * * OUTPUTS: pswRav1 [0..8] ACF obtained from L_av1 * pswNormRav1 r_av1 scaling factor * ***************************************************************************/ void predictor_values(Longword pL_av1[], Shortword pswRav1[], Shortword *pswNormRav1) { /*_________________________________________________________________________ | | | Automatic Variables | |_________________________________________________________________________| */ Shortword pswVpar[8], pswAav1[9]; /*_________________________________________________________________________ | | | Executable Code | |_________________________________________________________________________| */ schur_recursion(pL_av1, pswVpar); step_up(8, pswVpar, pswAav1); compute_rav1(pswAav1, pswRav1, pswNormRav1); } /**************************************************************************** * * FUNCTION: schur_recursion * * VERSION: 1.2 * * PURPOSE: Uses the Schur recursion to compute adaptive filter * reflection coefficients from an autorrelation function. * * INPUTS: pL_av1[0..8] autocorrelation function * * OUTPUTS: pswVpar[0..7] reflection coefficients * ***************************************************************************/ void schur_recursion(Longword pL_av1[], Shortword pswVpar[]) { /*_________________________________________________________________________ | | | Automatic Variables | |_________________________________________________________________________| */ Shortword pswAcf[9], pswPp[9], pswKk[9], swTemp; int i, k, m, n; /*_________________________________________________________________________ | | | Executable Code | |_________________________________________________________________________| */ /*** Schur recursion with 16-bit arithmetic ***/ if (pL_av1[0] == 0) { for (i = 0; i < 8; i++) pswVpar[i] = 0; return; } swTemp = norm_l(pL_av1[0]); for (k = 0; k <= 8; k++) pswAcf[k] = extract_h(L_shl(pL_av1[k], swTemp)); /*** Initialise array pp[..] and kk[..] for the recursion: ***/ for (i = 1; i <= 7; i++) pswKk[9 - i] = pswAcf[i]; for (i = 0; i <= 8; i++) pswPp[i] = pswAcf[i]; /*** Compute Parcor coefficients: ***/ for (n = 0; n < 8; n++) { if (pswPp[0] < abs_s(pswPp[1])) { for (i = n; i < 8; i++) pswVpar[i] = 0; return; } pswVpar[n] = divide_s(abs_s(pswPp[1]), pswPp[0]); if (pswPp[1] > 0) pswVpar[n] = negate(pswVpar[n]); if (n == 7) return; /*** Schur recursion: ***/ pswPp[0] = add(pswPp[0], mult_r(pswPp[1], pswVpar[n])); for (m = 1; m <= (7 - n); m++) { pswPp[m] = add(pswPp[1 + m], mult_r(pswKk[9 - m], pswVpar[n])); pswKk[9 - m] = add(pswKk[9 - m], mult_r(pswPp[1 + m], pswVpar[n])); } } } /**************************************************************************** * * FUNCTION: step_up * * VERSION: 1.2 * * PURPOSE: Computes the transversal filter coefficients from the * reflection coefficients. * * INPUTS: swNp filter order (2..8) * pswVpar[0..np-1] reflection coefficients * * OUTPUTS: pswAav1[0..np] transversal filter coefficients * ***************************************************************************/ void step_up(Shortword swNp, Shortword pswVpar[], Shortword pswAav1[]) { /*_________________________________________________________________________ | | | Automatic Variables | |_________________________________________________________________________| */ Longword pL_coef[9], pL_work[9]; Shortword swTemp; int i, m; /*_________________________________________________________________________ | | | Executable Code | |_________________________________________________________________________| */ /*** Initialisation of the step-up recursion ***/ pL_coef[0] = L_shl(0x4000L, 15); pL_coef[1] = L_shl(L_deposit_l(pswVpar[0]), 14); /*** Loop on the LPC analysis order: ***/ for (m = 2; m <= swNp; m++) { for (i = 1; i < m; i++) { swTemp = extract_h(pL_coef[m - i]); pL_work[i] = L_mac(pL_coef[i], pswVpar[m - 1], swTemp); } for (i = 1; i < m; i++) pL_coef[i] = pL_work[i]; pL_coef[m] = L_shl(L_deposit_l(pswVpar[m - 1]), 14); } /*** Keep the aav1[0..swNp] in 15 bits for the following subclause ***/ for (i = 0; i <= swNp; i++) pswAav1[i] = extract_h(L_shr(pL_coef[i], 3)); } /**************************************************************************** * * FUNCTION: compute_rav1 * * VERSION: 1.2 * * PURPOSE: Computes the autocorrelation function of the adaptive * filter coefficients. * * INPUTS: pswAav1[0..8] adaptive filter coefficients * * OUTPUTS: pswRav1[0..8] ACF of aav1 * pswNormRav1 r_av1 scaling factor * ***************************************************************************/ void compute_rav1(Shortword pswAav1[], Shortword pswRav1[], Shortword *pswNormRav1) { /*_________________________________________________________________________ | | | Automatic Variables | |_________________________________________________________________________| */ Longword pL_work[9]; int i, k; /*_________________________________________________________________________ | | | Executable Code | |_________________________________________________________________________| */ /*** Computation of the rav1[0..8] ***/ for (i = 0; i <= 8; i++) { pL_work[i] = 0; for (k = 0; k <= 8 - i; k++) pL_work[i] = L_mac(pL_work[i], pswAav1[k], pswAav1[k + i]); } if (pL_work[0] == 0) *pswNormRav1 = 0; else *pswNormRav1 = norm_l(pL_work[0]); for (i = 0; i <= 8; i++) pswRav1[i] = extract_h(L_shl(pL_work[i], *pswNormRav1)); } /**************************************************************************** * * FUNCTION: spectral_comparison * * VERSION: 1.2 * * PURPOSE: Computes the stat flag needed for the threshold * adaptation decision. * * INPUTS: pswRav1[0..8] ACF obtained from L_av1 * swNormRav1 rav1 scaling factor * pL_av0[0..8] ACF averaged over last four frames * * OUTPUTS: pswStat flag to indicate spectral stationarity * ***************************************************************************/ void spectral_comparison(Shortword pswRav1[], Shortword swNormRav1, Longword pL_av0[], Shortword *pswStat) { /*_________________________________________________________________________ | | | Automatic Variables | |_________________________________________________________________________| */ Longword L_dm, L_sump, L_temp; Shortword pswSav0[9], swShift, swDivShift, swTemp; int i; /*_________________________________________________________________________ | | | Executable Code | |_________________________________________________________________________| */ /*** Re-normalise L_av0[0..8] ***/ if (pL_av0[0] == 0) { for (i = 0; i <= 8; i++) pswSav0[i] = 4095; } else { swShift = sub(norm_l(pL_av0[0]), 3); for (i = 0; i <= 8; i++) pswSav0[i] = extract_h(L_shl(pL_av0[i], swShift)); } /*** compute partial sum of dm ***/ L_sump = 0; for (i = 1; i <= 8; i++) L_sump = L_mac(L_sump, pswRav1[i], pswSav0[i]); /*** compute the division of the partial sum by sav0[0] ***/ if (L_sump < 0) L_temp = L_negate(L_sump); else L_temp = L_sump; if (L_temp == 0) { L_dm = 0; swShift = 0; } else { pswSav0[0] = shl(pswSav0[0], 3); swShift = norm_l(L_temp); swTemp = extract_h(L_shl(L_temp, swShift)); if (pswSav0[0] >= swTemp) { swDivShift = 0; swTemp = divide_s(swTemp, pswSav0[0]); } else { swDivShift = 1; swTemp = sub(swTemp, pswSav0[0]); swTemp = divide_s(swTemp, pswSav0[0]); } if (swDivShift == 1) L_dm = 0x8000L; else L_dm = 0; L_dm = L_shl((L_add(L_dm, L_deposit_l(swTemp))), 1); if (L_sump < 0) L_dm = L_sub(0L, L_dm); } /*** Re-normalisation and final computation of L_dm ***/ L_dm = L_shl(L_dm, 14); L_dm = L_shr(L_dm, swShift); L_dm = L_add(L_dm, L_shl(L_deposit_l(pswRav1[0]), 11)); L_dm = L_shr(L_dm, swNormRav1); /*** Compute the difference and save L_dm ***/ L_temp = L_sub(L_dm, L_lastdm); L_lastdm = L_dm; if (L_temp < 0L) L_temp = L_negate(L_temp); /*** Evaluation of the state flag ***/ L_temp = L_sub(L_temp, 4456L); if (L_temp < 0) *pswStat = 1; else *pswStat = 0; } /**************************************************************************** * * FUNCTION: tone_detection * * VERSION: 1.2 * * PURPOSE: Computes the tone flag needed for the threshold * adaptation decision. * * INPUTS: pswRc[0..3] reflection coefficients calculated in the * speech encoder short term predictor * * OUTPUTS: pswTone flag to indicate a periodic signal component * ***************************************************************************/ void tone_detection(Shortword pswRc[4], Shortword *pswTone) { /*_________________________________________________________________________ | | | Automatic Variables | |_________________________________________________________________________| */ Longword L_num, L_den, L_temp; Shortword swTemp, swPredErr, pswA[3]; int i; /*_________________________________________________________________________ | | | Executable Code | |_________________________________________________________________________| */ *pswTone = 0; /*** Calculate filter coefficients ***/ step_up(2, pswRc, pswA); /*** Calculate ( a[1] * a[1] ) ***/ swTemp = shl(pswA[1], 3); L_den = L_mult(swTemp, swTemp); /*** Calculate ( 4*a[2] - a[1]*a[1] ) ***/ L_temp = L_shl(L_deposit_h(pswA[2]), 3); L_num = L_sub(L_temp, L_den); /*** Check if pole frequency is less than 385 Hz ***/ if (L_num <= 0) return; if (pswA[1] < 0) { swTemp = extract_h(L_den); L_temp = L_msu(L_num, swTemp, 3189); if (L_temp < 0) return; } /*** Calculate normalised prediction error ***/ swPredErr = 0x7fff; for (i = 0; i < 4; i++) { swTemp = mult(pswRc[i], pswRc[i]); swTemp = sub(32767, swTemp); swPredErr = mult(swPredErr, swTemp); } /*** Test if prediction error is smaller than threshold ***/ swTemp = sub(swPredErr, 1464); if (swTemp < 0) *pswTone = 1; } /**************************************************************************** * * FUNCTION: threshold_adaptation * * VERSION: 1.2 * * PURPOSE: Evaluates the secondary VAD decision. If speech is not * present then the noise model rvad and adaptive threshold * thvad are updated. * * INPUTS: swStat flag to indicate spectral stationarity * swPtch flag to indicate a periodic signal component * swTone flag to indicate a tone signal component * pswRav1[0..8] ACF obtained from l_av1 * swNormRav1 r_av1 scaling factor * swM_pvad mantissa of filtered signal energy * swE_pvad exponent of filtered signal energy * swM_acf0 mantissa of signal frame energy * swE_acf0 exponent of signal frame energy * * OUTPUTS: pswRvad[0..8] autocorrelated adaptive filter coefficients * pswNormRvad rvad scaling factor * pswM_thvad mantissa of decision threshold * pswE_thvad exponent of decision threshold * ***************************************************************************/ void threshold_adaptation(Shortword swStat, Shortword swPtch, Shortword swTone, Shortword pswRav1[], Shortword swNormRav1, Shortword swM_pvad, Shortword swE_pvad, Shortword swM_acf0, Shortword swE_acf0, Shortword pswRvad[], Shortword *pswNormRvad, Shortword *pswM_thvad, Shortword *pswE_thvad) { /*_________________________________________________________________________ | | | Automatic Variables | |_________________________________________________________________________| */ Longword L_temp; Shortword swTemp, swComp, swComp2, swM_temp, swE_temp; int i; /*_________________________________________________________________________ | | | Executable Code | |_________________________________________________________________________| */ swComp = 0; /*** Test if acf0 < pth; if yes set thvad to plev ***/ if (swE_acf0 < E_PTH) swComp = 1; if ((swE_acf0 == E_PTH) && (swM_acf0 < M_PTH)) swComp = 1; if (swComp == 1) { *pswE_thvad = E_PLEV; *pswM_thvad = M_PLEV; return; } /*** Test if an adaption is required ***/ if (swPtch == 1) swComp = 1; if (swStat == 0) swComp = 1; if (swTone == 1) swComp = 1; if (swComp == 1) { swAdaptCount = 0; return; } /*** Increment adaptcount ***/ swAdaptCount = add(swAdaptCount, 1); if (swAdaptCount <= 8) return; /*** computation of thvad-(thvad/dec) ***/ *pswM_thvad = sub(*pswM_thvad, shr(*pswM_thvad, 5)); if (*pswM_thvad < 0x4000) { *pswM_thvad = shl(*pswM_thvad, 1); *pswE_thvad = sub(*pswE_thvad, 1); } /*** computation of pvad*fac ***/ L_temp = L_mult(swM_pvad, 20889); L_temp = L_shr(L_temp, 15); swE_temp = add(swE_pvad, 1); if (L_temp > 0x7fffL) { L_temp = L_shr(L_temp, 1); swE_temp = add(swE_temp, 1); } swM_temp = extract_l(L_temp); /*** test if thvad < pavd*fac ***/ if (*pswE_thvad < swE_temp) swComp = 1; if ((*pswE_thvad == swE_temp) && (*pswM_thvad < swM_temp)) swComp = 1; /*** compute minimum(thvad+(thvad/inc), pvad*fac) when comp = 1 ***/ if (swComp == 1) { /*** compute thvad + (thvad/inc) ***/ L_temp = L_add(L_deposit_l(*pswM_thvad),L_deposit_l(shr(*pswM_thvad, 4))); if (L_temp > 0x7fffL) { *pswM_thvad = extract_l(L_shr(L_temp, 1)); *pswE_thvad = add(*pswE_thvad, 1); } else *pswM_thvad = extract_l(L_temp); swComp2 = 0; if (swE_temp < *pswE_thvad) swComp2 = 1; if ((swE_temp == *pswE_thvad) && (swM_temp < *pswM_thvad)) swComp2 = 1; if (swComp2 == 1) { *pswE_thvad = swE_temp; *pswM_thvad = swM_temp; } } /*** compute pvad + margin ***/ if (swE_pvad == E_MARGIN) { L_temp = L_add(L_deposit_l(swM_pvad), L_deposit_l(M_MARGIN)); swM_temp = extract_l(L_shr(L_temp, 1)); swE_temp = add(swE_pvad, 1); } else { if (swE_pvad > E_MARGIN) { swTemp = sub(swE_pvad, E_MARGIN); swTemp = shr(M_MARGIN, swTemp); L_temp = L_add(L_deposit_l(swM_pvad), L_deposit_l(swTemp)); if (L_temp > 0x7fffL) { swE_temp = add(swE_pvad, 1); swM_temp = extract_l(L_shr(L_temp, 1)); } else { swE_temp = swE_pvad; swM_temp = extract_l(L_temp); } } else { swTemp = sub(E_MARGIN, swE_pvad); swTemp = shr(swM_pvad, swTemp); L_temp = L_add(L_deposit_l(M_MARGIN), L_deposit_l(swTemp)); if (L_temp > 0x7fffL) { swE_temp = add(E_MARGIN, 1); swM_temp = extract_l(L_shr(L_temp, 1)); } else { swE_temp = E_MARGIN; swM_temp = extract_l(L_temp); } } } /*** Test if thvad > pvad + margin ***/ swComp = 0; if (*pswE_thvad > swE_temp) swComp = 1; if ((*pswE_thvad == swE_temp) && (*pswM_thvad > swM_temp)) swComp = 1; if (swComp == 1) { *pswE_thvad = swE_temp; *pswM_thvad = swM_temp; } /*** Normalise and retain rvad[0..8] in memory ***/ *pswNormRvad = swNormRav1; for (i = 0; i <= 8; i++) pswRvad[i] = pswRav1[i]; /*** Set adaptcount to adp + 1 ***/ swAdaptCount = 9; } /**************************************************************************** * * FUNCTION: vad_decision * * VERSION: 1.2 * * PURPOSE: Computes the VAD decision based on the comparison of the * floating point representations of pvad and thvad. * * INPUTS: swM_pvad mantissa of filtered signal energy * swE_pvad exponent of filtered signal energy * swM_thvad mantissa of decision threshold * swE_thvad exponent of decision threshold * * OUTPUTS: pswVvad vad decision before hangover is added * ***************************************************************************/ void vad_decision(Shortword swM_pvad, Shortword swE_pvad, Shortword swM_thvad, Shortword swE_thvad, Shortword *pswVvad) { /*_________________________________________________________________________ | | | Executable Code | |_________________________________________________________________________| */ *pswVvad = 0; if (swE_pvad > swE_thvad) *pswVvad = 1; if ((swE_pvad == swE_thvad) && (swM_pvad > swM_thvad)) *pswVvad = 1; } /**************************************************************************** * * FUNCTION: vad_hangover * * VERSION: 1.2 * * PURPOSE: Computes the final VAD decision for the current frame * being processed. * * INPUTS: swVvad vad decision before hangover is added * * OUTPUTS: pswVadFlag vad decision after hangover is added * ***************************************************************************/ void vad_hangover(Shortword swVvad, Shortword *pswVadFlag) { /*_________________________________________________________________________ | | | Executable Code | |_________________________________________________________________________| */ if (swVvad == 1) swBurstCount = add(swBurstCount, 1); else swBurstCount = 0; if (swBurstCount >= 3) { swHangCount = 5; swBurstCount = 3; } *pswVadFlag = swVvad; if (swHangCount >= 0) { *pswVadFlag = 1; swHangCount = sub(swHangCount, 1); } } /**************************************************************************** * * FUNCTION: periodicity_update * * VERSION: 1.2 * * PURPOSE: Computes the ptch flag needed for the threshold * adaptation decision for the next frame. * * INPUTS: pswLags[0..3] speech encoder long term predictor lags * * OUTPUTS: pswPtch Boolean voiced / unvoiced decision * ***************************************************************************/ void periodicity_update(Shortword pswLags[4], Shortword *pswPtch) { /*_________________________________________________________________________ | | | Automatic Variables | |_________________________________________________________________________| */ Shortword swMinLag, swMaxLag, swSmallLag, swLagCount, swTemp; int i, j; /*_________________________________________________________________________ | | | Executable Code | |_________________________________________________________________________| */ /*** Run loop for No. of sub-segments in the frame ***/ swLagCount = 0; for (i = 0; i <= 3; i++) { /*** Search the maximum and minimum of consecutive lags ***/ if (swOldLag > pswLags[i]) { swMinLag = pswLags[i]; swMaxLag = swOldLag; } else { swMinLag = swOldLag; swMaxLag = pswLags[i]; } /*** Compute smallag (modulo operation not defined) ***/ swSmallLag = swMaxLag; for (j = 0; j <= 2; j++) { if (swSmallLag >= swMinLag) swSmallLag = sub(swSmallLag, swMinLag); } /*** Minimum of smallag and minlag - smallag ***/ swTemp = sub(swMinLag, swSmallLag); if (swTemp < swSmallLag) swSmallLag = swTemp; if (swSmallLag < 2) swLagCount = add(swLagCount, 1); /*** Save the current LTP lag ***/ swOldLag = pswLags[i]; } /*** Update the veryoldlagcount and oldlagcount ***/ swVeryOldLagCount = swOldLagCount; swOldLagCount = swLagCount; /*** Make ptch decision ready for next frame ***/ swTemp = add(swOldLagCount, swVeryOldLagCount); if (swTemp >= 7) *pswPtch = 1; else *pswPtch = 0; }