|
|
@ -272,8 +272,7 @@ static void g729d_get_new_exc( |
|
|
|
|
|
|
|
|
|
|
|
ff_celp_convolve_circ(fc_new, fc_cur, phase_filter[dstate], subframe_size); |
|
|
|
ff_celp_convolve_circ(fc_new, fc_cur, phase_filter[dstate], subframe_size); |
|
|
|
|
|
|
|
|
|
|
|
for(i=0; i<subframe_size; i++) |
|
|
|
for (i = 0; i < subframe_size; i++) { |
|
|
|
{ |
|
|
|
|
|
|
|
out[i] = in[i]; |
|
|
|
out[i] = in[i]; |
|
|
|
out[i] -= (gain_code * fc_cur[i] + 0x2000) >> 14; |
|
|
|
out[i] -= (gain_code * fc_cur[i] + 0x2000) >> 14; |
|
|
|
out[i] += (gain_code * fc_new[i] + 0x2000) >> 14; |
|
|
|
out[i] += (gain_code * fc_new[i] + 0x2000) >> 14; |
|
|
@ -289,10 +288,10 @@ static void g729d_get_new_exc( |
|
|
|
*/ |
|
|
|
*/ |
|
|
|
static int g729d_onset_decision(int past_onset, const int16_t* past_gain_code) |
|
|
|
static int g729d_onset_decision(int past_onset, const int16_t* past_gain_code) |
|
|
|
{ |
|
|
|
{ |
|
|
|
if((past_gain_code[0] >> 1) > past_gain_code[1]) |
|
|
|
if ((past_gain_code[0] >> 1) > past_gain_code[1]) |
|
|
|
return 2; |
|
|
|
return 2; |
|
|
|
else |
|
|
|
|
|
|
|
return FFMAX(past_onset-1, 0); |
|
|
|
return FFMAX(past_onset-1, 0); |
|
|
|
} |
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
/**
|
|
|
@ -307,24 +306,25 @@ static int16_t g729d_voice_decision(int onset, int prev_voice_decision, const in |
|
|
|
{ |
|
|
|
{ |
|
|
|
int i, low_gain_pitch_cnt, voice_decision; |
|
|
|
int i, low_gain_pitch_cnt, voice_decision; |
|
|
|
|
|
|
|
|
|
|
|
if(past_gain_pitch[0] >= 14745) // 0.9
|
|
|
|
if (past_gain_pitch[0] >= 14745) { // 0.9
|
|
|
|
voice_decision = DECISION_VOICE; |
|
|
|
voice_decision = DECISION_VOICE; |
|
|
|
else if (past_gain_pitch[0] <= 9830) // 0.6
|
|
|
|
} else if (past_gain_pitch[0] <= 9830) { // 0.6
|
|
|
|
voice_decision = DECISION_NOISE; |
|
|
|
voice_decision = DECISION_NOISE; |
|
|
|
else |
|
|
|
} else { |
|
|
|
voice_decision = DECISION_INTERMEDIATE; |
|
|
|
voice_decision = DECISION_INTERMEDIATE; |
|
|
|
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
for(i=0, low_gain_pitch_cnt=0; i<6; i++) |
|
|
|
for (i = 0, low_gain_pitch_cnt = 0; i < 6; i++) |
|
|
|
if(past_gain_pitch[i] < 9830) |
|
|
|
if (past_gain_pitch[i] < 9830) |
|
|
|
low_gain_pitch_cnt++; |
|
|
|
low_gain_pitch_cnt++; |
|
|
|
|
|
|
|
|
|
|
|
if(low_gain_pitch_cnt > 2 && !onset) |
|
|
|
if (low_gain_pitch_cnt > 2 && !onset) |
|
|
|
voice_decision = DECISION_NOISE; |
|
|
|
voice_decision = DECISION_NOISE; |
|
|
|
|
|
|
|
|
|
|
|
if(!onset && voice_decision > prev_voice_decision + 1) |
|
|
|
if (!onset && voice_decision > prev_voice_decision + 1) |
|
|
|
voice_decision--; |
|
|
|
voice_decision--; |
|
|
|
|
|
|
|
|
|
|
|
if(onset && voice_decision < DECISION_VOICE) |
|
|
|
if (onset && voice_decision < DECISION_VOICE) |
|
|
|
voice_decision++; |
|
|
|
voice_decision++; |
|
|
|
|
|
|
|
|
|
|
|
return voice_decision; |
|
|
|
return voice_decision; |
|
|
@ -361,30 +361,30 @@ static av_cold int decoder_init(AVCodecContext * avctx) |
|
|
|
return AVERROR(ENOMEM); |
|
|
|
return AVERROR(ENOMEM); |
|
|
|
|
|
|
|
|
|
|
|
for (c = 0; c < avctx->channels; c++) { |
|
|
|
for (c = 0; c < avctx->channels; c++) { |
|
|
|
ctx->gain_coeff = 16384; // 1.0 in (1.14)
|
|
|
|
ctx->gain_coeff = 16384; // 1.0 in (1.14)
|
|
|
|
|
|
|
|
|
|
|
|
for (k = 0; k < MA_NP + 1; k++) { |
|
|
|
for (k = 0; k < MA_NP + 1; k++) { |
|
|
|
ctx->past_quantizer_outputs[k] = ctx->past_quantizer_output_buf[k]; |
|
|
|
ctx->past_quantizer_outputs[k] = ctx->past_quantizer_output_buf[k]; |
|
|
|
for (i = 1; i < 11; i++) |
|
|
|
for (i = 1; i < 11; i++) |
|
|
|
ctx->past_quantizer_outputs[k][i - 1] = (18717 * i) >> 3; |
|
|
|
ctx->past_quantizer_outputs[k][i - 1] = (18717 * i) >> 3; |
|
|
|
} |
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
ctx->lsp[0] = ctx->lsp_buf[0]; |
|
|
|
ctx->lsp[0] = ctx->lsp_buf[0]; |
|
|
|
ctx->lsp[1] = ctx->lsp_buf[1]; |
|
|
|
ctx->lsp[1] = ctx->lsp_buf[1]; |
|
|
|
memcpy(ctx->lsp[0], lsp_init, 10 * sizeof(int16_t)); |
|
|
|
memcpy(ctx->lsp[0], lsp_init, 10 * sizeof(int16_t)); |
|
|
|
|
|
|
|
|
|
|
|
ctx->exc = &ctx->exc_base[PITCH_DELAY_MAX+INTERPOL_LEN]; |
|
|
|
ctx->exc = &ctx->exc_base[PITCH_DELAY_MAX+INTERPOL_LEN]; |
|
|
|
|
|
|
|
|
|
|
|
ctx->pitch_delay_int_prev = PITCH_DELAY_MIN; |
|
|
|
ctx->pitch_delay_int_prev = PITCH_DELAY_MIN; |
|
|
|
|
|
|
|
|
|
|
|
/* random seed initialization */ |
|
|
|
/* random seed initialization */ |
|
|
|
ctx->rand_value = 21845; |
|
|
|
ctx->rand_value = 21845; |
|
|
|
|
|
|
|
|
|
|
|
/* quantized prediction error */ |
|
|
|
/* quantized prediction error */ |
|
|
|
for(i=0; i<4; i++) |
|
|
|
for (i = 0; i < 4; i++) |
|
|
|
ctx->quant_energy[i] = -14336; // -14 in (5.10)
|
|
|
|
ctx->quant_energy[i] = -14336; // -14 in (5.10)
|
|
|
|
|
|
|
|
|
|
|
|
ctx++; |
|
|
|
ctx++; |
|
|
|
} |
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
ff_audiodsp_init(&s->adsp); |
|
|
|
ff_audiodsp_init(&s->adsp); |
|
|
@ -441,286 +441,289 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame_ptr, |
|
|
|
} |
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
for (c = 0; c < avctx->channels; c++) { |
|
|
|
for (c = 0; c < avctx->channels; c++) { |
|
|
|
int frame_erasure = 0; ///< frame erasure detected during decoding
|
|
|
|
int frame_erasure = 0; ///< frame erasure detected during decoding
|
|
|
|
int bad_pitch = 0; ///< parity check failed
|
|
|
|
int bad_pitch = 0; ///< parity check failed
|
|
|
|
int is_periodic = 0; ///< whether one of the subframes is declared as periodic or not
|
|
|
|
int is_periodic = 0; ///< whether one of the subframes is declared as periodic or not
|
|
|
|
out_frame = (int16_t*)frame->data[c]; |
|
|
|
out_frame = (int16_t*)frame->data[c]; |
|
|
|
|
|
|
|
|
|
|
|
for (i=0; i < buf_size; i++) |
|
|
|
|
|
|
|
frame_erasure |= buf[i]; |
|
|
|
|
|
|
|
frame_erasure = !frame_erasure; |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
init_get_bits(&gb, buf, 8*buf_size); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ma_predictor = get_bits(&gb, 1); |
|
|
|
|
|
|
|
quantizer_1st = get_bits(&gb, VQ_1ST_BITS); |
|
|
|
|
|
|
|
quantizer_2nd_lo = get_bits(&gb, VQ_2ND_BITS); |
|
|
|
|
|
|
|
quantizer_2nd_hi = get_bits(&gb, VQ_2ND_BITS); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if(frame_erasure) |
|
|
|
|
|
|
|
lsf_restore_from_previous(ctx->lsfq, ctx->past_quantizer_outputs, |
|
|
|
|
|
|
|
ctx->ma_predictor_prev); |
|
|
|
|
|
|
|
else { |
|
|
|
|
|
|
|
lsf_decode(ctx->lsfq, ctx->past_quantizer_outputs, |
|
|
|
|
|
|
|
ma_predictor, |
|
|
|
|
|
|
|
quantizer_1st, quantizer_2nd_lo, quantizer_2nd_hi); |
|
|
|
|
|
|
|
ctx->ma_predictor_prev = ma_predictor; |
|
|
|
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
tmp = ctx->past_quantizer_outputs[MA_NP]; |
|
|
|
for (i = 0; i < buf_size; i++) |
|
|
|
memmove(ctx->past_quantizer_outputs + 1, ctx->past_quantizer_outputs, |
|
|
|
frame_erasure |= buf[i]; |
|
|
|
MA_NP * sizeof(int16_t*)); |
|
|
|
frame_erasure = !frame_erasure; |
|
|
|
ctx->past_quantizer_outputs[0] = tmp; |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ff_acelp_lsf2lsp(ctx->lsp[1], ctx->lsfq, 10); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ff_acelp_lp_decode(&lp[0][0], &lp[1][0], ctx->lsp[1], ctx->lsp[0], 10); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
FFSWAP(int16_t*, ctx->lsp[1], ctx->lsp[0]); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < 2; i++) { |
|
|
|
|
|
|
|
int gain_corr_factor; |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
uint8_t ac_index; ///< adaptive codebook index
|
|
|
|
|
|
|
|
uint8_t pulses_signs; ///< fixed-codebook vector pulse signs
|
|
|
|
|
|
|
|
int fc_indexes; ///< fixed-codebook indexes
|
|
|
|
|
|
|
|
uint8_t gc_1st_index; ///< gain codebook (first stage) index
|
|
|
|
|
|
|
|
uint8_t gc_2nd_index; ///< gain codebook (second stage) index
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ac_index = get_bits(&gb, format->ac_index_bits[i]); |
|
|
|
|
|
|
|
if(!i && format->parity_bit) |
|
|
|
|
|
|
|
bad_pitch = av_parity(ac_index >> 2) == get_bits1(&gb); |
|
|
|
|
|
|
|
fc_indexes = get_bits(&gb, format->fc_indexes_bits); |
|
|
|
|
|
|
|
pulses_signs = get_bits(&gb, format->fc_signs_bits); |
|
|
|
|
|
|
|
gc_1st_index = get_bits(&gb, format->gc_1st_index_bits); |
|
|
|
|
|
|
|
gc_2nd_index = get_bits(&gb, format->gc_2nd_index_bits); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if (frame_erasure) |
|
|
|
|
|
|
|
pitch_delay_3x = 3 * ctx->pitch_delay_int_prev; |
|
|
|
|
|
|
|
else if(!i) { |
|
|
|
|
|
|
|
if (bad_pitch) |
|
|
|
|
|
|
|
pitch_delay_3x = 3 * ctx->pitch_delay_int_prev; |
|
|
|
|
|
|
|
else |
|
|
|
|
|
|
|
pitch_delay_3x = ff_acelp_decode_8bit_to_1st_delay3(ac_index); |
|
|
|
|
|
|
|
} else { |
|
|
|
|
|
|
|
int pitch_delay_min = av_clip(ctx->pitch_delay_int_prev - 5, |
|
|
|
|
|
|
|
PITCH_DELAY_MIN, PITCH_DELAY_MAX - 9); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if(packet_type == FORMAT_G729D_6K4) |
|
|
|
init_get_bits(&gb, buf, 8*buf_size); |
|
|
|
pitch_delay_3x = ff_acelp_decode_4bit_to_2nd_delay3(ac_index, pitch_delay_min); |
|
|
|
|
|
|
|
else |
|
|
|
|
|
|
|
pitch_delay_3x = ff_acelp_decode_5_6_bit_to_2nd_delay3(ac_index, pitch_delay_min); |
|
|
|
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/* Round pitch delay to nearest (used everywhere except ff_acelp_interpolate). */ |
|
|
|
ma_predictor = get_bits(&gb, 1); |
|
|
|
pitch_delay_int[i] = (pitch_delay_3x + 1) / 3; |
|
|
|
quantizer_1st = get_bits(&gb, VQ_1ST_BITS); |
|
|
|
if (pitch_delay_int[i] > PITCH_DELAY_MAX) { |
|
|
|
quantizer_2nd_lo = get_bits(&gb, VQ_2ND_BITS); |
|
|
|
av_log(avctx, AV_LOG_WARNING, "pitch_delay_int %d is too large\n", pitch_delay_int[i]); |
|
|
|
quantizer_2nd_hi = get_bits(&gb, VQ_2ND_BITS); |
|
|
|
pitch_delay_int[i] = PITCH_DELAY_MAX; |
|
|
|
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if (frame_erasure) { |
|
|
|
if (frame_erasure) { |
|
|
|
ctx->rand_value = g729_prng(ctx->rand_value); |
|
|
|
lsf_restore_from_previous(ctx->lsfq, ctx->past_quantizer_outputs, |
|
|
|
fc_indexes = av_mod_uintp2(ctx->rand_value, format->fc_indexes_bits); |
|
|
|
ctx->ma_predictor_prev); |
|
|
|
|
|
|
|
} else { |
|
|
|
ctx->rand_value = g729_prng(ctx->rand_value); |
|
|
|
lsf_decode(ctx->lsfq, ctx->past_quantizer_outputs, |
|
|
|
pulses_signs = ctx->rand_value; |
|
|
|
ma_predictor, |
|
|
|
|
|
|
|
quantizer_1st, quantizer_2nd_lo, quantizer_2nd_hi); |
|
|
|
|
|
|
|
ctx->ma_predictor_prev = ma_predictor; |
|
|
|
} |
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
tmp = ctx->past_quantizer_outputs[MA_NP]; |
|
|
|
|
|
|
|
memmove(ctx->past_quantizer_outputs + 1, ctx->past_quantizer_outputs, |
|
|
|
|
|
|
|
MA_NP * sizeof(int16_t*)); |
|
|
|
|
|
|
|
ctx->past_quantizer_outputs[0] = tmp; |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ff_acelp_lsf2lsp(ctx->lsp[1], ctx->lsfq, 10); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ff_acelp_lp_decode(&lp[0][0], &lp[1][0], ctx->lsp[1], ctx->lsp[0], 10); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
FFSWAP(int16_t*, ctx->lsp[1], ctx->lsp[0]); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < 2; i++) { |
|
|
|
|
|
|
|
int gain_corr_factor; |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
uint8_t ac_index; ///< adaptive codebook index
|
|
|
|
|
|
|
|
uint8_t pulses_signs; ///< fixed-codebook vector pulse signs
|
|
|
|
|
|
|
|
int fc_indexes; ///< fixed-codebook indexes
|
|
|
|
|
|
|
|
uint8_t gc_1st_index; ///< gain codebook (first stage) index
|
|
|
|
|
|
|
|
uint8_t gc_2nd_index; ///< gain codebook (second stage) index
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ac_index = get_bits(&gb, format->ac_index_bits[i]); |
|
|
|
|
|
|
|
if (!i && format->parity_bit) |
|
|
|
|
|
|
|
bad_pitch = av_parity(ac_index >> 2) == get_bits1(&gb); |
|
|
|
|
|
|
|
fc_indexes = get_bits(&gb, format->fc_indexes_bits); |
|
|
|
|
|
|
|
pulses_signs = get_bits(&gb, format->fc_signs_bits); |
|
|
|
|
|
|
|
gc_1st_index = get_bits(&gb, format->gc_1st_index_bits); |
|
|
|
|
|
|
|
gc_2nd_index = get_bits(&gb, format->gc_2nd_index_bits); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if (frame_erasure) { |
|
|
|
|
|
|
|
pitch_delay_3x = 3 * ctx->pitch_delay_int_prev; |
|
|
|
|
|
|
|
} else if (!i) { |
|
|
|
|
|
|
|
if (bad_pitch) { |
|
|
|
|
|
|
|
pitch_delay_3x = 3 * ctx->pitch_delay_int_prev; |
|
|
|
|
|
|
|
} else { |
|
|
|
|
|
|
|
pitch_delay_3x = ff_acelp_decode_8bit_to_1st_delay3(ac_index); |
|
|
|
|
|
|
|
} |
|
|
|
|
|
|
|
} else { |
|
|
|
|
|
|
|
int pitch_delay_min = av_clip(ctx->pitch_delay_int_prev - 5, |
|
|
|
|
|
|
|
PITCH_DELAY_MIN, PITCH_DELAY_MAX - 9); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if (packet_type == FORMAT_G729D_6K4) { |
|
|
|
|
|
|
|
pitch_delay_3x = ff_acelp_decode_4bit_to_2nd_delay3(ac_index, pitch_delay_min); |
|
|
|
|
|
|
|
} else { |
|
|
|
|
|
|
|
pitch_delay_3x = ff_acelp_decode_5_6_bit_to_2nd_delay3(ac_index, pitch_delay_min); |
|
|
|
|
|
|
|
} |
|
|
|
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
memset(fc, 0, sizeof(int16_t) * SUBFRAME_SIZE); |
|
|
|
/* Round pitch delay to nearest (used everywhere except ff_acelp_interpolate). */ |
|
|
|
switch (packet_type) { |
|
|
|
pitch_delay_int[i] = (pitch_delay_3x + 1) / 3; |
|
|
|
case FORMAT_G729_8K: |
|
|
|
if (pitch_delay_int[i] > PITCH_DELAY_MAX) { |
|
|
|
ff_acelp_fc_pulse_per_track(fc, ff_fc_4pulses_8bits_tracks_13, |
|
|
|
av_log(avctx, AV_LOG_WARNING, "pitch_delay_int %d is too large\n", pitch_delay_int[i]); |
|
|
|
ff_fc_4pulses_8bits_track_4, |
|
|
|
pitch_delay_int[i] = PITCH_DELAY_MAX; |
|
|
|
fc_indexes, pulses_signs, 3, 3); |
|
|
|
} |
|
|
|
break; |
|
|
|
|
|
|
|
case FORMAT_G729D_6K4: |
|
|
|
|
|
|
|
ff_acelp_fc_pulse_per_track(fc, ff_fc_2pulses_9bits_track1_gray, |
|
|
|
|
|
|
|
ff_fc_2pulses_9bits_track2_gray, |
|
|
|
|
|
|
|
fc_indexes, pulses_signs, 1, 4); |
|
|
|
|
|
|
|
break; |
|
|
|
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
|
|
|
This filter enhances harmonic components of the fixed-codebook vector to |
|
|
|
|
|
|
|
improve the quality of the reconstructed speech. |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/ fc_v[i], i < pitch_delay |
|
|
|
if (frame_erasure) { |
|
|
|
fc_v[i] = < |
|
|
|
ctx->rand_value = g729_prng(ctx->rand_value); |
|
|
|
\ fc_v[i] + gain_pitch * fc_v[i-pitch_delay], i >= pitch_delay |
|
|
|
fc_indexes = av_mod_uintp2(ctx->rand_value, format->fc_indexes_bits); |
|
|
|
*/ |
|
|
|
|
|
|
|
ff_acelp_weighted_vector_sum(fc + pitch_delay_int[i], |
|
|
|
|
|
|
|
fc + pitch_delay_int[i], |
|
|
|
|
|
|
|
fc, 1 << 14, |
|
|
|
|
|
|
|
av_clip(ctx->past_gain_pitch[0], SHARP_MIN, SHARP_MAX), |
|
|
|
|
|
|
|
0, 14, |
|
|
|
|
|
|
|
SUBFRAME_SIZE - pitch_delay_int[i]); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
memmove(ctx->past_gain_pitch+1, ctx->past_gain_pitch, 5 * sizeof(int16_t)); |
|
|
|
ctx->rand_value = g729_prng(ctx->rand_value); |
|
|
|
ctx->past_gain_code[1] = ctx->past_gain_code[0]; |
|
|
|
pulses_signs = ctx->rand_value; |
|
|
|
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
if (frame_erasure) { |
|
|
|
|
|
|
|
ctx->past_gain_pitch[0] = (29491 * ctx->past_gain_pitch[0]) >> 15; // 0.90 (0.15)
|
|
|
|
|
|
|
|
ctx->past_gain_code[0] = ( 2007 * ctx->past_gain_code[0] ) >> 11; // 0.98 (0.11)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
gain_corr_factor = 0; |
|
|
|
memset(fc, 0, sizeof(int16_t) * SUBFRAME_SIZE); |
|
|
|
} else { |
|
|
|
switch (packet_type) { |
|
|
|
if (packet_type == FORMAT_G729D_6K4) { |
|
|
|
case FORMAT_G729_8K: |
|
|
|
ctx->past_gain_pitch[0] = cb_gain_1st_6k4[gc_1st_index][0] + |
|
|
|
ff_acelp_fc_pulse_per_track(fc, ff_fc_4pulses_8bits_tracks_13, |
|
|
|
cb_gain_2nd_6k4[gc_2nd_index][0]; |
|
|
|
ff_fc_4pulses_8bits_track_4, |
|
|
|
gain_corr_factor = cb_gain_1st_6k4[gc_1st_index][1] + |
|
|
|
fc_indexes, pulses_signs, 3, 3); |
|
|
|
cb_gain_2nd_6k4[gc_2nd_index][1]; |
|
|
|
break; |
|
|
|
|
|
|
|
case FORMAT_G729D_6K4: |
|
|
|
/* Without check below overflow can occur in ff_acelp_update_past_gain.
|
|
|
|
ff_acelp_fc_pulse_per_track(fc, ff_fc_2pulses_9bits_track1_gray, |
|
|
|
It is not issue for G.729, because gain_corr_factor in it's case is always |
|
|
|
ff_fc_2pulses_9bits_track2_gray, |
|
|
|
greater than 1024, while in G.729D it can be even zero. */ |
|
|
|
fc_indexes, pulses_signs, 1, 4); |
|
|
|
gain_corr_factor = FFMAX(gain_corr_factor, 1024); |
|
|
|
break; |
|
|
|
#ifndef G729_BITEXACT |
|
|
|
|
|
|
|
gain_corr_factor >>= 1; |
|
|
|
|
|
|
|
#endif |
|
|
|
|
|
|
|
} else { |
|
|
|
|
|
|
|
ctx->past_gain_pitch[0] = cb_gain_1st_8k[gc_1st_index][0] + |
|
|
|
|
|
|
|
cb_gain_2nd_8k[gc_2nd_index][0]; |
|
|
|
|
|
|
|
gain_corr_factor = cb_gain_1st_8k[gc_1st_index][1] + |
|
|
|
|
|
|
|
cb_gain_2nd_8k[gc_2nd_index][1]; |
|
|
|
|
|
|
|
} |
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
/* Decode the fixed-codebook gain. */ |
|
|
|
|
|
|
|
ctx->past_gain_code[0] = ff_acelp_decode_gain_code(&s->adsp, gain_corr_factor, |
|
|
|
|
|
|
|
fc, MR_ENERGY, |
|
|
|
|
|
|
|
ctx->quant_energy, |
|
|
|
|
|
|
|
ma_prediction_coeff, |
|
|
|
|
|
|
|
SUBFRAME_SIZE, 4); |
|
|
|
|
|
|
|
#ifdef G729_BITEXACT |
|
|
|
|
|
|
|
/*
|
|
|
|
/*
|
|
|
|
This correction required to get bit-exact result with |
|
|
|
This filter enhances harmonic components of the fixed-codebook vector to |
|
|
|
reference code, because gain_corr_factor in G.729D is |
|
|
|
improve the quality of the reconstructed speech. |
|
|
|
two times larger than in original G.729. |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
If bit-exact result is not issue then gain_corr_factor |
|
|
|
/ fc_v[i], i < pitch_delay |
|
|
|
can be simpler divided by 2 before call to g729_get_gain_code |
|
|
|
fc_v[i] = < |
|
|
|
instead of using correction below. |
|
|
|
\ fc_v[i] + gain_pitch * fc_v[i-pitch_delay], i >= pitch_delay |
|
|
|
*/ |
|
|
|
*/ |
|
|
|
if (packet_type == FORMAT_G729D_6K4) { |
|
|
|
ff_acelp_weighted_vector_sum(fc + pitch_delay_int[i], |
|
|
|
gain_corr_factor >>= 1; |
|
|
|
fc + pitch_delay_int[i], |
|
|
|
ctx->past_gain_code[0] >>= 1; |
|
|
|
fc, 1 << 14, |
|
|
|
|
|
|
|
av_clip(ctx->past_gain_pitch[0], SHARP_MIN, SHARP_MAX), |
|
|
|
|
|
|
|
0, 14, |
|
|
|
|
|
|
|
SUBFRAME_SIZE - pitch_delay_int[i]); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
memmove(ctx->past_gain_pitch+1, ctx->past_gain_pitch, 5 * sizeof(int16_t)); |
|
|
|
|
|
|
|
ctx->past_gain_code[1] = ctx->past_gain_code[0]; |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if (frame_erasure) { |
|
|
|
|
|
|
|
ctx->past_gain_pitch[0] = (29491 * ctx->past_gain_pitch[0]) >> 15; // 0.90 (0.15)
|
|
|
|
|
|
|
|
ctx->past_gain_code[0] = ( 2007 * ctx->past_gain_code[0] ) >> 11; // 0.98 (0.11)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
gain_corr_factor = 0; |
|
|
|
|
|
|
|
} else { |
|
|
|
|
|
|
|
if (packet_type == FORMAT_G729D_6K4) { |
|
|
|
|
|
|
|
ctx->past_gain_pitch[0] = cb_gain_1st_6k4[gc_1st_index][0] + |
|
|
|
|
|
|
|
cb_gain_2nd_6k4[gc_2nd_index][0]; |
|
|
|
|
|
|
|
gain_corr_factor = cb_gain_1st_6k4[gc_1st_index][1] + |
|
|
|
|
|
|
|
cb_gain_2nd_6k4[gc_2nd_index][1]; |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/* Without check below overflow can occur in ff_acelp_update_past_gain.
|
|
|
|
|
|
|
|
It is not issue for G.729, because gain_corr_factor in it's case is always |
|
|
|
|
|
|
|
greater than 1024, while in G.729D it can be even zero. */ |
|
|
|
|
|
|
|
gain_corr_factor = FFMAX(gain_corr_factor, 1024); |
|
|
|
|
|
|
|
#ifndef G729_BITEXACT |
|
|
|
|
|
|
|
gain_corr_factor >>= 1; |
|
|
|
|
|
|
|
#endif |
|
|
|
|
|
|
|
} else { |
|
|
|
|
|
|
|
ctx->past_gain_pitch[0] = cb_gain_1st_8k[gc_1st_index][0] + |
|
|
|
|
|
|
|
cb_gain_2nd_8k[gc_2nd_index][0]; |
|
|
|
|
|
|
|
gain_corr_factor = cb_gain_1st_8k[gc_1st_index][1] + |
|
|
|
|
|
|
|
cb_gain_2nd_8k[gc_2nd_index][1]; |
|
|
|
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/* Decode the fixed-codebook gain. */ |
|
|
|
|
|
|
|
ctx->past_gain_code[0] = ff_acelp_decode_gain_code(&s->adsp, gain_corr_factor, |
|
|
|
|
|
|
|
fc, MR_ENERGY, |
|
|
|
|
|
|
|
ctx->quant_energy, |
|
|
|
|
|
|
|
ma_prediction_coeff, |
|
|
|
|
|
|
|
SUBFRAME_SIZE, 4); |
|
|
|
|
|
|
|
#ifdef G729_BITEXACT |
|
|
|
|
|
|
|
/*
|
|
|
|
|
|
|
|
This correction required to get bit-exact result with |
|
|
|
|
|
|
|
reference code, because gain_corr_factor in G.729D is |
|
|
|
|
|
|
|
two times larger than in original G.729. |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
If bit-exact result is not issue then gain_corr_factor |
|
|
|
|
|
|
|
can be simpler divided by 2 before call to g729_get_gain_code |
|
|
|
|
|
|
|
instead of using correction below. |
|
|
|
|
|
|
|
*/ |
|
|
|
|
|
|
|
if (packet_type == FORMAT_G729D_6K4) { |
|
|
|
|
|
|
|
gain_corr_factor >>= 1; |
|
|
|
|
|
|
|
ctx->past_gain_code[0] >>= 1; |
|
|
|
|
|
|
|
} |
|
|
|
|
|
|
|
#endif |
|
|
|
} |
|
|
|
} |
|
|
|
#endif |
|
|
|
ff_acelp_update_past_gain(ctx->quant_energy, gain_corr_factor, 2, frame_erasure); |
|
|
|
} |
|
|
|
|
|
|
|
ff_acelp_update_past_gain(ctx->quant_energy, gain_corr_factor, 2, frame_erasure); |
|
|
|
/* Routine requires rounding to lowest. */ |
|
|
|
|
|
|
|
ff_acelp_interpolate(ctx->exc + i * SUBFRAME_SIZE, |
|
|
|
/* Routine requires rounding to lowest. */ |
|
|
|
ctx->exc + i * SUBFRAME_SIZE - pitch_delay_3x / 3, |
|
|
|
ff_acelp_interpolate(ctx->exc + i * SUBFRAME_SIZE, |
|
|
|
ff_acelp_interp_filter, 6, |
|
|
|
ctx->exc + i * SUBFRAME_SIZE - pitch_delay_3x / 3, |
|
|
|
(pitch_delay_3x % 3) << 1, |
|
|
|
ff_acelp_interp_filter, 6, |
|
|
|
10, SUBFRAME_SIZE); |
|
|
|
(pitch_delay_3x % 3) << 1, |
|
|
|
|
|
|
|
10, SUBFRAME_SIZE); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ff_acelp_weighted_vector_sum(ctx->exc + i * SUBFRAME_SIZE, |
|
|
|
|
|
|
|
ctx->exc + i * SUBFRAME_SIZE, fc, |
|
|
|
|
|
|
|
(!ctx->was_periodic && frame_erasure) ? 0 : ctx->past_gain_pitch[0], |
|
|
|
|
|
|
|
( ctx->was_periodic && frame_erasure) ? 0 : ctx->past_gain_code[0], |
|
|
|
|
|
|
|
1 << 13, 14, SUBFRAME_SIZE); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
memcpy(synth, ctx->syn_filter_data, 10 * sizeof(int16_t)); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if (ff_celp_lp_synthesis_filter( |
|
|
|
|
|
|
|
synth+10, |
|
|
|
|
|
|
|
&lp[i][1], |
|
|
|
|
|
|
|
ctx->exc + i * SUBFRAME_SIZE, |
|
|
|
|
|
|
|
SUBFRAME_SIZE, |
|
|
|
|
|
|
|
10, |
|
|
|
|
|
|
|
1, |
|
|
|
|
|
|
|
0, |
|
|
|
|
|
|
|
0x800)) |
|
|
|
|
|
|
|
/* Overflow occurred, downscale excitation signal... */ |
|
|
|
|
|
|
|
for (j = 0; j < 2 * SUBFRAME_SIZE + PITCH_DELAY_MAX + INTERPOL_LEN; j++) |
|
|
|
|
|
|
|
ctx->exc_base[j] >>= 2; |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/* ... and make synthesis again. */ |
|
|
|
|
|
|
|
if (packet_type == FORMAT_G729D_6K4) { |
|
|
|
|
|
|
|
int16_t exc_new[SUBFRAME_SIZE]; |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ctx->onset = g729d_onset_decision(ctx->onset, ctx->past_gain_code); |
|
|
|
|
|
|
|
ctx->voice_decision = g729d_voice_decision(ctx->onset, ctx->voice_decision, ctx->past_gain_pitch); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
g729d_get_new_exc(exc_new, ctx->exc + i * SUBFRAME_SIZE, fc, ctx->voice_decision, ctx->past_gain_code[0], SUBFRAME_SIZE); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ff_celp_lp_synthesis_filter( |
|
|
|
|
|
|
|
synth+10, |
|
|
|
|
|
|
|
&lp[i][1], |
|
|
|
|
|
|
|
exc_new, |
|
|
|
|
|
|
|
SUBFRAME_SIZE, |
|
|
|
|
|
|
|
10, |
|
|
|
|
|
|
|
0, |
|
|
|
|
|
|
|
0, |
|
|
|
|
|
|
|
0x800); |
|
|
|
|
|
|
|
} else { |
|
|
|
|
|
|
|
ff_celp_lp_synthesis_filter( |
|
|
|
|
|
|
|
synth+10, |
|
|
|
|
|
|
|
&lp[i][1], |
|
|
|
|
|
|
|
ctx->exc + i * SUBFRAME_SIZE, |
|
|
|
|
|
|
|
SUBFRAME_SIZE, |
|
|
|
|
|
|
|
10, |
|
|
|
|
|
|
|
0, |
|
|
|
|
|
|
|
0, |
|
|
|
|
|
|
|
0x800); |
|
|
|
|
|
|
|
} |
|
|
|
|
|
|
|
/* Save data (without postfilter) for use in next subframe. */ |
|
|
|
|
|
|
|
memcpy(ctx->syn_filter_data, synth+SUBFRAME_SIZE, 10 * sizeof(int16_t)); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/* Calculate gain of unfiltered signal for use in AGC. */ |
|
|
|
|
|
|
|
gain_before = 0; |
|
|
|
|
|
|
|
for (j = 0; j < SUBFRAME_SIZE; j++) |
|
|
|
|
|
|
|
gain_before += FFABS(synth[j+10]); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/* Call postfilter and also update voicing decision for use in next frame. */ |
|
|
|
|
|
|
|
ff_g729_postfilter( |
|
|
|
|
|
|
|
&s->adsp, |
|
|
|
|
|
|
|
&ctx->ht_prev_data, |
|
|
|
|
|
|
|
&is_periodic, |
|
|
|
|
|
|
|
&lp[i][0], |
|
|
|
|
|
|
|
pitch_delay_int[0], |
|
|
|
|
|
|
|
ctx->residual, |
|
|
|
|
|
|
|
ctx->res_filter_data, |
|
|
|
|
|
|
|
ctx->pos_filter_data, |
|
|
|
|
|
|
|
synth+10, |
|
|
|
|
|
|
|
SUBFRAME_SIZE); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/* Calculate gain of filtered signal for use in AGC. */ |
|
|
|
ff_acelp_weighted_vector_sum(ctx->exc + i * SUBFRAME_SIZE, |
|
|
|
gain_after = 0; |
|
|
|
ctx->exc + i * SUBFRAME_SIZE, fc, |
|
|
|
for(j=0; j<SUBFRAME_SIZE; j++) |
|
|
|
(!ctx->was_periodic && frame_erasure) ? 0 : ctx->past_gain_pitch[0], |
|
|
|
gain_after += FFABS(synth[j+10]); |
|
|
|
( ctx->was_periodic && frame_erasure) ? 0 : ctx->past_gain_code[0], |
|
|
|
|
|
|
|
1 << 13, 14, SUBFRAME_SIZE); |
|
|
|
|
|
|
|
|
|
|
|
ctx->gain_coeff = ff_g729_adaptive_gain_control( |
|
|
|
memcpy(synth, ctx->syn_filter_data, 10 * sizeof(int16_t)); |
|
|
|
gain_before, |
|
|
|
|
|
|
|
gain_after, |
|
|
|
if (ff_celp_lp_synthesis_filter( |
|
|
|
synth+10, |
|
|
|
synth+10, |
|
|
|
|
|
|
|
&lp[i][1], |
|
|
|
|
|
|
|
ctx->exc + i * SUBFRAME_SIZE, |
|
|
|
SUBFRAME_SIZE, |
|
|
|
SUBFRAME_SIZE, |
|
|
|
ctx->gain_coeff); |
|
|
|
10, |
|
|
|
|
|
|
|
1, |
|
|
|
|
|
|
|
0, |
|
|
|
|
|
|
|
0x800)) |
|
|
|
|
|
|
|
/* Overflow occurred, downscale excitation signal... */ |
|
|
|
|
|
|
|
for (j = 0; j < 2 * SUBFRAME_SIZE + PITCH_DELAY_MAX + INTERPOL_LEN; j++) |
|
|
|
|
|
|
|
ctx->exc_base[j] >>= 2; |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/* ... and make synthesis again. */ |
|
|
|
|
|
|
|
if (packet_type == FORMAT_G729D_6K4) { |
|
|
|
|
|
|
|
int16_t exc_new[SUBFRAME_SIZE]; |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ctx->onset = g729d_onset_decision(ctx->onset, ctx->past_gain_code); |
|
|
|
|
|
|
|
ctx->voice_decision = g729d_voice_decision(ctx->onset, ctx->voice_decision, ctx->past_gain_pitch); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
g729d_get_new_exc(exc_new, ctx->exc + i * SUBFRAME_SIZE, fc, ctx->voice_decision, ctx->past_gain_code[0], SUBFRAME_SIZE); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ff_celp_lp_synthesis_filter( |
|
|
|
|
|
|
|
synth+10, |
|
|
|
|
|
|
|
&lp[i][1], |
|
|
|
|
|
|
|
exc_new, |
|
|
|
|
|
|
|
SUBFRAME_SIZE, |
|
|
|
|
|
|
|
10, |
|
|
|
|
|
|
|
0, |
|
|
|
|
|
|
|
0, |
|
|
|
|
|
|
|
0x800); |
|
|
|
|
|
|
|
} else { |
|
|
|
|
|
|
|
ff_celp_lp_synthesis_filter( |
|
|
|
|
|
|
|
synth+10, |
|
|
|
|
|
|
|
&lp[i][1], |
|
|
|
|
|
|
|
ctx->exc + i * SUBFRAME_SIZE, |
|
|
|
|
|
|
|
SUBFRAME_SIZE, |
|
|
|
|
|
|
|
10, |
|
|
|
|
|
|
|
0, |
|
|
|
|
|
|
|
0, |
|
|
|
|
|
|
|
0x800); |
|
|
|
|
|
|
|
} |
|
|
|
|
|
|
|
/* Save data (without postfilter) for use in next subframe. */ |
|
|
|
|
|
|
|
memcpy(ctx->syn_filter_data, synth+SUBFRAME_SIZE, 10 * sizeof(int16_t)); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/* Calculate gain of unfiltered signal for use in AGC. */ |
|
|
|
|
|
|
|
gain_before = 0; |
|
|
|
|
|
|
|
for (j = 0; j < SUBFRAME_SIZE; j++) |
|
|
|
|
|
|
|
gain_before += FFABS(synth[j+10]); |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/* Call postfilter and also update voicing decision for use in next frame. */ |
|
|
|
|
|
|
|
ff_g729_postfilter( |
|
|
|
|
|
|
|
&s->adsp, |
|
|
|
|
|
|
|
&ctx->ht_prev_data, |
|
|
|
|
|
|
|
&is_periodic, |
|
|
|
|
|
|
|
&lp[i][0], |
|
|
|
|
|
|
|
pitch_delay_int[0], |
|
|
|
|
|
|
|
ctx->residual, |
|
|
|
|
|
|
|
ctx->res_filter_data, |
|
|
|
|
|
|
|
ctx->pos_filter_data, |
|
|
|
|
|
|
|
synth+10, |
|
|
|
|
|
|
|
SUBFRAME_SIZE); |
|
|
|
|
|
|
|
|
|
|
|
if (frame_erasure) |
|
|
|
/* Calculate gain of filtered signal for use in AGC. */ |
|
|
|
ctx->pitch_delay_int_prev = FFMIN(ctx->pitch_delay_int_prev + 1, PITCH_DELAY_MAX); |
|
|
|
gain_after = 0; |
|
|
|
else |
|
|
|
for (j = 0; j < SUBFRAME_SIZE; j++) |
|
|
|
ctx->pitch_delay_int_prev = pitch_delay_int[i]; |
|
|
|
gain_after += FFABS(synth[j+10]); |
|
|
|
|
|
|
|
|
|
|
|
memcpy(synth+8, ctx->hpf_z, 2*sizeof(int16_t)); |
|
|
|
ctx->gain_coeff = ff_g729_adaptive_gain_control( |
|
|
|
ff_acelp_high_pass_filter( |
|
|
|
gain_before, |
|
|
|
out_frame + i*SUBFRAME_SIZE, |
|
|
|
gain_after, |
|
|
|
ctx->hpf_f, |
|
|
|
synth+10, |
|
|
|
synth+10, |
|
|
|
SUBFRAME_SIZE, |
|
|
|
SUBFRAME_SIZE); |
|
|
|
ctx->gain_coeff); |
|
|
|
memcpy(ctx->hpf_z, synth+8+SUBFRAME_SIZE, 2*sizeof(int16_t)); |
|
|
|
|
|
|
|
} |
|
|
|
if (frame_erasure) { |
|
|
|
|
|
|
|
ctx->pitch_delay_int_prev = FFMIN(ctx->pitch_delay_int_prev + 1, PITCH_DELAY_MAX); |
|
|
|
|
|
|
|
} else { |
|
|
|
|
|
|
|
ctx->pitch_delay_int_prev = pitch_delay_int[i]; |
|
|
|
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
memcpy(synth+8, ctx->hpf_z, 2*sizeof(int16_t)); |
|
|
|
|
|
|
|
ff_acelp_high_pass_filter( |
|
|
|
|
|
|
|
out_frame + i*SUBFRAME_SIZE, |
|
|
|
|
|
|
|
ctx->hpf_f, |
|
|
|
|
|
|
|
synth+10, |
|
|
|
|
|
|
|
SUBFRAME_SIZE); |
|
|
|
|
|
|
|
memcpy(ctx->hpf_z, synth+8+SUBFRAME_SIZE, 2*sizeof(int16_t)); |
|
|
|
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
ctx->was_periodic = is_periodic; |
|
|
|
ctx->was_periodic = is_periodic; |
|
|
|
|
|
|
|
|
|
|
|
/* Save signal for use in next frame. */ |
|
|
|
/* Save signal for use in next frame. */ |
|
|
|
memmove(ctx->exc_base, ctx->exc_base + 2 * SUBFRAME_SIZE, (PITCH_DELAY_MAX+INTERPOL_LEN)*sizeof(int16_t)); |
|
|
|
memmove(ctx->exc_base, ctx->exc_base + 2 * SUBFRAME_SIZE, (PITCH_DELAY_MAX+INTERPOL_LEN)*sizeof(int16_t)); |
|
|
|
|
|
|
|
|
|
|
|
buf += packet_type == FORMAT_G729_8K ? G729_8K_BLOCK_SIZE : G729D_6K4_BLOCK_SIZE; |
|
|
|
buf += packet_type == FORMAT_G729_8K ? G729_8K_BLOCK_SIZE : G729D_6K4_BLOCK_SIZE; |
|
|
|
ctx++; |
|
|
|
ctx++; |
|
|
|
} |
|
|
|
} |
|
|
|
|
|
|
|
|
|
|
|
*got_frame_ptr = 1; |
|
|
|
*got_frame_ptr = 1; |
|
|
|