[FFmpeg-devel] [PATCH] AAC Encoder, Round 2
Kostya
kostya.shishkov
Sat Aug 23 17:31:30 CEST 2008
I'm back (feeling even worse than before but nm).
Here is $subj is in a form of diff against FFmpeg SVN.
-------------- next part --------------
Index: aacenc.c
===================================================================
--- aacenc.c (revision 14920)
+++ aacenc.c (working copy)
@@ -28,6 +28,7 @@
* TODOs:
* psy model selection with some option
* add sane pulse detection
+ * add temporal noise shaping
***********************************/
#include "avcodec.h"
@@ -118,6 +119,29 @@
swb_size_128_16, swb_size_128_16, swb_size_128_8
};
+/** spectral coefficients codebook information */
+static const struct {
+ int16_t maxval; ///< maximum possible value
+ int8_t range; ///< value used in vector calculation
+} aac_cb_info[] = {
+ { 0, -1 }, // zero codebook
+ { 1, 3 },
+ { 1, 3 },
+ { 2, 3 },
+ { 2, 3 },
+ { 4, 9 },
+ { 4, 9 },
+ { 7, 8 },
+ { 7, 8 },
+ { 12, 13 },
+ { 12, 13 },
+ { 8191, 17 },
+ { -1, -1 }, // reserved
+ { -1, -1 }, // perceptual noise substitution
+ { -1, -1 }, // intensity out-of-phase
+ { -1, -1 }, // intensity in-phase
+};
+
/** bits needed to code codebook run value for long windows */
static const uint8_t run_value_bits_long[64] = {
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
@@ -211,10 +235,6 @@
return -1;
}
s->samplerate_index = i;
- s->swb_sizes1024 = swb_size_1024[i];
- s->swb_num1024 = ff_aac_num_swb_1024[i];
- s->swb_sizes128 = swb_size_128[i];
- s->swb_num128 = ff_aac_num_swb_128[i];
dsputil_init(&s->dsp, avctx);
ff_mdct_init(&s->mdct1024, 11, 0);
@@ -229,7 +249,7 @@
s->cpe = av_mallocz(sizeof(ChannelElement) * aac_chan_configs[avctx->channels-1][0]);
if(ff_aac_psy_init(&s->psy, avctx, AAC_PSY_3GPP,
aac_chan_configs[avctx->channels-1][0], 0,
- s->swb_sizes1024, s->swb_num1024, s->swb_sizes128, s->swb_num128) < 0){
+ swb_size_1024[i], ff_aac_num_swb_1024[i], swb_size_128[i], ff_aac_num_swb_128[i]) < 0){
av_log(avctx, AV_LOG_ERROR, "Cannot initialize selected model.\n");
return -1;
}
@@ -239,14 +259,65 @@
return 0;
}
+static void apply_window_and_mdct(AVCodecContext *avctx, AACEncContext *s,
+ SingleChannelElement *sce, short *audio, int channel)
+{
+ int i, j, k;
+ const float * lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
+ const float * swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
+ const float * pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
+
+ if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
+ memcpy(s->output, sce->saved, sizeof(float)*1024);
+ if(sce->ics.window_sequence[0] == LONG_STOP_SEQUENCE){
+ memset(s->output, 0, sizeof(s->output[0]) * 448);
+ for(i = 448; i < 576; i++)
+ s->output[i] = sce->saved[i] * pwindow[i - 448];
+ for(i = 576; i < 704; i++)
+ s->output[i] = sce->saved[i];
+ }
+ if(sce->ics.window_sequence[0] != LONG_START_SEQUENCE){
+ j = channel;
+ for (i = 0; i < 1024; i++, j += avctx->channels){
+ s->output[i+1024] = audio[j] * lwindow[1024 - i - 1];
+ sce->saved[i] = audio[j] * lwindow[i];
+ }
+ }else{
+ j = channel;
+ for(i = 0; i < 448; i++, j += avctx->channels)
+ s->output[i+1024] = audio[j];
+ for(i = 448; i < 576; i++, j += avctx->channels)
+ s->output[i+1024] = audio[j] * swindow[576 - i - 1];
+ memset(s->output+1024+576, 0, sizeof(s->output[0]) * 448);
+ j = channel;
+ for(i = 0; i < 1024; i++, j += avctx->channels)
+ sce->saved[i] = audio[j];
+ }
+ ff_mdct_calc(&s->mdct1024, sce->coeffs, s->output);
+ }else{
+ j = channel;
+ for (k = 0; k < 1024; k += 128) {
+ for(i = 448 + k; i < 448 + k + 256; i++)
+ s->output[i - 448 - k] = (i < 1024)
+ ? sce->saved[i]
+ : audio[channel + (i-1024)*avctx->channels];
+ s->dsp.vector_fmul (s->output, k ? swindow : pwindow, 128);
+ s->dsp.vector_fmul_reverse(s->output+128, s->output+128, swindow, 128);
+ ff_mdct_calc(&s->mdct128, sce->coeffs + k, s->output);
+ }
+ j = channel;
+ for(i = 0; i < 1024; i++, j += avctx->channels)
+ sce->saved[i] = audio[j];
+ }
+}
+
/**
* Encode ics_info element.
* @see Table 4.6 (syntax of ics_info)
*/
-static void put_ics_info(AVCodecContext *avctx, IndividualChannelStream *info)
+static void put_ics_info(AACEncContext *s, IndividualChannelStream *info)
{
- AACEncContext *s = avctx->priv_data;
- int i;
+ int wg;
put_bits(&s->pb, 1, 0); // ics_reserved bit
put_bits(&s->pb, 2, info->window_sequence[0]);
@@ -256,12 +327,294 @@
put_bits(&s->pb, 1, 0); // no prediction
}else{
put_bits(&s->pb, 4, info->max_sfb);
- for(i = 1; i < info->num_windows; i++)
- put_bits(&s->pb, 1, info->group_len[i]);
+ for(wg = 0; wg < info->num_window_groups; wg++){
+ if(wg)
+ put_bits(&s->pb, 1, 0);
+ if(info->group_len[wg] > 1)
+ put_sbits(&s->pb, info->group_len[wg] - 1, 0xFF);
+ }
}
}
/**
+ * Encode MS data.
+ * @see 4.6.8.1 "Joint Coding - M/S Stereo"
+ */
+static void encode_ms_info(PutBitContext *pb, ChannelElement *cpe)
+{
+ int i, w, wg;
+
+ put_bits(pb, 2, cpe->ms_mode);
+ if(cpe->ms_mode == 1){
+ w = 0;
+ for(wg = 0; wg < cpe->ch[0].ics.num_window_groups; wg++){
+ for(i = 0; i < cpe->ch[0].ics.max_sfb; i++)
+ put_bits(pb, 1, cpe->ms_mask[w + i]);
+ w += cpe->ch[0].ics.group_len[wg]*16;
+ }
+ }
+}
+
+/**
+ * Calculate the number of bits needed to code all coefficient signs in current band.
+ */
+static int calculate_band_sign_bits(AACEncContext *s, SingleChannelElement *sce,
+ int group_len, int start, int size)
+{
+ int bits = 0;
+ int i, w;
+ for(w = 0; w < group_len; w++){
+ for(i = 0; i < size; i++){
+ if(sce->icoefs[start + i])
+ bits++;
+ }
+ start += 128;
+ }
+ return bits;
+}
+
+/**
+ * Calculate the number of bits needed to code given band with given codebook.
+ *
+ * @param s encoder context
+ * @param sce channel element
+ * @param group_len window group length
+ * @param start scalefactor band position in spectral coefficients
+ * @param size scalefactor band size
+ * @param cb codebook number
+ */
+static int calculate_band_bits(AACEncContext *s, SingleChannelElement *sce,
+ int group_len, int start, int size, int cb)
+{
+ int i, j, w;
+ int bits = 0, dim, idx;
+ int range = aac_cb_info[cb].range;
+
+ if(range == -1) return 0;
+ cb--;
+ dim = cb < FIRST_PAIR_BT ? 4 : 2;
+
+ if(IS_CODEBOOK_UNSIGNED(cb)){
+ int coef_abs[2];
+ for(w = 0; w < group_len; w++){
+ for(i = 0; i < size; i += dim){
+ idx = 0;
+ for(j = 0; j < dim; j++){
+ coef_abs[j] = FFABS(sce->icoefs[start+i+j]);
+ idx = idx * range + FFMIN(coef_abs[j], 16);
+ if(cb == ESC_BT && coef_abs[j] > 15)
+ bits += av_log2(coef_abs[j]) * 2 - 4 + 1;
+ }
+ bits += ff_aac_spectral_bits[cb][idx];
+ }
+ start += 128;
+ }
+ }else{
+ for(w = 0; w < group_len; w++){
+ for(i = 0; i < size; i += dim){
+ idx = sce->icoefs[start+i];
+ for(j = 1; j < dim; j++)
+ idx = idx * range + sce->icoefs[start+i+j];
+ //it turned out that all signed codebooks use the same offset for index coding
+ idx += 40;
+ bits += ff_aac_spectral_bits[cb][idx];
+ }
+ start += 128;
+ }
+ }
+ return bits;
+}
+
+/**
+ * Encode band info for single window group bands.
+ */
+static void encode_window_bands_info(AACEncContext *s, SingleChannelElement *sce,
+ int win, int group_len)
+{
+ BandCodingPath path[64];
+ int band_bits[64][12];
+ int maxval;
+ int w, swb, cb, start, start2, size;
+ int i, j;
+ const int max_sfb = sce->ics.max_sfb;
+ const int run_bits = sce->ics.num_windows == 1 ? 5 : 3;
+ const int run_esc = (1 << run_bits) - 1;
+ int bits, sbits, idx, count;
+ int stack[64], stack_len;
+
+ start = win*128;
+ for(swb = 0; swb < max_sfb; swb++){
+ maxval = 0;
+ start2 = start;
+ size = sce->ics.swb_sizes[swb];
+ if(sce->zeroes[win*16 + swb])
+ maxval = 0;
+ else{
+ for(w = 0; w < group_len; w++){
+ for(i = start2; i < start2 + size; i++){
+ maxval = FFMAX(maxval, FFABS(sce->icoefs[i]));
+ }
+ start2 += 128;
+ }
+ }
+ sbits = calculate_band_sign_bits(s, sce, group_len, start, size);
+ for(cb = 0; cb < 12; cb++){
+ if(aac_cb_info[cb].maxval < maxval)
+ band_bits[swb][cb] = INT_MAX;
+ else{
+ band_bits[swb][cb] = calculate_band_bits(s, sce, group_len, start, size, cb);
+ if(IS_CODEBOOK_UNSIGNED(cb-1)){
+ band_bits[swb][cb] += sbits;
+ }
+ }
+ }
+ start += sce->ics.swb_sizes[swb];
+ }
+ path[0].bits = 0;
+ for(i = 1; i <= max_sfb; i++)
+ path[i].bits = INT_MAX;
+ for(i = 0; i < max_sfb; i++){
+ for(cb = 0; cb < 12; cb++){
+ int sum = 0;
+ for(j = 1; j <= max_sfb - i; j++){
+ if(band_bits[i+j-1][cb] == INT_MAX)
+ break;
+ sum += band_bits[i+j-1][cb];
+ bits = sum + path[i].bits + run_value_bits[sce->ics.num_windows == 8][j];
+ if(bits < path[i+j].bits){
+ path[i+j].bits = bits;
+ path[i+j].codebook = cb;
+ path[i+j].prev_idx = i;
+ }
+ }
+ }
+ }
+ assert(path[max_sfb].bits != INT_MAX);
+
+ //convert resulting path from backward-linked list
+ stack_len = 0;
+ idx = max_sfb;
+ while(idx > 0){
+ stack[stack_len++] = idx;
+ idx = path[idx].prev_idx;
+ }
+
+ //perform actual band info encoding
+ start = 0;
+ for(i = stack_len - 1; i >= 0; i--){
+ put_bits(&s->pb, 4, path[stack[i]].codebook);
+ count = stack[i] - path[stack[i]].prev_idx;
+ memset(sce->zeroes + win*16 + start, !path[stack[i]].codebook, count);
+ //XXX: memset when band_type is also uint8_t
+ for(j = 0; j < count; j++){
+ sce->band_type[win*16 + start] = path[stack[i]].codebook;
+ start++;
+ }
+ while(count >= run_esc){
+ put_bits(&s->pb, run_bits, run_esc);
+ count -= run_esc;
+ }
+ put_bits(&s->pb, run_bits, count);
+ }
+}
+
+/**
+ * Encode the coefficients of one scalefactor band with selected codebook.
+ */
+static void encode_band_coeffs(AACEncContext *s, SingleChannelElement *sce,
+ int start, int size, int cb)
+{
+ const uint8_t *bits = ff_aac_spectral_bits [cb - 1];
+ const uint16_t *codes = ff_aac_spectral_codes[cb - 1];
+ const int range = aac_cb_info[cb].range;
+ const int dim = (cb < FIRST_PAIR_BT) ? 4 : 2;
+ int i, j, idx;
+
+ //do not encode zero or special codebooks
+ if(range == -1) return;
+
+ if(cb == ESC_BT){
+ int coef_abs[2];
+ for(i = start; i < start + size; i += 2){
+ idx = 0;
+ for(j = 0; j < 2; j++){
+ coef_abs[j] = FFABS(sce->icoefs[i+j]);
+ idx = idx*17 + FFMIN(coef_abs[j], 16);
+ }
+ put_bits(&s->pb, bits[idx], codes[idx]);
+ //output signs
+ for(j = 0; j < 2; j++)
+ if(sce->icoefs[i+j])
+ put_bits(&s->pb, 1, sce->icoefs[i+j] < 0);
+ //output escape values
+ for(j = 0; j < 2; j++)
+ if(coef_abs[j] > 15){
+ int len = av_log2(coef_abs[j]);
+
+ put_bits(&s->pb, len - 4 + 1, (1 << (len - 4 + 1)) - 2);
+ put_bits(&s->pb, len, coef_abs[j] & ((1 << len) - 1));
+ }
+ }
+ }else if(IS_CODEBOOK_UNSIGNED(cb)){
+ for(i = start; i < start + size; i += dim){
+ idx = FFABS(sce->icoefs[i]);
+ for(j = 1; j < dim; j++)
+ idx = idx * range + FFABS(sce->icoefs[i+j]);
+ put_bits(&s->pb, bits[idx], codes[idx]);
+ //output signs
+ for(j = 0; j < dim; j++)
+ if(sce->icoefs[i+j])
+ put_bits(&s->pb, 1, sce->icoefs[i+j] < 0);
+ }
+ }else{
+ for(i = start; i < start + size; i += dim){
+ idx = sce->icoefs[i];
+ for(j = 1; j < dim; j++)
+ idx = idx * range + sce->icoefs[i+j];
+ //it turned out that all signed codebooks use the same offset for index coding
+ idx += 40;
+ put_bits(&s->pb, bits[idx], codes[idx]);
+ }
+ }
+}
+
+/**
+ * Encode scalefactor band coding type.
+ */
+static void encode_band_info(AACEncContext *s, SingleChannelElement *sce)
+{
+ int w, wg;
+
+ w = 0;
+ for(wg = 0; wg < sce->ics.num_window_groups; wg++){
+ encode_window_bands_info(s, sce, w, sce->ics.group_len[wg]);
+ w += sce->ics.group_len[wg];
+ }
+}
+
+/**
+ * Encode scalefactors.
+ */
+static void encode_scale_factors(AVCodecContext *avctx, AACEncContext *s, SingleChannelElement *sce, int global_gain)
+{
+ int off = global_gain, diff;
+ int i, w, wg;
+
+ w = 0;
+ for(wg = 0; wg < sce->ics.num_window_groups; wg++){
+ for(i = 0; i < sce->ics.max_sfb; i++){
+ if(!sce->zeroes[w*16 + i]){
+ diff = sce->sf_idx[w*16 + i] - off + SCALE_DIFF_ZERO;
+ if(diff < 0 || diff > 120) av_log(avctx, AV_LOG_ERROR, "Scalefactor difference is too big to be coded\n");
+ off = sce->sf_idx[w*16 + i];
+ put_bits(&s->pb, ff_aac_scalefactor_bits[diff], ff_aac_scalefactor_code[diff]);
+ }
+ }
+ w += sce->ics.group_len[wg];
+ }
+}
+
+/**
* Encode pulse data.
*/
static void encode_pulses(AACEncContext *s, Pulse *pulse)
@@ -295,7 +648,7 @@
continue;
}
for(w2 = w; w2 < w + sce->ics.group_len[wg]; w2++){
- encode_band_coeffs(s, cpe, channel, start + w2*128,
+ encode_band_coeffs(s, sce, start + w2*128,
sce->ics.swb_sizes[i],
sce->band_type[w*16 + i]);
}
@@ -306,6 +659,44 @@
}
/**
+ * Encode one channel of audio data.
+ */
+static int encode_individual_channel(AVCodecContext *avctx, AACEncContext *s, SingleChannelElement *sce, int common_window)
+{
+ int g, w, wg;
+ int global_gain = 0, last = 256;
+
+ //determine global gain as standard recommends - the first scalefactor value
+ w = 0;
+ for(wg = 0; wg < sce->ics.num_window_groups; wg++){
+ for(g = sce->ics.max_sfb - 1; g >= 0; g--){
+ if(sce->sf_idx[w + g] == 256)
+ sce->sf_idx[w + g] = last;
+ else
+ last = sce->sf_idx[w + g];
+ }
+ for(g = 0; g < sce->ics.max_sfb; g++){
+ if(sce->sf_idx[w + g] == 256)
+ sce->sf_idx[w + g] = last;
+ else
+ last = sce->sf_idx[w + g];
+ }
+ w += sce->ics.group_len[wg]*16;
+ }
+ global_gain = last;
+
+ put_bits(&s->pb, 8, global_gain);
+ if(!common_window) put_ics_info(s, &sce->ics);
+ encode_band_info(s, sce);
+ encode_scale_factors(avctx, s, sce, global_gain);
+ encode_pulses(s, &sce->pulse);
+ put_bits(&s->pb, 1, 0); //tns
+ put_bits(&s->pb, 1, 0); //ssr
+ encode_spectral_coeffs(s, sce);
+ return 0;
+}
+
+/**
* Write some auxiliary information about the created AAC file.
*/
static void put_bitstream_info(AVCodecContext *avctx, AACEncContext *s, const char *name)
@@ -325,6 +716,80 @@
put_bits(&s->pb, 12 - padbits, 0);
}
+static int aac_encode_frame(AVCodecContext *avctx,
+ uint8_t *frame, int buf_size, void *data)
+{
+ AACEncContext *s = avctx->priv_data;
+ int16_t *samples = s->samples, *samples2, *la;
+ ChannelElement *cpe;
+ int i, j, chans, tag, start_ch;
+ const uint8_t *chan_map = aac_chan_configs[avctx->channels-1];
+ int chan_el_counter[4];
+
+ if(s->last_frame)
+ return 0;
+ if(data){
+ if((s->psy.flags & PSY_MODEL_NO_PREPROC) == PSY_MODEL_NO_PREPROC){
+ memcpy(s->samples + 1024 * avctx->channels, data, 1024 * avctx->channels * sizeof(s->samples[0]));
+ }else{
+ start_ch = 0;
+ samples2 = s->samples + 1024 * avctx->channels;
+ for(i = 0; i < chan_map[0]; i++){
+ tag = chan_map[i+1];
+ chans = tag == TYPE_CPE ? 2 : 1;
+ ff_aac_psy_preprocess(&s->psy, (uint16_t*)data + start_ch, samples2 + start_ch, i, tag);
+ start_ch += chans;
+ }
+ }
+ }
+ if(!avctx->frame_number){
+ memcpy(s->samples, s->samples + 1024 * avctx->channels, 1024 * avctx->channels * sizeof(s->samples[0]));
+ return 0;
+ }
+
+ init_put_bits(&s->pb, frame, buf_size*8);
+ if((avctx->frame_number & 0xFF)==1 && !(avctx->flags & CODEC_FLAG_BITEXACT)){
+ put_bitstream_info(avctx, s, LIBAVCODEC_IDENT);
+ }
+ start_ch = 0;
+ memset(chan_el_counter, 0, sizeof(chan_el_counter));
+ for(i = 0; i < chan_map[0]; i++){
+ tag = chan_map[i+1];
+ chans = tag == TYPE_CPE ? 2 : 1;
+ cpe = &s->cpe[i];
+ samples2 = samples + start_ch;
+ la = samples2 + 1024 * avctx->channels + start_ch;
+ if(!data) la = NULL;
+ ff_aac_psy_suggest_window(&s->psy, samples2, la, i, tag, cpe);
+ for(j = 0; j < chans; j++){
+ apply_window_and_mdct(avctx, s, &cpe->ch[j], samples2, j);
+ }
+ ff_aac_psy_analyze(&s->psy, i, tag, cpe);
+ put_bits(&s->pb, 3, tag);
+ put_bits(&s->pb, 4, chan_el_counter[tag]++);
+ if(chans == 2){
+ put_bits(&s->pb, 1, cpe->common_window);
+ if(cpe->common_window){
+ put_ics_info(s, &cpe->ch[0].ics);
+ encode_ms_info(&s->pb, cpe);
+ }
+ }
+ for(j = 0; j < chans; j++){
+ encode_individual_channel(avctx, s, &cpe->ch[j], cpe->common_window);
+ }
+ start_ch += chans;
+ }
+
+ put_bits(&s->pb, 3, TYPE_END);
+ flush_put_bits(&s->pb);
+ avctx->frame_bits = put_bits_count(&s->pb);
+
+ if(!data)
+ s->last_frame = 1;
+ memcpy(s->samples, s->samples + 1024 * avctx->channels, 1024 * avctx->channels * sizeof(s->samples[0]));
+ return put_bits_count(&s->pb)>>3;
+}
+
static av_cold int aac_encode_end(AVCodecContext *avctx)
{
AACEncContext *s = avctx->priv_data;
Index: aac.h
===================================================================
--- aac.h (revision 14920)
+++ aac.h (working copy)
@@ -133,6 +133,12 @@
AFTER_IMDCT = 3,
};
+#define SCALE_DIV_512 36 ///< scalefactor difference that corresponds to scale difference in 512 times
+#define SCALE_ONE_POS 140 ///< scalefactor index that corresponds to scale=1.0
+#define SCALE_MAX_POS 255 ///< scalefactor index maximum value
+#define SCALE_MAX_DIFF 60 ///< maximum scalefactor difference allowed by standard
+#define SCALE_DIFF_ZERO 60 ///< codebook index corresponding to zero scalefactor indices difference
+
/**
* Individual Channel Stream
*/
@@ -143,6 +149,7 @@
int num_window_groups;
uint8_t group_len[8];
const uint16_t *swb_offset; ///< table of offsets to the lowest spectral coefficient of a scalefactor band, sfb, for a particular window
+ const uint8_t *swb_sizes; ///< table of scalefactor band sizes for a particular window
int num_swb; ///< number of scalefactor window bands
int num_windows;
int tns_max_bands;
@@ -178,6 +185,7 @@
typedef struct {
int num_pulse;
+ int start;
int pos[4];
int amp[4];
} Pulse;
@@ -202,12 +210,16 @@
typedef struct {
IndividualChannelStream ics;
TemporalNoiseShaping tns;
- enum BandType band_type[120]; ///< band types
+ Pulse pulse;
+ enum BandType band_type[128]; ///< band types
int band_type_run_end[120]; ///< band type run end points
float sf[120]; ///< scalefactors
+ int sf_idx[128]; ///< scalefactor indices (used by encoder)
+ uint8_t zeroes[128]; ///< band is not coded (used by encoder)
DECLARE_ALIGNED_16(float, coeffs[1024]); ///< coefficients for IMDCT
- DECLARE_ALIGNED_16(float, saved[512]); ///< overlap
+ DECLARE_ALIGNED_16(float, saved[1024]); ///< overlap
DECLARE_ALIGNED_16(float, ret[1024]); ///< PCM output
+ DECLARE_ALIGNED_16(int, icoefs[1024]); ///< integer coefficients for encoding
} SingleChannelElement;
/**
@@ -215,6 +227,8 @@
*/
typedef struct {
// CPE specific
+ int common_window; ///< Set if channels share a common 'IndividualChannelStream' in bitstream.
+ int ms_mode; ///< Signals mid/side stereo flags coding mode (used by encoder)
uint8_t ms_mask[120]; ///< Set if mid/side stereo is used for each scalefactor window band
// shared
SingleChannelElement ch[2];
Index: aacpsy.c
===================================================================
--- aacpsy.c (revision 14920)
+++ aacpsy.c (working copy)
@@ -51,6 +51,23 @@
return av_clip((int)(pow(fabsf(coef) * Q, 0.75) + 0.4054), 0, 8191);
}
+/**
+ * Convert coefficients to integers.
+ * @return sum of coefficient absolute values
+ */
+static inline int quantize_coeffs(float *in, int *out, int size, int scale_idx)
+{
+ int i, sign, sum = 0;
+ const float Q = ff_aac_pow2sf_tab[200 - scale_idx + SCALE_ONE_POS - SCALE_DIV_512];
+ for(i = 0; i < size; i++){
+ sign = in[i] > 0.0;
+ out[i] = quant(in[i], Q);
+ sum += out[i];
+ if(sign) out[i] = -out[i];
+ }
+ return sum;
+}
+
static inline float get_approximate_quant_error(float *c, int size, int scale_idx)
{
int i;
@@ -68,11 +85,87 @@
}
/**
+ * Produce integer coefficients from scalefactors provided by the model.
+ */
+static void psy_create_output(AACPsyContext *apc, ChannelElement *cpe, int chans)
+{
+ int i, w, w2, wg, g, ch;
+ int start, sum, maxsfb, cmaxsfb;
+
+ for(ch = 0; ch < chans; ch++){
+ IndividualChannelStream *ics = &cpe->ch[ch].ics;
+ start = 0;
+ maxsfb = 0;
+ cpe->ch[ch].pulse.num_pulse = 0;
+ for(w = 0; w < ics->num_windows*16; w += 16){
+ for(g = 0; g < ics->num_swb; g++){
+ sum = 0;
+ //apply M/S
+ if(!ch && cpe->ms_mask[w + g]){
+ for(i = 0; i < ics->swb_sizes[g]; i++){
+ cpe->ch[0].coeffs[start+i] = (cpe->ch[0].coeffs[start+i] + cpe->ch[1].coeffs[start+i]) / 2.0;
+ cpe->ch[1].coeffs[start+i] = cpe->ch[0].coeffs[start+i] - cpe->ch[1].coeffs[start+i];
+ }
+ }
+ if(!cpe->ch[ch].zeroes[w + g])
+ sum = quantize_coeffs(cpe->ch[ch].coeffs + start,
+ cpe->ch[ch].icoefs + start,
+ ics->swb_sizes[g],
+ cpe->ch[ch].sf_idx[w + g]);
+ else
+ memset(cpe->ch[ch].icoefs + start, 0, ics->swb_sizes[g] * sizeof(cpe->ch[0].icoefs[0]));
+ cpe->ch[ch].zeroes[w + g] = !sum;
+ start += ics->swb_sizes[g];
+ }
+ for(cmaxsfb = ics->num_swb; cmaxsfb > 0 && cpe->ch[ch].zeroes[w+cmaxsfb-1]; cmaxsfb--);
+ maxsfb = FFMAX(maxsfb, cmaxsfb);
+ }
+ ics->max_sfb = maxsfb;
+
+ //adjust zero bands for window groups
+ w = 0;
+ for(wg = 0; wg < ics->num_window_groups; wg++){
+ for(g = 0; g < ics->max_sfb; g++){
+ i = 1;
+ for(w2 = 0; w2 < ics->group_len[wg]*16; w2 += 16){
+ if(!cpe->ch[ch].zeroes[w + w2 + g]){
+ i = 0;
+ break;
+ }
+ }
+ cpe->ch[ch].zeroes[w + g] = i;
+ }
+ w += ics->group_len[wg] * 16;
+ }
+ }
+
+ if(chans > 1 && cpe->common_window){
+ IndividualChannelStream *ics0 = &cpe->ch[0].ics;
+ IndividualChannelStream *ics1 = &cpe->ch[1].ics;
+ int msc = 0;
+ ics0->max_sfb = FFMAX(ics0->max_sfb, ics1->max_sfb);
+ ics1->max_sfb = ics0->max_sfb;
+ for(w = 0; w < ics0->num_windows*16; w += 16)
+ for(i = 0; i < ics0->max_sfb; i++)
+ if(cpe->ms_mask[w+i]) msc++;
+ if(msc == 0 || ics0->max_sfb == 0) cpe->ms_mode = 0;
+ else cpe->ms_mode = msc < ics0->max_sfb ? 1 : 2;
+ }
+}
+
+/**
* constants for 3GPP AAC psychoacoustic model
* @{
*/
+#define PSY_3GPP_C1 3.0f // log2(8.0)
+#define PSY_3GPP_C2 1.32192809488736234787f // log2(2.5)
+#define PSY_3GPP_C3 0.55935730170421255071f // 1 - C2/C1
+
#define PSY_3GPP_SPREAD_LOW 1.5f // spreading factor for ascending threshold spreading (15 dB/Bark)
#define PSY_3GPP_SPREAD_HI 3.0f // spreading factor for descending threshold spreading (30 dB/Bark)
+
+#define PSY_3GPP_RPEMIN 0.01f
+#define PSY_3GPP_RPELEV 2.0f
/**
* @}
*/
@@ -83,9 +176,33 @@
typedef struct Psy3gppBand{
float energy; ///< band energy
float ffac; ///< form factor
+ float thr; ///< energy threshold
+ float pe; ///< perceptual entropy
+ float a; ///< constant part in perceptual entropy
+ float b; ///< variable part in perceptual entropy
+ float nl; ///< predicted number of lines left after quantization
+ float min_snr; ///< minimal SNR
+ float thr_quiet; ///< threshold in quiet
}Psy3gppBand;
/**
+ * single/pair channel context for psychoacoustic model
+ */
+typedef struct Psy3gppChannel{
+ float a[2]; ///< parameter used for perceptual entropy - constant part
+ float b[2]; ///< parameter used for perceptual entropy - variable part
+ float pe[2]; ///< channel perceptual entropy
+ float thr[2]; ///< channel thresholds sum
+ Psy3gppBand band[2][128]; ///< bands information
+ Psy3gppBand prev_band[2][128]; ///< bands information from the previous frame
+
+ float win_energy[2]; ///< sliding average of channel energy
+ float iir_state[2][2]; ///< hi-pass IIR filter state
+ uint8_t next_grouping[2]; ///< stored grouping scheme for the next frame (in case of 8 short window sequence)
+ enum WindowSequence next_window_seq[2]; ///< window sequence to be used in the next frame
+}Psy3gppChannel;
+
+/**
* psychoacoustic model frame type-dependent coefficients
*/
typedef struct Psy3gppCoeffs{
@@ -96,9 +213,617 @@
}Psy3gppCoeffs;
/**
+ * 3GPP TS26.403-inspired psychoacoustic model specific data
+ */
+typedef struct Psy3gppContext{
+ Psy3gppCoeffs psy_coef[2];
+ int reservoir; ///< bit reservoir fullness
+ int avg_bits; ///< average frame size of bits for CBR
+ Psy3gppChannel *ch;
+}Psy3gppContext;
+
+/**
* Calculate Bark value for given line.
*/
static inline float calc_bark(float f)
{
return 13.3f * atanf(0.00076f * f) + 3.5f * atanf((f / 7500.0f) * (f / 7500.0f));
}
+
+#define ATH_ADD 4
+/**
+ * Calculate ATH value for given frequency.
+ * Borrowed from Lame.
+ */
+static inline float ath(float f, float add)
+{
+ f /= 1000.0f;
+ return 3.64 * pow(f, -0.8)
+ - 6.8 * exp(-0.6 * (f - 3.4) * (f - 3.4))
+ + 6.0 * exp(-0.15 * (f - 8.7) * (f - 8.7))
+ + (0.6 + 0.04 * add) * 0.001 * f * f * f * f;
+}
+
+static av_cold int psy_3gpp_init(AACPsyContext *apc, int elements)
+{
+ Psy3gppContext *pctx;
+ float barks[1024];
+ int i, j, g, start;
+ float prev, minscale, minath;
+ apc->model_priv_data = av_mallocz(sizeof(Psy3gppContext));
+ pctx = (Psy3gppContext*) apc->model_priv_data;
+
+ for(i = 0; i < 1024; i++)
+ barks[i] = calc_bark(i * apc->avctx->sample_rate / 2048.0);
+ minath = ath(3410, ATH_ADD);
+ for(j = 0; j < 2; j++){
+ Psy3gppCoeffs *coeffs = &pctx->psy_coef[j];
+ int bands = j ? apc->num_bands128 : apc->num_bands1024;
+ i = 0;
+ prev = 0.0;
+ for(g = 0; g < bands; g++){
+ i += j ? apc->bands128[g] : apc->bands1024[g];
+ coeffs->barks[g] = (barks[i - 1] + prev) / 2.0;
+ prev = barks[i - 1];
+ }
+ for(g = 0; g < bands - 1; g++){
+ coeffs->spread_low[g] = pow(10.0, -(coeffs->barks[g+1] - coeffs->barks[g]) * PSY_3GPP_SPREAD_LOW);
+ coeffs->spread_hi [g] = pow(10.0, -(coeffs->barks[g+1] - coeffs->barks[g]) * PSY_3GPP_SPREAD_HI);
+ }
+ start = 0;
+ for(g = 0; g < bands; g++){
+ int size = j ? apc->bands128[g] : apc->bands1024[g];
+ minscale = ath(apc->avctx->sample_rate * start / 1024.0, ATH_ADD);
+ for(i = 1; i < size; i++){
+ minscale = fminf(minscale, ath(apc->avctx->sample_rate * (start + i) / 1024.0 / 2.0, ATH_ADD));
+ }
+ coeffs->ath[g] = minscale - minath;
+ start += size;
+ }
+ }
+
+ pctx->avg_bits = apc->avctx->bit_rate * 1024 / apc->avctx->sample_rate;
+ pctx->ch = av_mallocz(sizeof(Psy3gppChannel) * elements);
+ return 0;
+}
+
+/**
+ * IIR filter used in block switching decision
+ */
+static float iir_filter(int in, float state[2])
+{
+ float ret;
+
+ ret = 0.7548f * (in - state[0]) + 0.5095f * state[1];
+ state[0] = in;
+ state[1] = ret;
+ return ret;
+}
+
+/**
+ * window grouping information stored as bits (0 - new group, 1 - group continues)
+ */
+static const uint8_t window_grouping[9] = {
+ 0xB6, 0x6C, 0xD8, 0xB2, 0x66, 0xC6, 0x96, 0x36, 0x36
+};
+
+/**
+ * Tell encoder which window types to use.
+ * @see 3GPP TS26.403 5.4.1 "Blockswitching"
+ */
+static void psy_3gpp_window(AACPsyContext *apc, int16_t *audio, int16_t *la,
+ int tag, int type, ChannelElement *cpe)
+{
+ int ch;
+ int chans = type == TYPE_CPE ? 2 : 1;
+ int i, j;
+ int br = apc->avctx->bit_rate / apc->avctx->channels;
+ int attack_ratio = (br <= 16000 + 8000*chans) ? 18 : 10;
+ Psy3gppContext *pctx = (Psy3gppContext*) apc->model_priv_data;
+ Psy3gppChannel *pch = &pctx->ch[tag];
+ uint8_t grouping[2];
+ enum WindowSequence win[2];
+ IndividualChannelStream *ics0 = &cpe->ch[0].ics, *ics1 = &cpe->ch[1].ics;
+
+ if(la && !(apc->flags & PSY_MODEL_NO_SWITCH)){
+ float s[8], v;
+ for(ch = 0; ch < chans; ch++){
+ enum WindowSequence last_window_sequence = cpe->ch[ch].ics.window_sequence[0];
+ int switch_to_eight = 0;
+ float sum = 0.0, sum2 = 0.0;
+ int attack_n = 0;
+ for(i = 0; i < 8; i++){
+ for(j = 0; j < 128; j++){
+ v = iir_filter(audio[(i*128+j)*apc->avctx->channels+ch], pch->iir_state[ch]);
+ sum += v*v;
+ }
+ s[i] = sum;
+ sum2 += sum;
+ }
+ for(i = 0; i < 8; i++){
+ if(s[i] > pch->win_energy[ch] * attack_ratio){
+ attack_n = i + 1;
+ switch_to_eight = 1;
+ break;
+ }
+ }
+ pch->win_energy[ch] = pch->win_energy[ch]*7/8 + sum2/64;
+
+ switch(last_window_sequence){
+ case ONLY_LONG_SEQUENCE:
+ win[ch] = switch_to_eight ? LONG_START_SEQUENCE : ONLY_LONG_SEQUENCE;
+ grouping[ch] = 0;
+ break;
+ case LONG_START_SEQUENCE:
+ win[ch] = EIGHT_SHORT_SEQUENCE;
+ grouping[ch] = pch->next_grouping[ch];
+ break;
+ case LONG_STOP_SEQUENCE:
+ win[ch] = ONLY_LONG_SEQUENCE;
+ grouping[ch] = 0;
+ break;
+ case EIGHT_SHORT_SEQUENCE:
+ win[ch] = switch_to_eight ? EIGHT_SHORT_SEQUENCE : LONG_STOP_SEQUENCE;
+ grouping[ch] = switch_to_eight ? pch->next_grouping[ch] : 0;
+ break;
+ }
+ pch->next_grouping[ch] = window_grouping[attack_n];
+ }
+ }else{
+ for(ch = 0; ch < chans; ch++){
+ IndividualChannelStream *ics = &cpe->ch[ch].ics;
+ win[ch] = (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE)
+ ? EIGHT_SHORT_SEQUENCE
+ : ONLY_LONG_SEQUENCE;
+ grouping[ch] = (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) ? window_grouping[0] : 0;
+ }
+ }
+
+ for(ch = 0; ch < chans; ch++){
+ IndividualChannelStream *ics = &cpe->ch[ch].ics;
+ ics->window_sequence[0] = win[ch];
+ ics->use_kb_window[0] = 1;
+ if(win[ch] != EIGHT_SHORT_SEQUENCE){
+ ics->num_windows = 1;
+ ics->swb_sizes = apc->bands1024;
+ ics->num_swb = apc->num_bands1024;
+ ics->num_window_groups = 1;
+ ics->group_len[0] = 1;
+ }else{
+ ics->num_windows = 8;
+ ics->swb_sizes = apc->bands128;
+ ics->num_swb = apc->num_bands128;
+ ics->num_window_groups = 0;
+ ics->group_len[0] = 1;
+ for(i = 0; i < 8; i++){
+ if((grouping[ch] >> i) & 1){
+ ics->group_len[ics->num_window_groups - 1]++;
+ }else{
+ ics->num_window_groups++;
+ ics->group_len[ics->num_window_groups - 1] = 1;
+ }
+ }
+ }
+ }
+ cpe->common_window = 0;
+ if(chans > 1
+ && ics0->window_sequence[0] == ics1->window_sequence[0]
+ && ics0->use_kb_window[0] == ics1->use_kb_window[0]
+ && !(ics0->window_sequence[0] == EIGHT_SHORT_SEQUENCE && grouping[0] != grouping[1]))
+ cpe->common_window = 1;
+ if(PSY_MODEL_MODE(apc->flags) > PSY_MODE_QUALITY){
+ av_log(apc->avctx, AV_LOG_ERROR, "Unknown mode %d, defaulting to CBR\n", PSY_MODEL_MODE(apc->flags));
+ }
+}
+
+/**
+ * Modify threshold by adding some value in loudness domain.
+ * @see 3GPP TS26.403 5.6.1.1.1 "Addition of noise with equal loudness"
+ */
+static inline float modify_thr(float thr, float r){
+ float t;
+ t = pow(thr, 0.25) + r;
+ return (t*t)*(t*t);
+}
+
+/**
+ * Calculate perceptual entropy and its corresponding values for one band.
+ * @see 3GPP TS26.403 5.6.1.3 "Calculation of the reduction value"
+ */
+static void calc_pe(Psy3gppBand *band, int band_width)
+{
+ if(band->energy <= band->thr){
+ band->a = 0.0f;
+ band->b = 0.0f;
+ band->nl = 0.0f;
+ return;
+ }
+ band->nl = band->ffac / pow(band->energy/band_width, 0.25);
+ if(band->energy >= band->thr * 8.0){
+ band->a = band->nl * log2(band->energy);
+ band->b = band->nl;
+ }else{
+ band->a = band->nl * (PSY_3GPP_C2 + PSY_3GPP_C3 * log2(band->energy));
+ band->b = band->nl * PSY_3GPP_C3;
+ }
+ band->pe = band->a - band->b * log2(band->thr);
+ band->min_snr = 1.0 / (pow(2.0, band->pe / band_width) - 1.5);
+ band->min_snr = av_clipf(band->min_snr, 1.26f, 316.2277f);
+}
+
+/**
+ * Determine scalefactor from band threshold and form factor.
+ * @see 3GPP TS26.403 5.4 5.6.2 "Scalefactor determination"
+ */
+static inline int determine_scalefactor(Psy3gppBand *band)
+{
+ //spec gives constant for lg() but we scaled it for log2()
+ return (int)(2.66667 * log2(6.75*band->thr/band->ffac));
+}
+
+/**
+ * Determine scalefactors and prepare coefficients for encoding.
+ * @see 3GPP TS26.403 5.4 "Psychoacoustic model"
+ */
+static void psy_3gpp_process(AACPsyContext *apc, int tag, int type, ChannelElement *cpe)
+{
+ int start;
+ int ch, w, wg, g, i;
+ int prev_scale;
+ Psy3gppContext *pctx = (Psy3gppContext*) apc->model_priv_data;
+ float pe_target;
+ int bits_avail;
+ int chans = type == TYPE_CPE ? 2 : 1;
+ Psy3gppChannel *pch = &pctx->ch[tag];
+
+ //calculate energies, initial thresholds and related values - 5.4.2 "Threshold Calculation"
+ memset(pch->band, 0, sizeof(pch->band));
+ for(ch = 0; ch < chans; ch++){
+ IndividualChannelStream *ics = &cpe->ch[ch].ics;
+ start = 0;
+ for(w = 0; w < ics->num_windows*16; w += 16){
+ for(g = 0; g < ics->num_swb; g++){
+ Psy3gppBand *band = &pch->band[ch][w+g];
+ for(i = 0; i < ics->swb_sizes[g]; i++)
+ band->energy += cpe->ch[ch].coeffs[start+i] * cpe->ch[ch].coeffs[start+i];
+ band->energy *= 1.0f / (512*512);
+ band->thr = band->energy * 0.001258925f;
+ start += ics->swb_sizes[g];
+ if(band->energy != 0.0){
+ float ffac = 0.0;
+
+ for(i = 0; i < ics->swb_sizes[g]; i++)
+ ffac += sqrt(FFABS(cpe->ch[ch].coeffs[start+i]));
+ band->ffac = ffac / sqrt(512.0);
+ }
+ }
+ }
+ }
+
+ //modify thresholds - spread, threshold in quiet - 5.4.3 "Spreaded Energy Calculation"
+ for(ch = 0; ch < chans; ch++){
+ IndividualChannelStream *ics = &cpe->ch[ch].ics;
+ Psy3gppCoeffs *coeffs = &pctx->psy_coef[ics->num_windows == 8];
+ for(w = 0; w < ics->num_windows*16; w += 16){
+ Psy3gppBand *band = &pch->band[ch][w];
+ for(g = 1; g < ics->num_swb; g++){
+ band[g].thr = FFMAX(band[g].thr, band[g-1].thr * coeffs->spread_low[g-1]);
+ }
+ for(g = ics->num_swb - 2; g >= 0; g--){
+ band[g].thr = FFMAX(band[g].thr, band[g+1].thr * coeffs->spread_hi [g+1]);
+ }
+ for(g = 0; g < ics->num_swb; g++){
+ band[g].thr_quiet = FFMAX(band[g].thr, coeffs->ath[g]);
+ band[g].thr_quiet = fmaxf(PSY_3GPP_RPEMIN*band[g].thr_quiet,
+ fminf(band[g].thr_quiet,
+ PSY_3GPP_RPELEV*pch->prev_band[ch][w+g].thr_quiet));
+ band[g].thr = FFMAX(band[g].thr, band[g].thr_quiet * 0.25);
+ }
+ }
+ }
+
+ // M/S detection - 5.5.2 "Mid/Side Stereo"
+ if(chans > 1 && cpe->common_window){
+ start = 0;
+ for(w = 0; w < cpe->ch[0].ics.num_windows*16; w += 16){
+ for(g = 0; g < cpe->ch[0].ics.num_swb; g++){
+ Psy3gppBand *band0 = &pch->band[0][w+g];
+ Psy3gppBand *band1 = &pch->band[1][w+g];
+ double en_m = 0.0, en_s = 0.0, ff_m = 0.0, ff_s = 0.0, minthr;
+ float m, s;
+
+ cpe->ms_mask[w+g] = 0;
+ if(band0->energy == 0.0 || band1->energy == 0.0)
+ continue;
+ for(i = 0; i < cpe->ch[0].ics.swb_sizes[g]; i++){
+ m = cpe->ch[0].coeffs[start+i] + cpe->ch[1].coeffs[start+i];
+ s = cpe->ch[0].coeffs[start+i] - cpe->ch[1].coeffs[start+i];
+ en_m += m*m;
+ en_s += s*s;
+ }
+ en_m *= 1.0f / (512*512*4);
+ en_s *= 1.0f / (512*512*4);
+ minthr = FFMIN(band0->thr, band1->thr);
+ if(minthr * minthr * band0->energy * band1->energy >= band0->thr * band1->thr * en_m * en_s){
+ cpe->ms_mask[w+g] = 1;
+ band0->energy = en_m;
+ band1->energy = en_s;
+ band0->thr = en_m * 0.001258925f;
+ band1->thr = en_s * 0.001258925f;
+ for(i = 0; i < cpe->ch[0].ics.swb_sizes[g]; i++){
+ m = cpe->ch[0].coeffs[start+i] + cpe->ch[1].coeffs[start+i];
+ s = cpe->ch[0].coeffs[start+i] - cpe->ch[1].coeffs[start+i];
+ ff_m += sqrt(fabs(m));
+ ff_s += sqrt(fabs(s));
+ }
+ band0->ffac = ff_m * (1.0f / 32.0f); // sqrt(512)*sqrt(2)
+ band1->ffac = ff_s * (1.0f / 32.0f);
+ }
+ }
+ }
+ }
+
+ for(ch = 0; ch < chans; ch++){
+ IndividualChannelStream *ics = &cpe->ch[ch].ics;
+ pch->a[ch] = pch->b[ch] = pch->pe[ch] = pch->thr[ch] = 0.0f;
+ for(w = 0; w < ics->num_windows*16; w += 16){
+ for(g = 0; g < ics->num_swb; g++){
+ Psy3gppBand *band = &pch->band[ch][w+g];
+ if(band->energy != 0.0)
+ calc_pe(band, ics->swb_sizes[g]);
+ if(band->thr < band->energy){
+ pch->a[ch] += band->a;
+ pch->b[ch] += band->b;
+ pch->pe[ch] += band->pe;
+ pch->thr[ch] += band->thr;
+ }
+ }
+ }
+ }
+
+ switch(PSY_MODEL_MODE(apc->flags)){
+ case PSY_MODE_CBR:
+ case PSY_MODE_ABR:
+ //bitrate reduction - 5.6.1 "Reduction of psychoacoustic requirements"
+ if(PSY_MODEL_MODE(apc->flags) != PSY_MODE_ABR){
+ pctx->reservoir += pctx->avg_bits - apc->avctx->frame_bits;
+ bits_avail = pctx->avg_bits + pctx->reservoir;
+ bits_avail = FFMIN(bits_avail, pctx->avg_bits * 1.5);
+ pe_target = 1.18f * bits_avail / apc->avctx->channels * chans;
+ }else{
+ pe_target = pctx->avg_bits / apc->avctx->channels * chans;
+ }
+ for(i = 0; i < 2; i++){
+ float t0, pe, r, a0 = 0.0f, pe0 = 0.0f, b0 = 0.0f;
+ for(ch = 0; ch < chans; ch++){
+ a0 += pch->a[ch];
+ b0 += pch->b[ch];
+ pe0 += pch->pe[ch];
+ }
+ if(pe0 == 0.0f) break;
+ t0 = pow(2.0, (a0 - pe0) / (4.0 * b0));
+ r = pow(2.0, (a0 - pe_target) / (4.0 * b0)) - t0;
+
+ //add correction factor to thresholds and recalculate perceptual entropy
+ for(ch = 0; ch < chans; ch++){
+ IndividualChannelStream *ics = &cpe->ch[ch].ics;
+ pch->a[ch] = pch->b[ch] = pch->pe[ch] = pch->thr[ch] = 0.0;
+ pe = 0.0f;
+ for(w = 0; w < ics->num_windows*16; w += 16){
+ for(g = 0; g < ics->num_swb; g++){
+ Psy3gppBand *band = &pch->band[ch][w+g];
+ band->thr = modify_thr(band->thr, r);
+ calc_pe(band, ics->swb_sizes[g]);
+ if(band->thr < band->energy){
+ pch->a[ch] += band->a;
+ pch->b[ch] += band->b;
+ pch->pe[ch] += band->pe;
+ pch->thr[ch] += band->thr;
+ }
+ }
+ }
+ }
+ }
+
+ //determine scalefactors - 5.6.2 "Scalefactor determination"
+ for(ch = 0; ch < chans; ch++){
+ IndividualChannelStream *ics = &cpe->ch[ch].ics;
+ for(w = 0; w < ics->num_windows*16; w += 16){
+ for(g = 0; g < ics->num_swb; g++){
+ Psy3gppBand *band = &pch->band[ch][w+g];
+ cpe->ch[ch].zeroes[w+g] = band->thr >= band->energy;
+ if(cpe->ch[ch].zeroes[w+g]) continue;
+ cpe->ch[ch].sf_idx[w+g] = determine_scalefactor(band);
+ }
+ }
+ }
+ break;
+ case PSY_MODE_QUALITY:
+ for(ch = 0; ch < chans; ch++){
+ IndividualChannelStream *ics = &cpe->ch[ch].ics;
+ start = 0;
+ for(w = 0; w < ics->num_windows*16; w += 16){
+ for(g = 0; g < ics->num_swb; g++){
+ Psy3gppBand *band = &pch->band[ch][w+g];
+ if(band->thr >= band->energy){
+ cpe->ch[ch].sf_idx[w+g] = 0;
+ cpe->ch[ch].zeroes[w+g] = 1;
+ }else{
+ cpe->ch[ch].zeroes[w+g] = 0;
+ cpe->ch[ch].sf_idx[w+g] = determine_scalefactor(band);
+ while(cpe->ch[ch].sf_idx[w+g] > 3){
+ float dist = get_approximate_quant_error(cpe->ch[ch].coeffs + start,
+ ics->swb_sizes[g],
+ SCALE_ONE_POS + cpe->ch[ch].sf_idx[w+g]);
+ if(dist < band->thr) break;
+ cpe->ch[ch].sf_idx[w+g] -= 3;
+ }
+ }
+ start += ics->swb_sizes[g];
+ }
+ }
+ }
+ break;
+ }
+
+ //limit scalefactors
+ for(ch = 0; ch < chans; ch++){
+ int min_scale = 256;
+ IndividualChannelStream *ics = &cpe->ch[ch].ics;
+ for(w = 0; w < ics->num_windows*16; w += 16)
+ for(g = 0; g < ics->num_swb; g++){
+ if(cpe->ch[ch].zeroes[w + g]) continue;
+ min_scale = FFMIN(min_scale, cpe->ch[ch].sf_idx[w + g]);
+ }
+ for(w = 0; w < ics->num_windows*16; w += 16)
+ for(g = 0; g < ics->num_swb; g++){
+ if(cpe->ch[ch].zeroes[w + g]) continue;
+ cpe->ch[ch].sf_idx[w + g] = FFMIN(cpe->ch[ch].sf_idx[w + g], min_scale + SCALE_MAX_DIFF);
+ }
+ for(w = 0; w < ics->num_windows*16; w += 16)
+ for(g = 0; g < ics->num_swb; g++){
+ if(cpe->ch[ch].zeroes[w + g])
+ cpe->ch[ch].sf_idx[w + g] = 256;
+ else
+ cpe->ch[ch].sf_idx[w + g] = av_clip(SCALE_ONE_POS + cpe->ch[ch].sf_idx[w + g],
+ 0,
+ SCALE_MAX_POS);
+ }
+
+ //adjust scalefactors for window groups
+ w = 0;
+ for(wg = 0; wg < ics->num_window_groups; wg++){
+ int min_scale = 256;
+
+ for(g = 0; g < ics->num_swb; g++){
+ for(i = w; i < w + ics->group_len[wg]*16; i += 16){
+ if(cpe->ch[ch].zeroes[i + g]) continue;
+ min_scale = FFMIN(min_scale, cpe->ch[ch].sf_idx[i + g]);
+ }
+ for(i = w; i < w + ics->group_len[wg]*16; i += 16)
+ cpe->ch[ch].sf_idx[i + g] = min_scale;
+ }
+ w += ics->group_len[wg] * 16;
+ }
+ }
+
+ memcpy(pch->prev_band, pch->band, sizeof(pch->band));
+ psy_create_output(apc, cpe, chans);
+}
+
+static av_cold void psy_3gpp_end(AACPsyContext *apc)
+{
+ Psy3gppContext *pctx = (Psy3gppContext*) apc->model_priv_data;
+ av_freep(&pctx->ch);
+ av_freep(&apc->model_priv_data);
+}
+
+static const AACPsyModel psy_models[AAC_NB_PSY_MODELS] =
+{
+ {
+ "3GPP TS 26.403-inspired model",
+ psy_3gpp_init,
+ psy_3gpp_window,
+ psy_3gpp_process,
+ psy_3gpp_end,
+ },
+};
+
+int av_cold ff_aac_psy_init(AACPsyContext *ctx, AVCodecContext *avctx,
+ enum AACPsyModelType model, int elements, int flags,
+ const uint8_t *bands1024, int num_bands1024,
+ const uint8_t *bands128, int num_bands128)
+{
+ int i;
+
+ if(model < 0 || model >= AAC_NB_PSY_MODELS){
+ av_log(avctx, AV_LOG_ERROR, "Invalid psy model\n");
+ return -1;
+ }
+
+#ifndef CONFIG_HARDCODED_TABLES
+ for (i = 0; i < 316; i++)
+ ff_aac_pow2sf_tab[i] = pow(2, (i - 200)/4.);
+#endif /* CONFIG_HARDCODED_TABLES */
+
+ ctx->avctx = avctx;
+ ctx->flags = flags;
+ ctx->bands1024 = bands1024;
+ ctx->num_bands1024 = num_bands1024;
+ ctx->bands128 = bands128;
+ ctx->num_bands128 = num_bands128;
+ ctx->model = &psy_models[model];
+
+ if(ctx->flags & PSY_MODEL_NO_ST_ATT || PSY_MODEL_MODE(ctx->flags) == PSY_MODE_QUALITY){
+ ctx->flags |= PSY_MODEL_NO_ST_ATT;
+ ctx->stereo_att = 0.5f;
+ }else{
+ ctx->stereo_att = av_clipf(avctx->bit_rate / elements / 192000.0, 0.0f, 0.5f);
+ }
+ if(ctx->flags & PSY_MODEL_NO_LOWPASS || PSY_MODEL_MODE(ctx->flags) == PSY_MODE_QUALITY){
+ ctx->flags |= PSY_MODEL_NO_LOWPASS;
+ }else{
+ int cutoff;
+ cutoff = avctx->bit_rate / elements / 8;
+ if(ff_lowpass_filter_init_coeffs(&ctx->lp_coeffs, avctx->sample_rate/2, cutoff) < 0){
+ ctx->flags |= PSY_MODEL_NO_LOWPASS;
+ }else{
+ ctx->lp_state = av_mallocz(sizeof(LPFilterState) * elements * 2);
+ }
+ }
+ if(ctx->model->init)
+ return ctx->model->init(ctx, elements);
+ return 0;
+}
+
+void ff_aac_psy_suggest_window(AACPsyContext *ctx, int16_t *audio, int16_t *la,
+ int tag, int type, ChannelElement *cpe)
+{
+ ctx->model->window(ctx, audio, la, tag, type, cpe);
+}
+
+void ff_aac_psy_analyze(AACPsyContext *ctx, int tag, int type, ChannelElement *cpe)
+{
+ ctx->model->process(ctx, tag, type, cpe);
+}
+
+void av_cold ff_aac_psy_end(AACPsyContext *ctx)
+{
+ av_freep(&ctx->lp_state);
+ if(ctx->model->end)
+ return ctx->model->end(ctx);
+}
+
+void ff_aac_psy_preprocess(AACPsyContext *ctx, int16_t *audio, int16_t *dest, int tag, int type)
+{
+ int chans = type == TYPE_CPE ? 2 : 1;
+ const int chstride = ctx->avctx->channels;
+ int i, ch;
+ float t[2];
+
+ if(chans == 1){
+ for(ch = 0; ch < chans; ch++){
+ for(i = 0; i < 1024; i++){
+ dest[i * chstride + ch] = audio[i * chstride + ch];
+ }
+ }
+ }else{
+ for(i = 0; i < 1024; i++){
+ if(ctx->flags & PSY_MODEL_NO_ST_ATT){
+ for(ch = 0; ch < 2; ch++)
+ t[ch] = audio[i * chstride + ch];
+ }else{
+ t[0] = audio[i * chstride + 0] * (0.5 + ctx->stereo_att) + audio[i * chstride + 1] * (0.5 - ctx->stereo_att);
+ t[1] = audio[i * chstride + 0] * (0.5 - ctx->stereo_att) + audio[i * chstride + 1] * (0.5 + ctx->stereo_att);
+ }
+ if(!(ctx->flags & PSY_MODEL_NO_LOWPASS)){
+ LPFilterState *is = (LPFilterState*)ctx->lp_state + tag*2;
+ for(ch = 0; ch < 2; ch++)
+ t[ch] = ff_lowpass_filter(&ctx->lp_coeffs, is + ch, t[ch]);
+ }
+ for(ch = 0; ch < 2; ch++)
+ dest[i * chstride + ch] = av_clip_int16(t[ch]);
+ }
+ }
+}
+
Index: aacpsy.h
===================================================================
--- aacpsy.h (revision 14920)
+++ aacpsy.h (working copy)
@@ -27,20 +27,109 @@
#include "lowpass.h"
enum AACPsyModelType{
- AAC_PSY_TEST, ///< a sample model to exercise encoder
AAC_PSY_3GPP, ///< model following recommendations from 3GPP TS 26.403
AAC_NB_PSY_MODELS ///< total number of psychoacoustic models, since it's not a part of the ABI new models can be added freely
};
+enum AACPsyModelMode{
+ PSY_MODE_CBR, ///< follow bitrate as closely as possible
+ PSY_MODE_ABR, ///< try to achieve bitrate but actual bitrate may differ significantly
+ PSY_MODE_QUALITY, ///< try to achieve set quality instead of bitrate
+};
+
+#define PSY_MODEL_MODE_MASK 0x0000000F ///< bit fields for storing mode (CBR, ABR, VBR)
+#define PSY_MODEL_NO_SWITCH 0x00000020 ///< disable window switching
+#define PSY_MODEL_NO_ST_ATT 0x00000040 ///< disable stereo attenuation
+#define PSY_MODEL_NO_LOWPASS 0x00000080 ///< disable low-pass filtering
+
+#define PSY_MODEL_NO_PREPROC (PSY_MODEL_NO_ST_ATT | PSY_MODEL_NO_LOWPASS)
+
+#define PSY_MODEL_MODE(a) ((a) & PSY_MODEL_MODE_MASK)
+
/**
* context used by psychoacoustic model
*/
typedef struct AACPsyContext {
AVCodecContext *avctx; ///< encoder context
+
+ int flags; ///< model flags
+ const uint8_t *bands1024; ///< scalefactor band sizes for long (1024 samples) frame
+ int num_bands1024; ///< number of scalefactor bands for long frame
+ const uint8_t *bands128; ///< scalefactor band sizes for short (128 samples) frame
+ int num_bands128; ///< number of scalefactor bands for short frame
+
+ const struct AACPsyModel *model; ///< pointer to the psychoacoustic model implementation
+ void* model_priv_data; ///< psychoacoustic model implementation private data
+
+ float stereo_att; ///< stereo attenuation factor
+ LPFilterCoeffs lp_coeffs; ///< lowpass filter coefficients
+ LPFilterState *lp_state; ///< lowpass filter state
}AACPsyContext;
+typedef struct AACPsyModel {
+ const char *name;
+ int (*init) (AACPsyContext *apc, int elements);
+ void (*window) (AACPsyContext *apc, int16_t *audio, int16_t *la, int tag, int type, ChannelElement *cpe);
+ void (*process)(AACPsyContext *apc, int tag, int type, ChannelElement *cpe);
+ void (*end) (AACPsyContext *apc);
+}AACPsyModel;
+
/**
+ * Initialize psychoacoustic model.
+ *
+ * @param ctx model context
+ * @param avctx codec context
+ * @param model model implementation that will be used
+ * @param elements number of channel elements (single channel or channel pair) to handle by model
+ * @param flags model flags, may be ignored by model if unsupported
+ * @param bands1024 scalefactor band lengths for long (1024 samples) frame
+ * @param num_bands1024 number of scalefactor bands for long frame
+ * @param bands128 scalefactor band lengths for short (128 samples) frame
+ * @param num_bands128 number of scalefactor bands for short frame
+ *
+ * @return zero if successful, a negative value if not
+ */
+int ff_aac_psy_init(AACPsyContext *ctx, AVCodecContext *avctx,
+ enum AACPsyModelType model, int elements, int flags,
+ const uint8_t *bands1024, int num_bands1024,
+ const uint8_t *bands128, int num_bands128);
+
+/**
+ * Preprocess audio frame in order to compress it better.
+ *
+ * @param ctx model context
+ * @param audio samples to preprocess
+ * @param dest place to put filtered samples
+ * @param tag number of channel element to analyze
+ * @param type channel element type (e.g. ID_SCE or ID_CPE)
+ */
+void ff_aac_psy_preprocess(AACPsyContext *ctx, int16_t *audio, int16_t *dest, int tag, int type);
+
+/**
+ * Set window sequence and related parameters for channel element.
+ *
+ * @param ctx model context
+ * @param audio samples for the current frame
+ * @param la lookahead samples (NULL when unavailable)
+ * @param tag number of channel element to analyze
+ * @param type channel element type (e.g. ID_SCE or ID_CPE)
+ * @param cpe pointer to the current channel element
+ */
+void ff_aac_psy_suggest_window(AACPsyContext *ctx, int16_t *audio, int16_t *la, int tag, int type, ChannelElement *cpe);
+
+/**
+ * Perform psychoacoustic analysis and output coefficients in integer form
+ * along with scalefactors, M/S flags, etc.
+ *
+ * @param ctx model context
+ * @param tag number of channel element to analyze
+ * @param type channel element type (e.g. ID_SCE or ID_CPE)
+ * @param cpe pointer to the current channel element
+ */
+void ff_aac_psy_analyze(AACPsyContext *ctx, int tag, int type, ChannelElement *cpe);
+
+/**
* Cleanup model context at the end.
*
* @param ctx model context
More information about the ffmpeg-devel
mailing list