AMR-WB support by (Johannes Carlsson <joca at rixmail dot se>)

Originally committed as revision 2311 to svn://svn.ffmpeg.org/ffmpeg/trunk

configure

@@ -102,6 +102,7 @@ SLIBSUF=".so"
 EXESUF=""
 risky="yes"
 amr_nb="no"
+amr_wb="no"
 amr_nb_fixed="no"
 sunmlib="no"
 
@@ -850,6 +851,7 @@ fi
 echo "Sun medialib support"  $sunmlib
 echo "AMR-NB float support"  $amr_nb
 echo "AMR-NB fixed support"  $amr_nb_fixed
+echo "AMR-WB float support"  $amr_wb
 
 echo "Creating config.mak and config.h"
 
@@ -1137,6 +1139,17 @@ if test "$source_path_used" = "yes" ; then
 fi
 echo "SRC_PATH=$source_path" >> config.mak
 
+if test "$amr_wb" = "yes" ; then
+  echo "#define AMR_WB 1" >> $TMPH
+  echo "AMR_WB=yes" >> config.mak
+  echo
+  echo "AMR WB FLOAT NOTICE ! Make sure you have downloaded TS26.204"
+  echo "V5.1.0 from "
+  echo "http://www.3gpp.org/ftp/Specs/archive/26_series/26.204/26204-510.zip"
+  echo "and extracted the source to libavcodec/amrwb_float"
+  echo
+fi
+
 if test "$amr_nb" = "yes" ; then
   echo "#define AMR_NB 1" >> $TMPH
   echo "AMR_NB=yes" >> config.mak

doc/ffmpeg-doc.texi

@@ -716,6 +716,8 @@ solutions.
 @tab Real 28800 bit/s codec
 @item AMR-NB                 @tab X    @tab X
 @tab supported through an external library
+@item AMR-WB                 @tab X    @tab X
+@tab supported through an external library
 @item DV audio               @tab      @tab X
 @item Id RoQ DPCM            @tab      @tab X
 @tab used in Quake III, Jedi Knight 2, other computer games

libavcodec/Makefile

@@ -31,6 +31,15 @@ CLEANAMR=cleanamrfloat
 endif
 endif
 
+ifeq ($(AMR_WB),yes)
+OBJS+= amr.o amrwb_float/dec_acelp.o amrwb_float/dec_dtx.o amrwb_float/dec_gain.o \
+		amrwb_float/dec_if.o amrwb_float/dec_lpc.o amrwb_float/dec_main.o \
+		amrwb_float/dec_rom.o amrwb_float/dec_util.o amrwb_float/enc_acelp.o \
+		amrwb_float/enc_dtx.o amrwb_float/enc_gain.o amrwb_float/enc_if.o \
+		amrwb_float/enc_lpc.o amrwb_float/enc_main.o amrwb_float/enc_rom.o \
+		amrwb_float/enc_util.o amrwb_float/if_rom.o
+endif
+CLEANAMRWB=cleanamrwbfloat
 ASM_OBJS=
 
 # codecs which are patented in some non free countries like the us
@@ -202,6 +211,9 @@ cleanamr:
 cleanamrfloat:
 	rm -f amr_float/*.o
 
+cleanamrwbfloat:
+	$(MAKE) -C amrwb_float -f makefile.gcc clean
+
 # api example program
 apiexample: apiexample.c $(LIB)
 	$(CC) $(CFLAGS) -o $@ $< $(LIB) $(EXTRALIBS) -lm

libavcodec/allcodecs.c

@@ -138,6 +138,12 @@ void avcodec_register_all(void)
     register_avcodec(&amr_nb_decoder);
     register_avcodec(&amr_nb_encoder);
 #endif /* AMR_NB */
+
+#ifdef AMR_WB
+    register_avcodec(&amr_wb_decoder);
+    register_avcodec(&amr_wb_encoder);
+#endif /* AMR_WB */
+
     /* pcm codecs */
 
 #define PCM_CODEC(id, name) \

libavcodec/amr.c

@@ -17,31 +17,42 @@
  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
  /*
-    This code implements amr-nb audio encoder/decoder through external reference
+    This code implements amr-nb and amr-wb audio encoder/decoder through external reference
     code from www.3gpp.org. The licence of the code from 3gpp is unclear so you
     have to download the code separately. Two versions exists: One fixed-point
     and one with floats. For some reason the float-encoder is significant faster
     atleast on a P4 1.5GHz (0.9s instead of 9.9s on a 30s audio clip at MR102).
+    Both float and fixed point is supported for amr-nb, but only float for
+    amr-wb.
     
+    --AMR-NB--
     The fixed-point (TS26.073) can be downloaded from:
-    http://www.3gpp.org/ftp/Specs/latest/Rel-5/26_series/26073-510.zip
+    http://www.3gpp.org/ftp/Specs/archive/26_series/26.073/26073-510.zip
     Extract the soure into ffmpeg/libavcodec/amr
-    To use the float version run "./configure" with "--enable-amr-nb-fixed"
+    To use the fixed version run "./configure" with "--enable-amr-nb-fixed"
     
     The float version (default) can be downloaded from:
-    http://www.3gpp.org/ftp/Specs/latest/Rel-5/26_series/26104-510.zip
+    http://www.3gpp.org/ftp/Specs/archive/26_series/26.104/26104-510.zip
     Extract the soure into ffmpeg/libavcodec/amr_float
 
     The specification for amr-nb can be found in TS 26.071
     (http://www.3gpp.org/ftp/Specs/html-info/26071.htm) and some other
     info at http://www.3gpp.org/ftp/Specs/html-info/26-series.htm
     
-    In the future support for AMR-WB might also be included here.
+    --AMR-WB--
-    Reference code exist in TS26.173 and TS 26.204.
+    The reference code can be downloaded from:
+    http://www.3gpp.org/ftp/Specs/archive/26_series/26.204/26204-510.zip
+    It should be extracted to "libavcodec/amrwb_float". Enable it with
+    "--enable-amr-wb".
+    
+    The specification for amr-wb can be downloaded from:
+    http://www.3gpp.org/ftp/Specs/archive/26_series/26.171/26171-500.zip
+    
+    If someone want to use the fixed point version it can be downloaded
+    from: http://www.3gpp.org/ftp/Specs/archive/26_series/26.173/26173-571.zip
  
  */
-#define DEBUG
+
-//#define AMR_NB_FIXED
 #include "../config.h"
 #include "avcodec.h"
 
@@ -86,7 +97,7 @@ static enum Mode getBitrateMode(int bitrate)
                            
                          };
     int i;
-    for(i=0;i<sizeof(rates);i++)
+    for(i=0;i<8;i++)
     {
         if(rates[i].startrate<=bitrate && rates[i].stoprate>=bitrate)
         {
@@ -148,17 +159,19 @@ static int amr_nb_encode_init(AVCodecContext * avctx)
     
     if(avctx->sample_rate!=8000)
     {
-#ifdef DEBUG
+        if(avctx->debug)
-        printf("Only 8000Hz sample rate supported\n");
+        {
-#endif
+            fprintf(stderr, "Only 8000Hz sample rate supported\n");
+        }
         return -1;
     }
 
     if(avctx->channels!=1)
     {
-#ifdef DEBUG
+        if(avctx->debug)
-        printf("Only mono supported\n");
+        {
-#endif
+            fprintf(stderr, "Only mono supported\n");
+        }
         return -1;
     }
 
@@ -167,9 +180,10 @@ static int amr_nb_encode_init(AVCodecContext * avctx)
 
     if(Speech_Encode_Frame_init(&s->enstate, 0, "encoder") || sid_sync_init (&s->sidstate))
     {
-#ifdef DEBUG
+        if(avctx->debug)
-        printf("Speech_Encode_Frame_init error\n");
+        {
-#endif
+            fprintf(stderr, "Speech_Encode_Frame_init error\n");
+        }
         return -1;
     }
 
@@ -343,17 +357,19 @@ static int amr_nb_encode_init(AVCodecContext * avctx)
     
     if(avctx->sample_rate!=8000)
     {
-#ifdef DEBUG
+        if(avctx->debug)
-        printf("Only 8000Hz sample rate supported\n");
+        {
-#endif
+            fprintf(stderr, "Only 8000Hz sample rate supported\n");
+        }
         return -1;
     }
 
     if(avctx->channels!=1)
     {
-#ifdef DEBUG
+        if(avctx->debug)
-        printf("Only mono supported\n");
+        {
-#endif
+            fprintf(stderr, "Only mono supported\n");
+        }
         return -1;
     }
 
@@ -363,7 +379,10 @@ static int amr_nb_encode_init(AVCodecContext * avctx)
     s->enstate=Encoder_Interface_init(0);
     if(!s->enstate)
     {
-        printf("Encoder_Interface_init error\n");
+        if(avctx->debug)
+        {
+            fprintf(stderr, "Encoder_Interface_init error\n");
+        }
         return -1;
     }
 
@@ -458,3 +477,176 @@ AVCodec amr_nb_encoder =
     amr_nb_encode_close,
     NULL,
 };
+
+/* -----------AMR wideband ------------*/
+#ifdef AMR_WB
+
+#ifdef _TYPEDEF_H
+//To avoid duplicate typedefs from typdef in amr-nb
+#define typedef_h
+#endif
+
+#include "amrwb_float/enc_if.h"
+#include "amrwb_float/dec_if.h"
+
+/* Common code for fixed and float version*/
+typedef struct AMRWB_bitrates
+{
+    int startrate;
+    int stoprate;
+    int mode;
+    
+} AMRWB_bitrates;
+
+static int getWBBitrateMode(int bitrate)
+{
+    /* Adjusted so that all bitrates can be used from commandline where
+       only a multiple of 1000 can be specified*/
+    AMRWB_bitrates rates[]={ {0,7999,0}, //6.6kHz
+                           {8000,9999,1},//8.85
+                           {10000,13000,2},//12.65
+                           {13001,14999,3},//14.25
+                           {15000,17000,4},//15.85
+                           {17001,18000,5},//18.25
+                           {18001,22000,6},//19.85
+                           {22001,23000,7},//23.05
+                           {23001,24000,8},//23.85
+                           
+                         };
+    int i;
+
+    for(i=0;i<9;i++)
+    {
+        if(rates[i].startrate<=bitrate && rates[i].stoprate>=bitrate)
+        {
+            return(rates[i].mode);
+        }
+    }
+    /*Return highest possible*/
+    return(8);
+}
+
+
+typedef struct AMRWBContext {
+    int frameCount;
+    void *state;
+    int mode;
+    Word16 allow_dtx;
+} AMRWBContext;
+
+static int amr_wb_encode_init(AVCodecContext * avctx)
+{
+    AMRWBContext *s = (AMRWBContext*)avctx->priv_data;
+    s->frameCount=0;
+    
+    if(avctx->sample_rate!=16000)
+    {
+        if(avctx->debug)
+        {
+            fprintf(stderr, "Only 16000Hz sample rate supported\n");
+        }
+        return -1;
+    }
+
+    if(avctx->channels!=1)
+    {
+        if(avctx->debug)
+        {
+            fprintf(stderr, "Only mono supported\n");
+        }
+        return -1;
+    }
+
+    avctx->frame_size=320;
+    avctx->coded_frame= avcodec_alloc_frame();
+
+    s->state = E_IF_init();
+    s->mode=getWBBitrateMode(avctx->bit_rate);
+    s->allow_dtx=0;
+
+    return 0;
+}
+
+static int amr_wb_encode_close(AVCodecContext * avctx)
+{
+    AMRWBContext *s = (AMRWBContext*) avctx->priv_data;
+    E_IF_exit(s->state);
+    av_freep(&avctx->coded_frame);
+    s->frameCount++;
+    return 0;
+}
+
+static int amr_wb_encode_frame(AVCodecContext *avctx,
+			    unsigned char *frame/*out*/, int buf_size, void *data/*in*/)
+{
+    AMRWBContext *s = (AMRWBContext*) avctx->priv_data;
+    int size = E_IF_encode(s->state, s->mode, data, frame, s->allow_dtx);
+    return size;
+}
+
+static int amr_wb_decode_init(AVCodecContext * avctx)
+{
+    AMRWBContext *s = (AMRWBContext *)avctx->priv_data;
+    s->frameCount=0;
+    s->state = D_IF_init();
+    return 0;
+}
+
+extern const UWord8 block_size[];
+
+static int amr_wb_decode_frame(AVCodecContext * avctx,
+            void *data, int *data_size,
+            uint8_t * buf, int buf_size)
+{
+    AMRWBContext *s = (AMRWBContext*)avctx->priv_data;
+
+    uint8_t*amrData=buf;
+    int offset=0;
+    int mode;
+    int packet_size;
+    *data_size=0;
+
+    while(offset<buf_size)
+    {
+        s->frameCount++;
+        mode = (Word16)((amrData[offset] >> 3) & 0x0F);
+        packet_size = block_size[mode];
+        D_IF_decode( s->state, &amrData[offset], data+*data_size, _good_frame);
+    	*data_size+=320*2;
+        offset+=packet_size; 
+    }
+    return buf_size;
+}
+
+static int amr_wb_decode_close(AVCodecContext * avctx)
+{
+    AMRWBContext *s = (AMRWBContext *)avctx->priv_data;
+    D_IF_exit(s->state);
+    return 0;
+}
+
+AVCodec amr_wb_decoder =
+{
+    "amr_wb",
+    CODEC_TYPE_AUDIO,
+    CODEC_ID_AMR_WB,
+    sizeof(AMRWBContext),
+    amr_wb_decode_init,
+    NULL,
+    amr_wb_decode_close,
+    amr_wb_decode_frame,
+};
+
+AVCodec amr_wb_encoder =
+{
+    "amr_wb",
+    CODEC_TYPE_AUDIO,
+    CODEC_ID_AMR_WB,
+    sizeof(AMRWBContext),
+    amr_wb_encode_init,
+    amr_wb_encode_frame,
+    amr_wb_encode_close,
+    NULL,
+};
+
+#endif //AMR_WB

libavcodec/avcodec.h

@@ -95,6 +95,8 @@ enum CodecID {
 
 	/* AMR */
     CODEC_ID_AMR_NB,
+    CODEC_ID_AMR_WB,
+
     /* RealAudio codecs*/
     CODEC_ID_RA_144,
     CODEC_ID_RA_288,
@@ -1403,6 +1405,8 @@ extern AVCodec indeo3_decoder;
 extern AVCodec vp3_decoder;
 extern AVCodec amr_nb_decoder;
 extern AVCodec amr_nb_encoder;
+extern AVCodec amr_wb_encoder;
+extern AVCodec amr_wb_decoder;
 extern AVCodec aac_decoder;
 extern AVCodec mpeg4aac_decoder;
 extern AVCodec asv1_decoder;

libavformat/Makefile

@@ -21,7 +21,7 @@ ifeq ($(CONFIG_RISKY),yes)
 OBJS+= asf.o
 endif
 
-ifeq ($(AMR_NB),yes)
+ifeq ($(AMR_NB),yes) | ifeq ($(AMR_NB_FIXED),yes) | ifeq ($(AMR_WB),yes)
 OBJS+= amr.o
 endif
 

libavformat/allformats.c

@@ -57,7 +57,7 @@ void av_register_all(void)
     ipmovie_init();
     wc3_init();
 
-#ifdef AMR_NB
+#if defined(AMR_NB) || defined(AMR_NB_FIXED) || defined(AMR_WB)
     amr_init();
 #endif
     yuv4mpeg_init();

libavformat/amr.c

@@ -20,34 +20,34 @@
 /*
 Write and read amr data according to RFC3267, http://www.ietf.org/rfc/rfc3267.txt?number=3267
 
-Only amr narrowband (not amr-wb) is supported for now.
+Only mono files are supported.
 
 */
 #include "avformat.h"
-#include "avi.h"
 
 static const unsigned char AMR_header [] = "#!AMR\n";
-
+static const unsigned char AMRWB_header [] = "#!AMR-WB\n";
-/* AMR_FILE header */
-static int put_amr_header(ByteIOContext *pb, AVCodecContext *enc)
-{
-    put_tag(pb, AMR_header);       /* magic number */
-    return 0;
-}
 
 static int amr_write_header(AVFormatContext *s)
 {
     ByteIOContext *pb = &s->pb;
+    AVCodecContext *enc = &s->streams[0]->codec;
 
     s->priv_data = NULL;
 
-    /* format header */
+    if (enc->codec_id == CODEC_ID_AMR_NB)
-    if (put_amr_header(pb, &s->streams[0]->codec) < 0) {
+    {
-        return -1;
+        put_tag(pb, AMR_header);       /* magic number */
+    }
+    else if(enc->codec_id == CODEC_ID_AMR_WB)
+    {
+        put_tag(pb, AMRWB_header);       /* magic number */
+    }
+    else
+    {
+        //This is an error!
     }
-
     put_flush_packet(pb);
-
     return 0;
 }
 
@@ -66,10 +66,13 @@ static int amr_write_trailer(AVFormatContext *s)
 
 static int amr_probe(AVProbeData *p)
 {
-    /* check file header */
+    //Only check for "#!AMR" which could be amr-wb, amr-nb. 
-    if (p->buf_size < 6)
+    //This will also trigger multichannel files: "#!AMR_MC1.0\n" and 
+    //"#!AMR-WB_MC1.0\n" (not supported)
+
+    if (p->buf_size < 5)
         return 0;
-    if(memcmp(p->buf,AMR_header,6)==0)
+    if(memcmp(p->buf,AMR_header,5)==0)
         return AVPROBE_SCORE_MAX;
     else
         return 0;
@@ -81,24 +84,44 @@ static int amr_read_header(AVFormatContext *s,
 {
     ByteIOContext *pb = &s->pb;
     AVStream *st;
-    uint8_t header[6];
+    uint8_t header[9];
 
     get_buffer(pb, header, 6);
 
     if(memcmp(header,AMR_header,6)!=0)
+    {
+        get_buffer(pb, header+6, 3);
+        if(memcmp(header,AMRWB_header,9)!=0)
         {
             return -1;
         }
-
         st = av_new_stream(s, 0);
         if (!st)
+        {
             return AVERROR_NOMEM;
+        }
+    
+        st->codec.codec_type = CODEC_TYPE_AUDIO;
+        st->codec.codec_tag = CODEC_ID_AMR_WB;
+        st->codec.codec_id = CODEC_ID_AMR_WB;
+        st->codec.channels = 1;
+        st->codec.sample_rate = 16000;
+    }
+    else
+    {
+        st = av_new_stream(s, 0);
+        if (!st)
+        {
+            return AVERROR_NOMEM;
+        }
     
         st->codec.codec_type = CODEC_TYPE_AUDIO;
         st->codec.codec_tag = CODEC_ID_AMR_NB;
         st->codec.codec_id = CODEC_ID_AMR_NB;
         st->codec.channels = 1;
         st->codec.sample_rate = 8000;
+    }
+
     return 0;
 }
 
@@ -107,26 +130,30 @@ static int amr_read_header(AVFormatContext *s,
 static int amr_read_packet(AVFormatContext *s,
                           AVPacket *pkt)
 {
-    static uint16_t packed_size[16] = {12, 13, 15, 17, 19, 20, 26, 31, 5, 0, 0, 0, 0, 0, 0, 0};
+    AVCodecContext *enc = &s->streams[0]->codec;
+
+    if (enc->codec_id == CODEC_ID_AMR_NB)
+    {
+        const static uint8_t packed_size[16] = {12, 13, 15, 17, 19, 20, 26, 31, 5, 0, 0, 0, 0, 0, 0, 0};
         uint8_t toc, q, ft;
         int read;
         int size;
     
         if (url_feof(&s->pb))
+        {
             return -EIO;
-
+        }
-    toc=0;
     
         toc=get_byte(&s->pb);
-
         q  = (toc >> 2) & 0x01;
         ft = (toc >> 3) & 0x0F;
     
         size=packed_size[ft];
-    //printf("amr_read_packet size=%d\n",size);
     
         if (av_new_packet(pkt, size+1))
+        {
             return -EIO;
+        }
         pkt->stream_index = 0;
         
         pkt->data[0]=toc;
@@ -140,6 +167,45 @@ static int amr_read_packet(AVFormatContext *s,
         }
     
         return 0;
+    }
+    else if(enc->codec_id == CODEC_ID_AMR_WB)
+    {
+        static uint8_t packed_size[16] = {18, 24, 33, 37, 41, 47, 51, 59, 61, 6, 6, 0, 0, 0, 1, 1};
+        uint8_t toc, mode;
+        int read;
+        int size;
+    
+        if (url_feof(&s->pb))
+        {
+            return -EIO;
+        }
+    
+        toc=get_byte(&s->pb);
+        mode = (uint8_t)((toc >> 3) & 0x0F);
+        size = packed_size[mode];
+    
+        if ( (size==0) || av_new_packet(pkt, size))
+        {
+            return -EIO;
+        }
+    
+        pkt->stream_index = 0;
+        pkt->data[0]=toc;
+    
+        read = get_buffer(&s->pb, pkt->data+1, size-1);
+    
+        if (read != (size-1))
+        {
+            av_free_packet(pkt);
+            return -EIO;
+        }
+    
+        return 0;
+    }
+    else
+    {
+        return -EIO;
+    }
 }
 
 static int amr_read_close(AVFormatContext *s)
@@ -150,7 +216,7 @@ static int amr_read_close(AVFormatContext *s)
 static AVInputFormat amr_iformat = {
     "amr",
     "3gpp amr file format",
-    0,
+    0, /*priv_data_size*/
     amr_probe,
     amr_read_header,
     amr_read_packet,

libavformat/mov.c

@@ -121,6 +121,7 @@ static const CodecTag mov_audio_tags[] = {
     { CODEC_ID_AAC, MKTAG('m', 'p', '4', 'a') }, /* MPEG 4 AAC or audio ? */
     /* The standard for mpeg4 audio is still not normalised AFAIK anyway */
     { CODEC_ID_AMR_NB, MKTAG('s', 'a', 'm', 'r') }, /* AMR-NB 3gp */
+    { CODEC_ID_AMR_WB, MKTAG('s', 'a', 'w', 'b') }, /* AMR-WB 3gp */
     { CODEC_ID_NONE, 0 },
 };
 
@@ -850,10 +851,10 @@ static int mov_read_stsd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
 #endif
 	} else {
             st->codec.codec_id = codec_get_id(mov_audio_tags, format);
-	    if(st->codec.codec_id==CODEC_ID_AMR_NB) //from TS26.244
+	    if(st->codec.codec_id==CODEC_ID_AMR_NB || st->codec.codec_id==CODEC_ID_AMR_WB) //from TS26.244
 	    {
 #ifdef DEBUG
-               printf("AMR-NB audio identified!!\n");
+               printf("AMR-NB or AMR-WB audio identified!!\n");
 #endif
                get_be32(pb);get_be32(pb); //Reserved_8
                get_be16(pb);//Reserved_2
@@ -864,40 +865,25 @@ static int mov_read_stsd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
 
                 //AMRSpecificBox.(10 bytes)
                
-#ifdef DEBUG
-               int damr_size=
-#endif
                get_be32(pb); //size
-#ifdef DEBUG
-               int damr_type=
-#endif
                get_be32(pb); //type=='damr'
-#ifdef DEBUG
-               int damr_vendor=
-#endif
                get_be32(pb); //vendor
                get_byte(pb); //decoder version
                get_be16(pb); //mode_set
                get_byte(pb); //mode_change_period
                get_byte(pb); //frames_per_sample
 
-#ifdef DEBUG
-               printf("Audio: damr_type=%c%c%c%c damr_size=%d damr_vendor=%c%c%c%c\n",
-                       (damr_type >> 24) & 0xff,
-                       (damr_type >> 16) & 0xff,
-                       (damr_type >> 8) & 0xff,
-                       (damr_type >> 0) & 0xff,
-		       damr_size, 
-                       (damr_vendor >> 24) & 0xff,
-                       (damr_vendor >> 16) & 0xff,
-                       (damr_vendor >> 8) & 0xff,
-                       (damr_vendor >> 0) & 0xff
-		       );
-#endif
-
                st->duration = AV_NOPTS_VALUE;//Not possible to get from this info, must count number of AMR frames
+               if(st->codec.codec_id==CODEC_ID_AMR_NB)
+               {
                    st->codec.sample_rate=8000;
                    st->codec.channels=1;
+               }
+               else //AMR-WB
+               {
+                   st->codec.sample_rate=16000;
+                   st->codec.channels=1;
+               }
                st->codec.bits_per_sample=16;
                st->codec.bit_rate=0; /*It is not possible to tell this before we have 
                                        an audio frame and even then every frame can be different*/