SIPR16k decoder

Originally committed as revision 21234 to svn://svn.ffmpeg.org/ffmpeg/trunk

Changelog

@@ -48,7 +48,7 @@ version <next>:
 - R210 decoder
 - Auravision Aura 1 and 2 decoders
 - Deluxe Paint Animation playback system
-- SIPR decoding for modes 8k5, 6k5 and 5k0
+- SIPR decoder
 - Adobe Filmstrip muxer and demuxer
 
 

doc/general.texi

@@ -620,7 +620,6 @@ following image formats are supported:
 @item RealAudio 3.0 (dnet)   @tab IX  @tab  X
     @tab Real low bitrate AC-3 codec
 @item RealAudio SIPR / ACELP.NET @tab     @tab  X
-    @tab 16 kbps mode not yet supported
 @item Shorten                @tab     @tab  X
 @item Sierra VMD audio       @tab     @tab  X
     @tab Used in Sierra VMD files.

libavcodec/Makefile

@@ -263,7 +263,8 @@ OBJS-$(CONFIG_SGI_ENCODER)             += sgienc.o rle.o
 OBJS-$(CONFIG_SHORTEN_DECODER)         += shorten.o
 OBJS-$(CONFIG_SIPR_DECODER)            += sipr.o acelp_pitch_delay.o \
                                           celp_math.o acelp_vectors.o \
-                                          acelp_filters.o celp_filters.o lsp.o
+                                          acelp_filters.o celp_filters.o lsp.o \
+                                          sipr16k.o
 OBJS-$(CONFIG_SMACKAUD_DECODER)        += smacker.o
 OBJS-$(CONFIG_SMACKER_DECODER)         += smacker.o
 OBJS-$(CONFIG_SMC_DECODER)             += smc.o

libavcodec/avcodec.h

@@ -30,7 +30,7 @@
 #include "libavutil/avutil.h"
 
 #define LIBAVCODEC_VERSION_MAJOR 52
-#define LIBAVCODEC_VERSION_MINOR 47
+#define LIBAVCODEC_VERSION_MINOR 48
 #define LIBAVCODEC_VERSION_MICRO  0
 
 #define LIBAVCODEC_VERSION_INT  AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \

libavcodec/sipr.c

@@ -51,6 +51,7 @@ typedef struct {
 
     /* bitstream parameters */
     uint8_t number_of_fc_indexes;
+    uint8_t ma_predictor_bits;  ///< size in bits of the switched MA predictor
 
     /** size in bits of the i-th stage vector of quantizer */
     uint8_t vq_indexes_bits[5];
@@ -64,6 +65,22 @@ typedef struct {
 } SiprModeParam;
 
 static const SiprModeParam modes[MODE_COUNT] = {
+    [MODE_16k] = {
+        .mode_name          = "16k",
+        .bits_per_frame     = 160,
+        .subframe_count     = SUBFRAME_COUNT_16k,
+        .frames_per_packet  = 1,
+        .pitch_sharp_factor = 0.00,
+
+        .number_of_fc_indexes = 10,
+        .ma_predictor_bits    = 1,
+        .vq_indexes_bits      = {7, 8, 7, 7, 7},
+        .pitch_delay_bits     = {9, 6},
+        .gp_index_bits        = 4,
+        .fc_index_bits        = {4, 5, 4, 5, 4, 5, 4, 5, 4, 5},
+        .gc_index_bits        = 5
+    },
+
     [MODE_8k5] = {
         .mode_name          = "8k5",
         .bits_per_frame     = 152,
@@ -72,6 +89,7 @@ static const SiprModeParam modes[MODE_COUNT] = {
         .pitch_sharp_factor = 0.8,
 
         .number_of_fc_indexes = 3,
+        .ma_predictor_bits    = 0,
         .vq_indexes_bits      = {6, 7, 7, 7, 5},
         .pitch_delay_bits     = {8, 5, 5},
         .gp_index_bits        = 0,
@@ -87,6 +105,7 @@ static const SiprModeParam modes[MODE_COUNT] = {
         .pitch_sharp_factor = 0.8,
 
         .number_of_fc_indexes = 3,
+        .ma_predictor_bits    = 0,
         .vq_indexes_bits      = {6, 7, 7, 7, 5},
         .pitch_delay_bits     = {8, 5, 5},
         .gp_index_bits        = 0,
@@ -102,6 +121,7 @@ static const SiprModeParam modes[MODE_COUNT] = {
         .pitch_sharp_factor = 0.85,
 
         .number_of_fc_indexes = 1,
+        .ma_predictor_bits    = 0,
         .vq_indexes_bits      = {6, 7, 7, 7, 5},
         .pitch_delay_bits     = {8, 5, 8, 5, 5},
         .gp_index_bits        = 0,
@@ -173,6 +193,8 @@ static void decode_parameters(SiprParameters* parms, GetBitContext *pgb,
 {
     int i, j;
 
+    parms->ma_pred_switch           = get_bits(pgb, p->ma_predictor_bits);
+
     for (i = 0; i < 5; i++)
         parms->vq_indexes[i]        = get_bits(pgb, p->vq_indexes_bits[i]);
 
@@ -490,6 +512,9 @@ static av_cold int sipr_decoder_init(AVCodecContext * avctx)
 
     av_log(avctx, AV_LOG_DEBUG, "Mode: %s\n", modes[ctx->mode].mode_name);
 
+    if (ctx->mode == MODE_16k)
+        ff_sipr_init_16k(ctx);
+
     for (i = 0; i < LP_FILTER_ORDER; i++)
         ctx->lsp_history[i] = cos((i+1) * M_PI / (LP_FILTER_ORDER + 1));
 
@@ -498,12 +523,6 @@ static av_cold int sipr_decoder_init(AVCodecContext * avctx)
 
     avctx->sample_fmt = SAMPLE_FMT_FLT;
 
-    if (ctx->mode == MODE_16k) {
-        av_log(avctx, AV_LOG_ERROR, "decoding 16kbps SIPR files is not "
-                                    "supported yet.\n");
-        return -1;
-    }
-
     dsputil_init(&ctx->dsp, avctx);
 
     return 0;
@@ -518,6 +537,7 @@ static int sipr_decode_frame(AVCodecContext *avctx, void *datap,
     const SiprModeParam *mode_par = &modes[ctx->mode];
     GetBitContext gb;
     float *data = datap;
+    int subframe_size = ctx->mode == MODE_16k ? L_SUBFR_16k : SUBFR_SIZE;
     int i;
 
     ctx->avctx = avctx;
@@ -529,7 +549,7 @@ static int sipr_decode_frame(AVCodecContext *avctx, void *datap,
         *data_size = 0;
         return -1;
     }
-    if (*data_size < SUBFR_SIZE * mode_par->subframe_count * sizeof(float)) {
+    if (*data_size < subframe_size * mode_par->subframe_count * sizeof(float)) {
         av_log(avctx, AV_LOG_ERROR,
                "Error processing packet: output buffer (%d) too small\n",
                *data_size);
@@ -542,12 +562,16 @@ static int sipr_decode_frame(AVCodecContext *avctx, void *datap,
 
     for (i = 0; i < mode_par->frames_per_packet; i++) {
         decode_parameters(&parm, &gb, mode_par);
-        decode_frame(ctx, &parm, data);
 
-        data += SUBFR_SIZE * mode_par->subframe_count;
+        if (ctx->mode == MODE_16k)
+            ff_sipr_decode_frame_16k(ctx, &parm, data);
+        else
+            decode_frame(ctx, &parm, data);
+
+        data += subframe_size * mode_par->subframe_count;
     }
 
-    *data_size = mode_par->frames_per_packet * SUBFR_SIZE *
+    *data_size = mode_par->frames_per_packet * subframe_size *
         mode_par->subframe_count * sizeof(float);
 
     return mode_par->bits_per_frame >> 3;

libavcodec/sipr.h

@@ -28,6 +28,11 @@
 #include "dsputil.h"
 #include "acelp_pitch_delay.h"
 
+#define LP_FILTER_ORDER_16k  16
+#define L_SUBFR_16k          80
+#define PITCH_MIN            30
+#define PITCH_MAX            281
+
 #define LSFQ_DIFF_MIN        (0.0125 * M_PI)
 
 #define LP_FILTER_ORDER      10
@@ -38,6 +43,8 @@
 /**  Subframe size for all modes except 16k */
 #define SUBFR_SIZE           48
 
+#define SUBFRAME_COUNT_16k   2
+
 typedef enum {
     MODE_16k,
     MODE_8k5,
@@ -53,9 +60,9 @@ typedef struct {
     SiprMode mode;
 
     float past_pitch_gain;
-    float lsf_history[LP_FILTER_ORDER];
+    float lsf_history[LP_FILTER_ORDER_16k];
 
-    float excitation[L_INTERPOL + PITCH_DELAY_MAX + 5*SUBFR_SIZE];
+    float excitation[L_INTERPOL + PITCH_MAX + 2 * L_SUBFR_16k];
 
     DECLARE_ALIGNED_16(float, synth_buf[LP_FILTER_ORDER + 5*SUBFR_SIZE + 6]);
 
@@ -70,14 +77,31 @@ typedef struct {
     float postfilter_agc;
     float postfilter_mem5k0[PITCH_DELAY_MAX + LP_FILTER_ORDER];
     float postfilter_syn5k0[LP_FILTER_ORDER + SUBFR_SIZE*5];
+
+    /* 16k */
+    int pitch_lag_prev;
+    float iir_mem[LP_FILTER_ORDER_16k+1];
+    float filt_buf[2][LP_FILTER_ORDER_16k+1];
+    float *filt_mem[2];
+    float mem_preemph[LP_FILTER_ORDER_16k];
+    float synth[LP_FILTER_ORDER_16k];
+    double lsp_history_16k[16];
 } SiprContext;
 
 typedef struct {
+    int ma_pred_switch;        ///< switched moving average predictor
     int vq_indexes[5];
     int pitch_delay[5];        ///< pitch delay
     int gp_index[5];           ///< adaptive-codebook gain indexes
     int16_t fc_indexes[5][10]; ///< fixed-codebook indexes
     int gc_index[5];           ///< fixed-codebook gain indexes
 } SiprParameters;
 
+extern const float ff_pow_0_5[16];
+
+void ff_sipr_init_16k(SiprContext *ctx);
+
+void ff_sipr_decode_frame_16k(SiprContext *ctx, SiprParameters *params,
+                              float *out_data);
+
 #endif /* AVCODEC_SIPR_H */