Add HE-AAC v2 support to the AAC decoder.

Originally committed as revision 23647 to svn://svn.ffmpeg.org/ffmpeg/trunk

Changelog

@@ -11,6 +11,7 @@ version <next>:
 - CODEC_CAP_EXPERIMENTAL added
 - Demuxer for On2's IVF format
 - Pictor/PC Paint decoder
+- HE-AAC v2 decoder
 
 
 

libavcodec/Makefile

@@ -42,7 +42,7 @@ OBJS-$(CONFIG_VAAPI)                   += vaapi.o
 OBJS-$(CONFIG_VDPAU)                   += vdpau.o
 
 # decoders/encoders/hardware accelerators
-OBJS-$(CONFIG_AAC_DECODER)             += aacdec.o aactab.o aacsbr.o
+OBJS-$(CONFIG_AAC_DECODER)             += aacdec.o aactab.o aacsbr.o ps.o
 OBJS-$(CONFIG_AAC_ENCODER)             += aacenc.o aaccoder.o    \
                                           aacpsy.o aactab.o      \
                                           psymodel.o iirfilter.o \
@@ -667,5 +667,6 @@ $(SUBDIR)mpegaudiodec.o: $(SUBDIR)mpegaudio_tables.h
 $(SUBDIR)mpegaudiodec_float.o: $(SUBDIR)mpegaudio_tables.h
 $(SUBDIR)motionpixels.o: $(SUBDIR)motionpixels_tables.h
 $(SUBDIR)pcm.o: $(SUBDIR)pcm_tables.h
+$(SUBDIR)ps.o: $(SUBDIR)ps_tables.h
 $(SUBDIR)qdm2.o: $(SUBDIR)qdm2_tables.h
 endif

libavcodec/aacdec.c

@@ -67,7 +67,7 @@
  * Y (not in this code) Layer-2
  * Y (not in this code) Layer-3
  * N                    SinuSoidal Coding (Transient, Sinusoid, Noise)
- * N (planned)          Parametric Stereo
+ * Y                    Parametric Stereo
  * N                    Direct Stream Transfer
  *
  * Note: - HE AAC v1 comprises LC AAC with Spectral Band Replication.
@@ -200,7 +200,8 @@ static av_cold int che_configure(AACContext *ac,
         ff_aac_sbr_ctx_init(&ac->che[type][id]->sbr);
         if (type != TYPE_CCE) {
             ac->output_data[(*channels)++] = ac->che[type][id]->ch[0].ret;
-            if (type == TYPE_CPE) {
+            if (type == TYPE_CPE ||
+                (type == TYPE_SCE && ac->m4ac.ps == 1)) {
                 ac->output_data[(*channels)++] = ac->che[type][id]->ch[1].ret;
             }
         }
@@ -228,6 +229,7 @@ static av_cold int output_configure(AACContext *ac,
     AVCodecContext *avctx = ac->avctx;
     int i, type, channels = 0, ret;
 
+    if (new_che_pos != che_pos)
     memcpy(che_pos, new_che_pos, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));
 
     if (channel_config) {
@@ -471,6 +473,8 @@ static int decode_audio_specific_config(AACContext *ac, void *data,
         av_log(ac->avctx, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->m4ac.sampling_index);
         return -1;
     }
+    if (ac->m4ac.sbr == 1 && ac->m4ac.ps == -1)
+        ac->m4ac.ps = 1;
 
     skip_bits_long(&gb, i);
 
@@ -1667,6 +1671,10 @@ static int decode_extension_payload(AACContext *ac, GetBitContext *gb, int cnt,
             av_log(ac->avctx, AV_LOG_ERROR, "Implicit SBR was found with a first occurrence after the first frame.\n");
             skip_bits_long(gb, 8 * cnt - 4);
             return res;
+        } else if (ac->m4ac.ps == -1 && ac->output_configured < OC_LOCKED && ac->avctx->channels == 1) {
+            ac->m4ac.sbr = 1;
+            ac->m4ac.ps = 1;
+            output_configure(ac, ac->che_pos, ac->che_pos, ac->m4ac.chan_config, ac->output_configured);
         } else {
             ac->m4ac.sbr = 1;
         }
@@ -1946,8 +1954,10 @@ static int parse_adts_frame_header(AACContext *ac, GetBitContext *gb)
         } else if (ac->output_configured != OC_LOCKED) {
             ac->output_configured = OC_NONE;
         }
-        if (ac->output_configured != OC_LOCKED)
+        if (ac->output_configured != OC_LOCKED) {
             ac->m4ac.sbr = -1;
+            ac->m4ac.ps  = -1;
+        }
         ac->m4ac.sample_rate     = hdr_info.sample_rate;
         ac->m4ac.sampling_index  = hdr_info.sampling_index;
         ac->m4ac.object_type     = hdr_info.object_type;

libavcodec/aacsbr.c

@@ -31,6 +31,7 @@
 #include "aacsbr.h"
 #include "aacsbrdata.h"
 #include "fft.h"
+#include "ps.h"
 
 #include <stdint.h>
 #include <float.h>
@@ -120,6 +121,8 @@ av_cold void ff_aac_sbr_init(void)
 
     for (n = 0; n < 320; n++)
         sbr_qmf_window_ds[n] = sbr_qmf_window_us[2*n];
+
+    ff_ps_init();
 }
 
 av_cold void ff_aac_sbr_ctx_init(SpectralBandReplication *sbr)
@@ -130,6 +133,7 @@ av_cold void ff_aac_sbr_ctx_init(SpectralBandReplication *sbr)
     sbr->data[1].synthesis_filterbank_samples_offset = SBR_SYNTHESIS_BUF_SIZE - (1280 - 128);
     ff_mdct_init(&sbr->mdct, 7, 1, 1.0/64);
     ff_mdct_init(&sbr->mdct_ana, 7, 1, -2.0);
+    ff_ps_ctx_init(&sbr->ps);
 }
 
 av_cold void ff_aac_sbr_ctx_close(SpectralBandReplication *sbr)
@@ -890,7 +894,6 @@ static void read_sbr_extension(AACContext *ac, SpectralBandReplication *sbr,
                                GetBitContext *gb,
                                int bs_extension_id, int *num_bits_left)
 {
-//TODO - implement ps_data for parametric stereo parsing
     switch (bs_extension_id) {
     case EXTENSION_ID_PS:
         if (!ac->m4ac.ps) {
@@ -898,8 +901,8 @@ static void read_sbr_extension(AACContext *ac, SpectralBandReplication *sbr,
             skip_bits_long(gb, *num_bits_left); // bs_fill_bits
             *num_bits_left = 0;
         } else {
-#if 0
-            *num_bits_left -= ff_ps_data(gb, ps);
+#if 1
+            *num_bits_left -= ff_ps_read_data(ac->avctx, gb, &sbr->ps, *num_bits_left);
 #else
             av_log_missing_feature(ac->avctx, "Parametric Stereo is", 0);
             skip_bits_long(gb, *num_bits_left); // bs_fill_bits
@@ -1008,6 +1011,11 @@ static unsigned int read_sbr_data(AACContext *ac, SpectralBandReplication *sbr,
             num_bits_left -= 2;
             read_sbr_extension(ac, sbr, gb, get_bits(gb, 2), &num_bits_left); // bs_extension_id
         }
+        if (num_bits_left < 0) {
+            av_log(ac->avctx, AV_LOG_ERROR, "SBR Extension over read.\n");
+        }
+        if (num_bits_left > 0)
+            skip_bits(gb, num_bits_left);
     }
 
     return get_bits_count(gb) - cnt;
@@ -1166,7 +1174,7 @@ static void sbr_qmf_analysis(DSPContext *dsp, FFTContext *mdct, const float *in,
  * (14496-3 sp04 p206)
  */
 static void sbr_qmf_synthesis(DSPContext *dsp, FFTContext *mdct,
-                              float *out, float X[2][32][64],
+                              float *out, float X[2][38][64],
                               float mdct_buf[2][64],
                               float *v0, int *v_off, const unsigned int div,
                               float bias, float scale)
@@ -1402,7 +1410,7 @@ static int sbr_hf_gen(AACContext *ac, SpectralBandReplication *sbr,
 }
 
 /// Generate the subband filtered lowband
-static int sbr_x_gen(SpectralBandReplication *sbr, float X[2][32][64],
+static int sbr_x_gen(SpectralBandReplication *sbr, float X[2][38][64],
                      const float X_low[32][40][2], const float Y[2][38][64][2],
                      int ch)
 {
@@ -1424,7 +1432,7 @@ static int sbr_x_gen(SpectralBandReplication *sbr, float X[2][32][64],
     }
 
     for (k = 0; k < sbr->kx[1]; k++) {
-        for (i = i_Temp; i < i_f; i++) {
+        for (i = i_Temp; i < 38; i++) {
             X[0][i][k] = X_low[k][i + ENVELOPE_ADJUSTMENT_OFFSET][0];
             X[1][i][k] = X_low[k][i + ENVELOPE_ADJUSTMENT_OFFSET][1];
         }
@@ -1740,6 +1748,16 @@ void ff_sbr_apply(AACContext *ac, SpectralBandReplication *sbr, int id_aac,
         /* synthesis */
         sbr_x_gen(sbr, sbr->X[ch], sbr->X_low, sbr->data[ch].Y, ch);
     }
+
+    if (ac->m4ac.ps == 1) {
+        if (sbr->ps.start) {
+            ff_ps_apply(ac->avctx, &sbr->ps, sbr->X[0], sbr->X[1], sbr->kx[1] + sbr->m[1]);
+        } else {
+            memcpy(sbr->X[1], sbr->X[0], sizeof(sbr->X[0]));
+        }
+        nch = 2;
+    }
+
     sbr_qmf_synthesis(&ac->dsp, &sbr->mdct, L, sbr->X[0], sbr->qmf_filter_scratch,
                       sbr->data[0].synthesis_filterbank_samples,
                       &sbr->data[0].synthesis_filterbank_samples_offset,

libavcodec/avcodec.h

@@ -30,7 +30,7 @@
 #include "libavutil/avutil.h"
 
 #define LIBAVCODEC_VERSION_MAJOR 52
-#define LIBAVCODEC_VERSION_MINOR 76
+#define LIBAVCODEC_VERSION_MINOR 77
 #define LIBAVCODEC_VERSION_MICRO  0
 
 #define LIBAVCODEC_VERSION_INT  AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \

libavcodec/mpeg4audio.c

@@ -131,6 +131,14 @@ int ff_mpeg4audio_get_config(MPEG4AudioConfig *c, const uint8_t *buf, int buf_si
                 get_bits1(&gb); // skip 1 bit
         }
     }
+
+    //PS requires SBR
+    if (!c->sbr)
+        c->ps = 0;
+    //Limit implicit PS to the HE-AACv2 Profile
+    if ((c->ps == -1 && c->object_type != AOT_AAC_LC) || c->channels & ~0x01)
+        c->ps = 0;
+
     return specific_config_bitindex;
 }
 

libavcodec/ps.h

+/*
+ * MPEG-4 Parametric Stereo definitions and declarations
+ * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_PS_H
+#define AVCODEC_PS_H
+
+#include <stdint.h>
+
+#define PS_MAX_NUM_ENV 5
+#define PS_MAX_NR_IIDICC 34
+#define PS_MAX_NR_IPDOPD 17
+#define PS_MAX_SSB 91
+#define PS_MAX_AP_BANDS 50
+#define PS_QMF_TIME_SLOTS 32
+#define PS_MAX_DELAY 14
+#define PS_AP_LINKS 3
+#define PS_MAX_AP_DELAY 5
+
+typedef struct {
+    int    start;
+    int    enable_iid;
+    int    iid_mode;
+    int    iid_quant;
+    int    nr_iid_par;
+    int    nr_ipdopd_par;
+    int    enable_icc;
+    int    icc_mode;
+    int    nr_icc_par;
+    int    enable_ext;
+    int    frame_class;
+    int    num_env_old;
+    int    num_env;
+    int    enable_ipdopd;
+    int    border_position[PS_MAX_NUM_ENV+1];
+    int8_t iid_par[PS_MAX_NUM_ENV][PS_MAX_NR_IIDICC]; //<Inter-channel Intensity Difference Parameters
+    int8_t icc_par[PS_MAX_NUM_ENV][PS_MAX_NR_IIDICC]; //<Inter-Channel Coherence Parameters
+    /* ipd/opd is iid/icc sized so that the same functions can handle both */
+    int8_t ipd_par[PS_MAX_NUM_ENV][PS_MAX_NR_IIDICC]; //<Inter-channel Phase Difference Parameters
+    int8_t opd_par[PS_MAX_NUM_ENV][PS_MAX_NR_IIDICC]; //<Overall Phase Difference Parameters
+    int    is34bands;
+    int    is34bands_old;
+
+    float  in_buf[5][44][2];
+    float  delay[PS_MAX_SSB][PS_QMF_TIME_SLOTS + PS_MAX_DELAY][2];
+    float  ap_delay[PS_MAX_AP_BANDS][PS_AP_LINKS][PS_QMF_TIME_SLOTS + PS_MAX_AP_DELAY][2];
+    float  peak_decay_nrg[34];
+    float  power_smooth[34];
+    float  peak_decay_diff_smooth[34];
+    float  H11[2][PS_MAX_NUM_ENV+1][PS_MAX_NR_IIDICC];
+    float  H12[2][PS_MAX_NUM_ENV+1][PS_MAX_NR_IIDICC];
+    float  H21[2][PS_MAX_NUM_ENV+1][PS_MAX_NR_IIDICC];
+    float  H22[2][PS_MAX_NUM_ENV+1][PS_MAX_NR_IIDICC];
+    int8_t opd_hist[PS_MAX_NR_IIDICC];
+    int8_t ipd_hist[PS_MAX_NR_IIDICC];
+} PSContext;
+
+void ff_ps_init(void);
+void ff_ps_ctx_init(PSContext *ps);
+int ff_ps_read_data(AVCodecContext *avctx, GetBitContext *gb, PSContext *ps, int bits_left);
+int ff_ps_apply(AVCodecContext *avctx, PSContext *ps, float L[2][38][64], float R[2][38][64], int top);
+
+#endif /* AVCODEC_PS_H */

libavcodec/ps_tablegen.c

+/*
+ * Generate a header file for hardcoded Parametric Stereo tables
+ *
+ * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdlib.h>
+#define CONFIG_HARDCODED_TABLES 0
+#include "ps_tablegen.h"
+#include "tableprint.h"
+
+void write_float_3d_array (const void *p, int b, int c, int d)
+{
+    int i;
+    const float *f = p;
+    for (i = 0; i < b; i++) {
+        printf("{\n");
+        write_float_2d_array(f, c, d);
+        printf("},\n");
+        f += c * d;
+    }
+}
+
+void write_float_4d_array (const void *p, int a, int b, int c, int d)
+{
+    int i;
+    const float *f = p;
+    for (i = 0; i < a; i++) {
+        printf("{\n");
+        write_float_3d_array(f, b, c, d);
+        printf("},\n");
+        f += b * c * d;
+    }
+}
+
+int main(void)
+{
+    ps_tableinit();
+
+    write_fileheader();
+
+    printf("static const float pd_re_smooth[8*8*8] = {\n");
+    write_float_array(pd_re_smooth, 8*8*8);
+    printf("};\n");
+    printf("static const float pd_im_smooth[8*8*8] = {\n");
+    write_float_array(pd_im_smooth, 8*8*8);
+    printf("};\n");
+
+    printf("static const float HA[46][8][4] = {\n");
+    write_float_3d_array(HA, 46, 8, 4);
+    printf("};\n");
+    printf("static const float HB[46][8][4] = {\n");
+    write_float_3d_array(HB, 46, 8, 4);
+    printf("};\n");
+
+    printf("static const float f20_0_8[8][7][2] = {\n");
+    write_float_3d_array(f20_0_8, 8, 7, 2);
+    printf("};\n");
+    printf("static const float f34_0_12[12][7][2] = {\n");
+    write_float_3d_array(f34_0_12, 12, 7, 2);
+    printf("};\n");
+    printf("static const float f34_1_8[8][7][2] = {\n");
+    write_float_3d_array(f34_1_8, 8, 7, 2);
+    printf("};\n");
+    printf("static const float f34_2_4[4][7][2] = {\n");
+    write_float_3d_array(f34_2_4, 4, 7, 2);
+    printf("};\n");
+
+    printf("static const float Q_fract_allpass[2][50][3][2] = {\n");
+    write_float_4d_array(Q_fract_allpass, 2, 50, 3, 2);
+    printf("};\n");
+    printf("static const float phi_fract[2][50][2] = {\n");
+    write_float_3d_array(phi_fract, 2, 50, 2);
+    printf("};\n");
+
+    return 0;
+}

libavcodec/ps_tablegen.h

+/*
+ * Header file for hardcoded Parametric Stereo tables
+ *
+ * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef PS_TABLEGEN_H
+#define PS_TABLEGEN_H
+
+#include <stdint.h>
+#include <math.h>
+
+#if CONFIG_HARDCODED_TABLES
+#define ps_tableinit()
+#include "libavcodec/ps_tables.h"
+#else
+#include "../libavutil/common.h"
+#ifndef M_SQRT1_2
+#define M_SQRT1_2      0.70710678118654752440  /* 1/sqrt(2) */
+#endif
+#ifndef M_PI
+#define M_PI           3.14159265358979323846  /* pi */
+#endif
+#ifndef M_SQRT2
+#define M_SQRT2        1.41421356237309504880  /* sqrt(2) */
+#endif
+#define NR_ALLPASS_BANDS20 30
+#define NR_ALLPASS_BANDS34 50
+#define PS_AP_LINKS 3
+static float pd_re_smooth[8*8*8];
+static float pd_im_smooth[8*8*8];
+static float HA[46][8][4];
+static float HB[46][8][4];
+static float f20_0_8 [ 8][7][2];
+static float f34_0_12[12][7][2];
+static float f34_1_8 [ 8][7][2];
+static float f34_2_4 [ 4][7][2];
+static float Q_fract_allpass[2][50][3][2];
+static float phi_fract[2][50][2];
+
+static const float g0_Q8[] = {
+    0.00746082949812f, 0.02270420949825f, 0.04546865930473f, 0.07266113929591f,
+    0.09885108575264f, 0.11793710567217f, 0.125f
+};
+
+static const float g0_Q12[] = {
+    0.04081179924692f, 0.03812810994926f, 0.05144908135699f, 0.06399831151592f,
+    0.07428313801106f, 0.08100347892914f, 0.08333333333333f
+};
+
+static const float g1_Q8[] = {
+    0.01565675600122f, 0.03752716391991f, 0.05417891378782f, 0.08417044116767f,
+    0.10307344158036f, 0.12222452249753f, 0.125f
+};
+
+static const float g2_Q4[] = {
+    -0.05908211155639f, -0.04871498374946f, 0.0f,   0.07778723915851f,
+     0.16486303567403f,  0.23279856662996f, 0.25f
+};
+
+static void make_filters_from_proto(float (*filter)[7][2], const float *proto, int bands)
+{
+    int q, n;
+    for (q = 0; q < bands; q++) {
+        for (n = 0; n < 7; n++) {
+            double theta = 2 * M_PI * (q + 0.5) * (n - 6) / bands;
+            filter[q][n][0] = proto[n] *  cos(theta);
+            filter[q][n][1] = proto[n] * -sin(theta);
+        }
+    }
+}
+
+static void ps_tableinit(void)
+{
+    static const float ipdopd_sin[] = { 0, M_SQRT1_2, 1,  M_SQRT1_2,  0, -M_SQRT1_2, -1, -M_SQRT1_2 };
+    static const float ipdopd_cos[] = { 1, M_SQRT1_2, 0, -M_SQRT1_2, -1, -M_SQRT1_2,  0,  M_SQRT1_2 };
+    int pd0, pd1, pd2;
+
+    static const float iid_par_dequant[] = {
+        //iid_par_dequant_default
+        0.05623413251903, 0.12589254117942, 0.19952623149689, 0.31622776601684,
+        0.44668359215096, 0.63095734448019, 0.79432823472428, 1,
+        1.25892541179417, 1.58489319246111, 2.23872113856834, 3.16227766016838,
+        5.01187233627272, 7.94328234724282, 17.7827941003892,
+        //iid_par_dequant_fine
+        0.00316227766017, 0.00562341325190, 0.01,             0.01778279410039,
+        0.03162277660168, 0.05623413251903, 0.07943282347243, 0.11220184543020,
+        0.15848931924611, 0.22387211385683, 0.31622776601684, 0.39810717055350,
+        0.50118723362727, 0.63095734448019, 0.79432823472428, 1,
+        1.25892541179417, 1.58489319246111, 1.99526231496888, 2.51188643150958,
+        3.16227766016838, 4.46683592150963, 6.30957344480193, 8.91250938133745,
+        12.5892541179417, 17.7827941003892, 31.6227766016838, 56.2341325190349,
+        100,              177.827941003892, 316.227766016837,
+    };
+    static const float icc_invq[] = {
+        1, 0.937,      0.84118,    0.60092,    0.36764,   0,      -0.589,    -1
+    };
+    static const float acos_icc_invq[] = {
+        0, 0.35685527, 0.57133466, 0.92614472, 1.1943263, M_PI/2, 2.2006171, M_PI
+    };
+    int iid, icc;
+
+    int k, m;
+    static const int8_t f_center_20[] = {
+        -3, -1, 1, 3, 5, 7, 10, 14, 18, 22,
+    };
+    static const int8_t f_center_34[] = {
+         2,  6, 10, 14, 18, 22, 26, 30,
+        34,-10, -6, -2, 51, 57, 15, 21,
+        27, 33, 39, 45, 54, 66, 78, 42,
+       102, 66, 78, 90,102,114,126, 90,
+    };
+    static const float fractional_delay_links[] = { 0.43f, 0.75f, 0.347f };
+    const float fractional_delay_gain = 0.39f;
+
+    for (pd0 = 0; pd0 < 8; pd0++) {
+        float pd0_re = ipdopd_cos[pd0];
+        float pd0_im = ipdopd_sin[pd0];
+        for (pd1 = 0; pd1 < 8; pd1++) {
+            float pd1_re = ipdopd_cos[pd1];
+            float pd1_im = ipdopd_sin[pd1];
+            for (pd2 = 0; pd2 < 8; pd2++) {
+                float pd2_re = ipdopd_cos[pd2];
+                float pd2_im = ipdopd_sin[pd2];
+                float re_smooth = 0.25f * pd0_re + 0.5f * pd1_re + pd2_re;
+                float im_smooth = 0.25f * pd0_im + 0.5f * pd1_im + pd2_im;
+                float pd_mag = 1 / sqrt(im_smooth * im_smooth + re_smooth * re_smooth);
+                pd_re_smooth[pd0*64+pd1*8+pd2] = re_smooth * pd_mag;
+                pd_im_smooth[pd0*64+pd1*8+pd2] = im_smooth * pd_mag;
+            }
+        }
+    }
+
+    for (iid = 0; iid < 46; iid++) {
+        float c = iid_par_dequant[iid]; //<Linear Inter-channel Intensity Difference
+        float c1 = (float)M_SQRT2 / sqrtf(1.0f + c*c);
+        float c2 = c * c1;
+        for (icc = 0; icc < 8; icc++) {
+            /*if (PS_BASELINE || ps->icc_mode < 3)*/ {
+                float alpha = 0.5f * acos_icc_invq[icc];
+                float beta  = alpha * (c1 - c2) * (float)M_SQRT1_2;
+                HA[iid][icc][0] = c2 * cosf(beta + alpha);
+                HA[iid][icc][1] = c1 * cosf(beta - alpha);
+                HA[iid][icc][2] = c2 * sinf(beta + alpha);
+                HA[iid][icc][3] = c1 * sinf(beta - alpha);
+            } /* else */ {
+                float alpha, gamma, mu, rho;
+                float alpha_c, alpha_s, gamma_c, gamma_s;
+                rho = FFMAX(icc_invq[icc], 0.05f);
+                alpha = 0.5f * atan2f(2.0f * c * rho, c*c - 1.0f);
+                mu = c + 1.0f / c;
+                mu = sqrtf(1 + (4 * rho * rho - 4)/(mu * mu));
+                gamma = atanf(sqrtf((1.0f - mu)/(1.0f + mu)));
+                if (alpha < 0) alpha += M_PI/2;
+                alpha_c = cosf(alpha);
+                alpha_s = sinf(alpha);
+                gamma_c = cosf(gamma);
+                gamma_s = sinf(gamma);
+                HB[iid][icc][0] =  M_SQRT2 * alpha_c * gamma_c;
+                HB[iid][icc][1] =  M_SQRT2 * alpha_s * gamma_c;
+                HB[iid][icc][2] = -M_SQRT2 * alpha_s * gamma_s;
+                HB[iid][icc][3] =  M_SQRT2 * alpha_c * gamma_s;
+            }
+        }
+    }
+
+    for (k = 0; k < NR_ALLPASS_BANDS20; k++) {
+        double f_center, theta;
+        if (k < FF_ARRAY_ELEMS(f_center_20))
+            f_center = f_center_20[k] * 0.125;
+        else
+            f_center = k - 6.5f;
+        for (m = 0; m < PS_AP_LINKS; m++) {
+            theta = -M_PI * fractional_delay_links[m] * f_center;
+            Q_fract_allpass[0][k][m][0] = cos(theta);
+            Q_fract_allpass[0][k][m][1] = sin(theta);
+        }
+        theta = -M_PI*fractional_delay_gain*f_center;
+        phi_fract[0][k][0] = cos(theta);
+        phi_fract[0][k][1] = sin(theta);
+    }
+    for (k = 0; k < NR_ALLPASS_BANDS34; k++) {
+        double f_center, theta;
+        if (k < FF_ARRAY_ELEMS(f_center_34))
+            f_center = f_center_34[k] / 24.;
+        else
+            f_center = k - 26.5f;
+        for (m = 0; m < PS_AP_LINKS; m++) {
+            theta = -M_PI * fractional_delay_links[m] * f_center;
+            Q_fract_allpass[1][k][m][0] = cos(theta);
+            Q_fract_allpass[1][k][m][1] = sin(theta);
+        }
+        theta = -M_PI*fractional_delay_gain*f_center;
+        phi_fract[1][k][0] = cos(theta);
+        phi_fract[1][k][1] = sin(theta);
+    }
+
+    make_filters_from_proto(f20_0_8,  g0_Q8,   8);
+    make_filters_from_proto(f34_0_12, g0_Q12, 12);
+    make_filters_from_proto(f34_1_8,  g1_Q8,   8);
+    make_filters_from_proto(f34_2_4,  g2_Q4,   4);
+}
+#endif /* CONFIG_HARDCODED_TABLES */
+
+#endif /* PS_TABLEGEN_H */

libavcodec/psdata.c

+/*
+ * MPEG-4 Parametric Stereo data tables
+ * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+static const uint8_t huff_iid_df1_bits[] = {
+    18, 18, 18, 18, 18, 18, 18, 18, 18, 17, 18, 17, 17, 16, 16, 15, 14, 14,
+    13, 12, 12, 11, 10, 10,  8,  7,  6,  5,  4,  3,  1,  3,  4,  5,  6,  7,
+     8,  9, 10, 11, 11, 12, 13, 14, 14, 15, 16, 16, 17, 17, 18, 17, 18, 18,
+    18, 18, 18, 18, 18, 18, 18,
+};
+
+static const uint32_t huff_iid_df1_codes[] = {
+    0x01FEB4, 0x01FEB5, 0x01FD76, 0x01FD77, 0x01FD74, 0x01FD75, 0x01FE8A,
+    0x01FE8B, 0x01FE88, 0x00FE80, 0x01FEB6, 0x00FE82, 0x00FEB8, 0x007F42,
+    0x007FAE, 0x003FAF, 0x001FD1, 0x001FE9, 0x000FE9, 0x0007EA, 0x0007FB,
+    0x0003FB, 0x0001FB, 0x0001FF, 0x00007C, 0x00003C, 0x00001C, 0x00000C,
+    0x000000, 0x000001, 0x000001, 0x000002, 0x000001, 0x00000D, 0x00001D,
+    0x00003D, 0x00007D, 0x0000FC, 0x0001FC, 0x0003FC, 0x0003F4, 0x0007EB,
+    0x000FEA, 0x001FEA, 0x001FD6, 0x003FD0, 0x007FAF, 0x007F43, 0x00FEB9,
+    0x00FE83, 0x01FEB7, 0x00FE81, 0x01FE89, 0x01FE8E, 0x01FE8F, 0x01FE8C,
+    0x01FE8D, 0x01FEB2, 0x01FEB3, 0x01FEB0, 0x01FEB1,
+};
+
+static const uint8_t huff_iid_dt1_bits[] = {
+    16, 16, 16, 16, 16, 16, 16, 16, 16, 15, 15, 15, 15, 15, 15, 14, 14, 13,
+    13, 13, 12, 12, 11, 10,  9,  9,  7,  6,  5,  3,  1,  2,  5,  6,  7,  8,
+     9, 10, 11, 11, 12, 12, 13, 13, 14, 14, 15, 15, 15, 15, 16, 16, 16, 16,
+    16, 16, 16, 16, 16, 16, 16,
+};
+
+static const uint16_t huff_iid_dt1_codes[] = {
+    0x004ED4, 0x004ED5, 0x004ECE, 0x004ECF, 0x004ECC, 0x004ED6, 0x004ED8,
+    0x004F46, 0x004F60, 0x002718, 0x002719, 0x002764, 0x002765, 0x00276D,
+    0x0027B1, 0x0013B7, 0x0013D6, 0x0009C7, 0x0009E9, 0x0009ED, 0x0004EE,
+    0x0004F7, 0x000278, 0x000139, 0x00009A, 0x00009F, 0x000020, 0x000011,
+    0x00000A, 0x000003, 0x000001, 0x000000, 0x00000B, 0x000012, 0x000021,
+    0x00004C, 0x00009B, 0x00013A, 0x000279, 0x000270, 0x0004EF, 0x0004E2,
+    0x0009EA, 0x0009D8, 0x0013D7, 0x0013D0, 0x0027B2, 0x0027A2, 0x00271A,
+    0x00271B, 0x004F66, 0x004F67, 0x004F61, 0x004F47, 0x004ED9, 0x004ED7,
+    0x004ECD, 0x004ED2, 0x004ED3, 0x004ED0, 0x004ED1,
+};
+
+static const uint8_t huff_iid_df0_bits[] = {
+    17, 17, 17, 17, 16, 15, 13, 10,  9,  7,  6,  5,  4,  3,  1,  3,  4,  5,
+     6,  6,  8, 11, 13, 14, 14, 15, 17, 18, 18,
+};
+
+static const uint32_t huff_iid_df0_codes[] = {
+    0x01FFFB, 0x01FFFC, 0x01FFFD, 0x01FFFA, 0x00FFFC, 0x007FFC, 0x001FFD,
+    0x0003FE, 0x0001FE, 0x00007E, 0x00003C, 0x00001D, 0x00000D, 0x000005,
+    0x000000, 0x000004, 0x00000C, 0x00001C, 0x00003D, 0x00003E, 0x0000FE,
+    0x0007FE, 0x001FFC, 0x003FFC, 0x003FFD, 0x007FFD, 0x01FFFE, 0x03FFFE,
+    0x03FFFF,
+};
+
+static const uint8_t huff_iid_dt0_bits[] = {
+    19, 19, 19, 20, 20, 20, 17, 15, 12, 10,  8,  6,  4,  2,  1,  3,  5,  7,
+     9, 11, 13, 14, 17, 19, 20, 20, 20, 20, 20,
+};
+
+static const uint32_t huff_iid_dt0_codes[] = {
+    0x07FFF9, 0x07FFFA, 0x07FFFB, 0x0FFFF8, 0x0FFFF9, 0x0FFFFA, 0x01FFFD,
+    0x007FFE, 0x000FFE, 0x0003FE, 0x0000FE, 0x00003E, 0x00000E, 0x000002,
+    0x000000, 0x000006, 0x00001E, 0x00007E, 0x0001FE, 0x0007FE, 0x001FFE,
+    0x003FFE, 0x01FFFC, 0x07FFF8, 0x0FFFFB, 0x0FFFFC, 0x0FFFFD, 0x0FFFFE,
+    0x0FFFFF,
+};
+
+static const uint8_t huff_icc_df_bits[] = {
+    14, 14, 12, 10, 7, 5, 3, 1, 2, 4, 6, 8, 9, 11, 13,
+};
+
+static const uint16_t huff_icc_df_codes[] = {
+    0x3FFF, 0x3FFE, 0x0FFE, 0x03FE, 0x007E, 0x001E, 0x0006, 0x0000,
+    0x0002, 0x000E, 0x003E, 0x00FE, 0x01FE, 0x07FE, 0x1FFE,
+};
+
+static const uint8_t huff_icc_dt_bits[] = {
+    14, 13, 11, 9, 7, 5, 3, 1, 2, 4, 6, 8, 10, 12, 14,
+};
+
+static const uint16_t huff_icc_dt_codes[] = {
+    0x3FFE, 0x1FFE, 0x07FE, 0x01FE, 0x007E, 0x001E, 0x0006, 0x0000,
+    0x0002, 0x000E, 0x003E, 0x00FE, 0x03FE, 0x0FFE, 0x3FFF,
+};
+
+static const uint8_t huff_ipd_df_bits[] = {
+    1, 3, 4, 4, 4, 4, 4, 4,
+};
+
+static const uint8_t huff_ipd_df_codes[] = {
+    0x01, 0x00, 0x06, 0x04, 0x02, 0x03, 0x05, 0x07,
+};
+
+static const uint8_t huff_ipd_dt_bits[] = {
+    1, 3, 4, 5, 5, 4, 4, 3,
+};
+
+static const uint8_t huff_ipd_dt_codes[] = {
+    0x01, 0x02, 0x02, 0x03, 0x02, 0x00, 0x03, 0x03,
+};
+
+static const uint8_t huff_opd_df_bits[] = {
+    1, 3, 4, 4, 5, 5, 4, 3,
+};
+
+static const uint8_t huff_opd_df_codes[] = {
+    0x01, 0x01, 0x06, 0x04, 0x0F, 0x0E, 0x05, 0x00,
+};
+
+static const uint8_t huff_opd_dt_bits[] = {
+    1, 3, 4, 5, 5, 4, 4, 3,
+};
+
+static const uint8_t huff_opd_dt_codes[] = {
+    0x01, 0x02, 0x01, 0x07, 0x06, 0x00, 0x02, 0x03,
+};
+
+static const int8_t huff_offset[] = {
+    30, 30,
+    14, 14,
+    7, 7,
+    0, 0,
+    0, 0,
+};
+
+///Table 8.48
+static const int8_t k_to_i_20[] = {
+     1,  0,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 14, 15,
+    15, 15, 16, 16, 16, 16, 17, 17, 17, 17, 17, 18, 18, 18, 18, 18, 18, 18, 18,
+    18, 18, 18, 18, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19,
+    19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19
+};
+///Table 8.49
+static const int8_t k_to_i_34[] = {
+     0,  1,  2,  3,  4,  5,  6,  6,  7,  2,  1,  0, 10, 10,  4,  5,  6,  7,  8,
+     9, 10, 11, 12,  9, 14, 11, 12, 13, 14, 15, 16, 13, 16, 17, 18, 19, 20, 21,
+    22, 22, 23, 23, 24, 24, 25, 25, 26, 26, 27, 27, 27, 28, 28, 28, 29, 29, 29,
+    30, 30, 30, 31, 31, 31, 31, 32, 32, 32, 32, 33, 33, 33, 33, 33, 33, 33, 33,
+    33, 33, 33, 33, 33, 33, 33, 33, 33, 33, 33, 33, 33, 33, 33
+};
+
+static const float g1_Q2[] = {
+    0.0f,  0.01899487526049f, 0.0f, -0.07293139167538f,
+    0.0f,  0.30596630545168f, 0.5f
+};

libavcodec/sbr.h

@@ -31,6 +31,7 @@
 
 #include <stdint.h>
 #include "fft.h"
+#include "ps.h"
 
 /**
  * Spectral Band Replication header - spectrum parameters that invoke a reset if they differ from the previous header.
@@ -133,6 +134,7 @@ typedef struct {
     ///The number of frequency bands in f_master
     unsigned           n_master;
     SBRData            data[2];
+    PSContext          ps;
     ///N_Low and N_High respectively, the number of frequency bands for low and high resolution
     unsigned           n[2];
     ///Number of noise floor bands
@@ -157,7 +159,7 @@ typedef struct {
     ///QMF output of the HF generator
     float              X_high[64][40][2];
     ///QMF values of the reconstructed signal
-    DECLARE_ALIGNED(16, float, X)[2][2][32][64];
+    DECLARE_ALIGNED(16, float, X)[2][2][38][64];
     ///Zeroth coefficient used to filter the subband signals
     float              alpha0[64][2];
     ///First coefficient used to filter the subband signals