aacdec: Add support for LD (Low Delay) AAC

Changelog

@@ -35,6 +35,7 @@ version 10:
 - incomplete Voxware MetaSound decoder
 - WebP decoder
 - Error Resilient AAC syntax (ER AAC LC) decoding
+- Low Delay AAC (ER AAC LD) decoding
 
 
 version 9:

libavcodec/aac.h

@@ -289,6 +289,7 @@ typedef struct AACContext {
      */
     FFTContext mdct;
     FFTContext mdct_small;
+    FFTContext mdct_ld;
     FFTContext mdct_ltp;
     FmtConvertContext fmt_conv;
     AVFloatDSPContext fdsp;

libavcodec/aacdec.c

@@ -817,6 +817,13 @@ static int decode_audio_specific_config(AACContext *ac,
                m4ac->sampling_index);
         return AVERROR_INVALIDDATA;
     }
+    if (m4ac->object_type == AOT_ER_AAC_LD &&
+        (m4ac->sampling_index < 3 || m4ac->sampling_index > 7)) {
+        av_log(avctx, AV_LOG_ERROR,
+               "invalid low delay sampling rate index %d\n",
+               m4ac->sampling_index);
+        return AVERROR_INVALIDDATA;
+    }
 
     skip_bits_long(&gb, i);
 
@@ -825,6 +832,7 @@ static int decode_audio_specific_config(AACContext *ac,
     case AOT_AAC_LC:
     case AOT_AAC_LTP:
     case AOT_ER_AAC_LC:
+    case AOT_ER_AAC_LD:
         if ((ret = decode_ga_specific_config(ac, avctx, &gb,
                                             m4ac, m4ac->chan_config)) < 0)
             return ret;
@@ -985,12 +993,15 @@ static av_cold int aac_decode_init(AVCodecContext *avctx)
                     352);
 
     ff_mdct_init(&ac->mdct,       11, 1, 1.0 / (32768.0 * 1024.0));
+    ff_mdct_init(&ac->mdct_ld,    10, 1, 1.0 / (32768.0 * 512.0));
     ff_mdct_init(&ac->mdct_small,  8, 1, 1.0 / (32768.0 * 128.0));
     ff_mdct_init(&ac->mdct_ltp,   11, 0, -2.0 * 32768.0);
     // window initialization
     ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
+    ff_kbd_window_init(ff_aac_kbd_long_512,  4.0, 512);
     ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128);
     ff_init_ff_sine_windows(10);
+    ff_init_ff_sine_windows( 9);
     ff_init_ff_sine_windows( 7);
 
     cbrt_tableinit();
@@ -1063,6 +1074,14 @@ static int decode_ics_info(AACContext *ac, IndividualChannelStream *ics,
     }
     ics->window_sequence[1] = ics->window_sequence[0];
     ics->window_sequence[0] = get_bits(gb, 2);
+    if (ac->oc[1].m4ac.object_type == AOT_ER_AAC_LD &&
+        ics->window_sequence[0] != ONLY_LONG_SEQUENCE) {
+        av_log(ac->avctx, AV_LOG_ERROR,
+               "AAC LD is only defined for ONLY_LONG_SEQUENCE but "
+               "window sequence %d found.\n", ics->window_sequence[0]);
+        ics->window_sequence[0] = ONLY_LONG_SEQUENCE;
+        return AVERROR_INVALIDDATA;
+    }
     ics->use_kb_window[1]   = ics->use_kb_window[0];
     ics->use_kb_window[0]   = get_bits1(gb);
     ics->num_window_groups  = 1;
@@ -1086,8 +1105,15 @@ static int decode_ics_info(AACContext *ac, IndividualChannelStream *ics,
     } else {
         ics->max_sfb               = get_bits(gb, 6);
         ics->num_windows           = 1;
-        ics->swb_offset            =    ff_swb_offset_1024[ac->oc[1].m4ac.sampling_index];
+        if (ac->oc[1].m4ac.object_type == AOT_ER_AAC_LD) {
-        ics->num_swb               =   ff_aac_num_swb_1024[ac->oc[1].m4ac.sampling_index];
+            ics->swb_offset        =     ff_swb_offset_512[ac->oc[1].m4ac.sampling_index];
+            ics->num_swb           =    ff_aac_num_swb_512[ac->oc[1].m4ac.sampling_index];
+            if (!ics->num_swb || !ics->swb_offset)
+                return AVERROR_BUG;
+        } else {
+            ics->swb_offset        =    ff_swb_offset_1024[ac->oc[1].m4ac.sampling_index];
+            ics->num_swb           =   ff_aac_num_swb_1024[ac->oc[1].m4ac.sampling_index];
+        }
         ics->tns_max_bands         = ff_tns_max_bands_1024[ac->oc[1].m4ac.sampling_index];
         ics->predictor_present     = get_bits1(gb);
         ics->predictor_reset_group = 0;
@@ -1102,6 +1128,11 @@ static int decode_ics_info(AACContext *ac, IndividualChannelStream *ics,
                        "Prediction is not allowed in AAC-LC.\n");
                 return AVERROR_INVALIDDATA;
             } else {
+                if (ac->oc[1].m4ac.object_type == AOT_ER_AAC_LD) {
+                    av_log(ac->avctx, AV_LOG_ERROR,
+                           "LTP in ER AAC LD not yet implemented.\n");
+                    return AVERROR_PATCHWELCOME;
+                }
                 if ((ics->ltp.present = get_bits(gb, 1)))
                     decode_ltp(&ics->ltp, gb, ics->max_sfb);
             }
@@ -2314,6 +2345,25 @@ static void imdct_and_windowing(AACContext *ac, SingleChannelElement *sce)
     }
 }
 
+static void imdct_and_windowing_ld(AACContext *ac, SingleChannelElement *sce)
+{
+    IndividualChannelStream *ics = &sce->ics;
+    float *in    = sce->coeffs;
+    float *out   = sce->ret;
+    float *saved = sce->saved;
+    const float *lwindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_long_512 : ff_sine_512;
+    float *buf  = ac->buf_mdct;
+
+    // imdct
+    ac->mdct.imdct_half(&ac->mdct_ld, buf, in);
+
+    // window overlapping
+    ac->fdsp.vector_fmul_window(out, saved, buf, lwindow_prev, 256);
+
+    // buffer update
+    memcpy(saved, buf + 256, 256 * sizeof(float));
+}
+
 /**
  * Apply dependent channel coupling (applied before IMDCT).
  *
@@ -2410,6 +2460,11 @@ static void apply_channel_coupling(AACContext *ac, ChannelElement *cc,
 static void spectral_to_sample(AACContext *ac)
 {
     int i, type;
+    void (*imdct_and_window)(AACContext *ac, SingleChannelElement *sce);
+    if (ac->oc[1].m4ac.object_type == AOT_ER_AAC_LD)
+        imdct_and_window = imdct_and_windowing_ld;
+    else
+        imdct_and_window = imdct_and_windowing;
     for (type = 3; type >= 0; type--) {
         for (i = 0; i < MAX_ELEM_ID; i++) {
             ChannelElement *che = ac->che[type][i];
@@ -2431,11 +2486,11 @@ static void spectral_to_sample(AACContext *ac)
                 if (type <= TYPE_CPE)
                     apply_channel_coupling(ac, che, type, i, BETWEEN_TNS_AND_IMDCT, apply_dependent_coupling);
                 if (type != TYPE_CCE || che->coup.coupling_point == AFTER_IMDCT) {
-                    imdct_and_windowing(ac, &che->ch[0]);
+                    imdct_and_window(ac, &che->ch[0]);
                     if (ac->oc[1].m4ac.object_type == AOT_AAC_LTP)
                         update_ltp(ac, &che->ch[0]);
                     if (type == TYPE_CPE) {
-                        imdct_and_windowing(ac, &che->ch[1]);
+                        imdct_and_window(ac, &che->ch[1]);
                         if (ac->oc[1].m4ac.object_type == AOT_AAC_LTP)
                             update_ltp(ac, &che->ch[1]);
                     }
@@ -2503,6 +2558,9 @@ static int aac_decode_er_frame(AVCodecContext *avctx, void *data,
     int samples = 1024;
     int chan_config = ac->oc[1].m4ac.chan_config;
 
+    if (ac->oc[1].m4ac.object_type == AOT_ER_AAC_LD)
+        samples >>= 1;
+
     ac->frame = data;
 
     if ((err = frame_configure_elements(avctx)) < 0)
@@ -2757,6 +2815,7 @@ static av_cold int aac_decode_close(AVCodecContext *avctx)
 
     ff_mdct_end(&ac->mdct);
     ff_mdct_end(&ac->mdct_small);
+    ff_mdct_end(&ac->mdct_ld);
     ff_mdct_end(&ac->mdct_ltp);
     return 0;
 }

libavcodec/aactab.c

@@ -34,12 +34,17 @@
 #include <stdint.h>
 
 DECLARE_ALIGNED(32, float,  ff_aac_kbd_long_1024)[1024];
+DECLARE_ALIGNED(32, float,  ff_aac_kbd_long_512 )[512];
 DECLARE_ALIGNED(32, float,  ff_aac_kbd_short_128)[128];
 
 const uint8_t ff_aac_num_swb_1024[] = {
     41, 41, 47, 49, 49, 51, 47, 47, 43, 43, 43, 40, 40
 };
 
+const uint8_t ff_aac_num_swb_512[] = {
+     0,  0,  0, 36, 36, 37, 31, 31,  0,  0,  0,  0,  0
+};
+
 const uint8_t ff_aac_num_swb_128[] = {
     12, 12, 12, 14, 14, 14, 15, 15, 15, 15, 15, 15, 15
 };
@@ -1114,6 +1119,14 @@ static const uint16_t swb_offset_1024_48[] = {
     928, 1024
 };
 
+static const uint16_t swb_offset_512_48[] = {
+      0,   4,   8,  12,  16,  20,  24,  28,
+     32,  36,  40,  44,  48,  52,  56,  60,
+     68,  76,  84,  92, 100, 112, 124, 136,
+    148, 164, 184, 208, 236, 268, 300, 332,
+    364, 396, 428, 460, 512
+};
+
 static const uint16_t swb_offset_128_48[] = {
      0,   4,   8,  12,  16,  20,  28,  36,
     44,  56,  68,  80,  96, 112, 128
@@ -1129,6 +1142,14 @@ static const uint16_t swb_offset_1024_32[] = {
     928, 960, 992, 1024
 };
 
+static const uint16_t swb_offset_512_32[] = {
+      0,   4,   8,  12,  16,  20,  24,  28,
+     32,  36,  40,  44,  48,  52,  56,  64,
+     72,  80,  88,  96, 108, 120, 132, 144,
+    160, 176, 192, 212, 236, 260, 288, 320,
+    352, 384, 416, 448, 480, 512
+};
+
 static const uint16_t swb_offset_1024_24[] = {
       0,   4,   8,  12,  16,  20,  24,  28,
      32,  36,  40,  44,  52,  60,  68,  76,
@@ -1138,6 +1159,13 @@ static const uint16_t swb_offset_1024_24[] = {
     600, 652, 704, 768, 832, 896, 960, 1024
 };
 
+static const uint16_t swb_offset_512_24[] = {
+      0,   4,   8,  12,  16,  20,  24,  28,
+     32,  36,  40,  44,  52,  60,  68,  80,
+     92, 104, 120, 140, 164, 192, 224, 256,
+    288, 320, 352, 384, 416, 448, 480, 512,
+};
+
 static const uint16_t swb_offset_128_24[] = {
      0,   4,   8,  12,  16,  20,  24,  28,
     36,  44,  52,  64,  76,  92, 108, 128
@@ -1179,6 +1207,14 @@ const uint16_t * const ff_swb_offset_1024[] = {
     swb_offset_1024_8
 };
 
+const uint16_t * const ff_swb_offset_512[] = {
+    NULL,               NULL,               NULL,
+    swb_offset_512_48,  swb_offset_512_48,  swb_offset_512_32,
+    swb_offset_512_24,  swb_offset_512_24,  NULL,
+    NULL,               NULL,               NULL,
+    NULL
+};
+
 const uint16_t * const ff_swb_offset_128[] = {
     /* The last entry on the following row is swb_offset_128_64 but is a
        duplicate of swb_offset_128_96. */

libavcodec/aactab.h

@@ -45,6 +45,7 @@
  * @{
  */
 DECLARE_ALIGNED(32, extern float,  ff_aac_kbd_long_1024)[1024];
+DECLARE_ALIGNED(32, extern float,  ff_aac_kbd_long_512 )[512];
 DECLARE_ALIGNED(32, extern float,  ff_aac_kbd_short_128)[128];
 // @}
 
@@ -52,6 +53,7 @@ DECLARE_ALIGNED(32, extern float,  ff_aac_kbd_short_128)[128];
  * @{
  */
 extern const uint8_t ff_aac_num_swb_1024[];
+extern const uint8_t ff_aac_num_swb_512 [];
 extern const uint8_t ff_aac_num_swb_128 [];
 // @}
 
@@ -69,6 +71,7 @@ extern const float *ff_aac_codebook_vector_vals[];
 extern const uint16_t *ff_aac_codebook_vector_idx[];
 
 extern const uint16_t * const ff_swb_offset_1024[13];
+extern const uint16_t * const ff_swb_offset_512 [13];
 extern const uint16_t * const ff_swb_offset_128 [13];
 
 extern const uint8_t ff_tns_max_bands_1024[13];

libavcodec/version.h

@@ -27,7 +27,7 @@
  */
 
 #define LIBAVCODEC_VERSION_MAJOR 55
-#define LIBAVCODEC_VERSION_MINOR 19
+#define LIBAVCODEC_VERSION_MINOR 20
 #define LIBAVCODEC_VERSION_MICRO  0
 
 #define LIBAVCODEC_VERSION_INT  AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \