#include "libavutil/intfloat.h"
#include "libavutil/intreadwrite.h"
#include "avcodec.h"
#include "internal.h"
#include "get_bits.h"
#include "put_bits.h"
#include "wmaprodata.h"
#include "dsputil.h"
#include "fmtconvert.h"
#include "sinewin.h"
#include "wma.h"
Go to the source code of this file.
Data Structures | |
| struct | WMAProChannelCtx |
| frame specific decoder context for a single channel More... | |
| struct | WMAProChannelGrp |
| channel group for channel transformations More... | |
| struct | WMAProDecodeCtx |
| main decoder context More... | |
Defines | |
| #define | WMAPRO_MAX_CHANNELS 8 |
| current decoder limitations | |
| #define | MAX_SUBFRAMES 32 |
| max number of subframes per channel | |
| #define | MAX_BANDS 29 |
| max number of scale factor bands | |
| #define | MAX_FRAMESIZE 32768 |
| maximum compressed frame size | |
| #define | WMAPRO_BLOCK_MIN_BITS 6 |
| log2 of min block size | |
| #define | WMAPRO_BLOCK_MAX_BITS 12 |
| log2 of max block size | |
| #define | WMAPRO_BLOCK_MAX_SIZE (1 << WMAPRO_BLOCK_MAX_BITS) |
| maximum block size | |
| #define | WMAPRO_BLOCK_SIZES (WMAPRO_BLOCK_MAX_BITS - WMAPRO_BLOCK_MIN_BITS + 1) |
| possible block sizes | |
| #define | VLCBITS 9 |
| #define | SCALEVLCBITS 8 |
| #define | VEC4MAXDEPTH ((HUFF_VEC4_MAXBITS+VLCBITS-1)/VLCBITS) |
| #define | VEC2MAXDEPTH ((HUFF_VEC2_MAXBITS+VLCBITS-1)/VLCBITS) |
| #define | VEC1MAXDEPTH ((HUFF_VEC1_MAXBITS+VLCBITS-1)/VLCBITS) |
| #define | SCALEMAXDEPTH ((HUFF_SCALE_MAXBITS+SCALEVLCBITS-1)/SCALEVLCBITS) |
| #define | SCALERLMAXDEPTH ((HUFF_SCALE_RL_MAXBITS+VLCBITS-1)/VLCBITS) |
| #define | PRINT(a, b) av_log(s->avctx, AV_LOG_DEBUG, " %s = %d\n", a, b); |
| #define | PRINT_HEX(a, b) av_log(s->avctx, AV_LOG_DEBUG, " %s = %x\n", a, b); |
Functions | |
| static void av_cold | dump_context (WMAProDecodeCtx *s) |
| helper function to print the most important members of the context | |
| static av_cold int | decode_end (AVCodecContext *avctx) |
| Uninitialize the decoder and free all resources. | |
| static av_cold int | decode_init (AVCodecContext *avctx) |
| Initialize the decoder. | |
| static int | decode_subframe_length (WMAProDecodeCtx *s, int offset) |
| Decode the subframe length. | |
| static int | decode_tilehdr (WMAProDecodeCtx *s) |
| Decode how the data in the frame is split into subframes. | |
| static void | decode_decorrelation_matrix (WMAProDecodeCtx *s, WMAProChannelGrp *chgroup) |
| Calculate a decorrelation matrix from the bitstream parameters. | |
| static int | decode_channel_transform (WMAProDecodeCtx *s) |
| Decode channel transformation parameters. | |
| static int | decode_coeffs (WMAProDecodeCtx *s, int c) |
| Extract the coefficients from the bitstream. | |
| static int | decode_scale_factors (WMAProDecodeCtx *s) |
| Extract scale factors from the bitstream. | |
| static void | inverse_channel_transform (WMAProDecodeCtx *s) |
| Reconstruct the individual channel data. | |
| static void | wmapro_window (WMAProDecodeCtx *s) |
| Apply sine window and reconstruct the output buffer. | |
| static int | decode_subframe (WMAProDecodeCtx *s) |
| Decode a single subframe (block). | |
| static int | decode_frame (WMAProDecodeCtx *s, int *got_frame_ptr) |
| Decode one WMA frame. | |
| static int | remaining_bits (WMAProDecodeCtx *s, GetBitContext *gb) |
| Calculate remaining input buffer length. | |
| static void | save_bits (WMAProDecodeCtx *s, GetBitContext *gb, int len, int append) |
| Fill the bit reservoir with a (partial) frame. | |
| static int | decode_packet (AVCodecContext *avctx, void *data, int *got_frame_ptr, AVPacket *avpkt) |
| Decode a single WMA packet. | |
| static void | flush (AVCodecContext *avctx) |
| Clear decoder buffers (for seeking). | |
Variables | |
| static VLC | sf_vlc |
| scale factor DPCM vlc | |
| static VLC | sf_rl_vlc |
| scale factor run length vlc | |
| static VLC | vec4_vlc |
| 4 coefficients per symbol | |
| static VLC | vec2_vlc |
| 2 coefficients per symbol | |
| static VLC | vec1_vlc |
| 1 coefficient per symbol | |
| static VLC | coef_vlc [2] |
| coefficient run length vlc codes | |
| static float | sin64 [33] |
| sinus table for decorrelation | |
| AVCodec | ff_wmapro_decoder |
| wmapro decoder | |
The decoding therefore consists of the following steps:
The compressed wmapro bitstream is split into individual packets. Every such packet contains one or more wma frames. The compressed frames may have a variable length and frames may cross packet boundaries. Common to all wmapro frames is the number of samples that are stored in a frame. The number of samples and a few other decode flags are stored as extradata that has to be passed to the decoder.
The wmapro frames themselves are again split into a variable number of subframes. Every subframe contains the data for 2^N time domain samples where N varies between 7 and 12.
Example wmapro bitstream (in samples):
|| packet 0 || packet 1 || packet 2 packets --------------------------------------------------- || frame 0 || frame 1 || frame 2 || frames --------------------------------------------------- || | | || | | | || || subframes of channel 0 --------------------------------------------------- || | | || | | | || || subframes of channel 1 ---------------------------------------------------
The frame layouts for the individual channels of a wma frame does not need to be the same.
However, if the offsets and lengths of several subframes of a frame are the same, the subframes of the channels can be grouped. Every group may then use special coding techniques like M/S stereo coding to improve the compression ratio. These channel transformations do not need to be applied to a whole subframe. Instead, they can also work on individual scale factor bands (see below). The coefficients that carry the audio signal in the frequency domain are transmitted as huffman-coded vectors with 4, 2 and 1 elements. In addition to that, the encoder can switch to a runlevel coding scheme by transmitting subframe_length / 128 zero coefficients.
Before the audio signal can be converted to the time domain, the coefficients have to be rescaled and inverse quantized. A subframe is therefore split into several scale factor bands that get scaled individually. Scale factors are submitted for every frame but they might be shared between the subframes of a channel. Scale factors are initially DPCM-coded. Once scale factors are shared, the differences are transmitted as runlevel codes. Every subframe length and offset combination in the frame layout shares a common quantization factor that can be adjusted for every channel by a modifier. After the inverse quantization, the coefficients get processed by an IMDCT. The resulting values are then windowed with a sine window and the first half of the values are added to the second half of the output from the previous subframe in order to reconstruct the output samples.
Definition in file wmaprodec.c.
| #define MAX_BANDS 29 |
| #define MAX_FRAMESIZE 32768 |
| #define MAX_SUBFRAMES 32 |
| #define PRINT | ( | a, | |||
| b | ) | av_log(s->avctx, AV_LOG_DEBUG, " %s = %d\n", a, b); |
| #define PRINT_HEX | ( | a, | |||
| b | ) | av_log(s->avctx, AV_LOG_DEBUG, " %s = %x\n", a, b); |
| #define SCALEMAXDEPTH ((HUFF_SCALE_MAXBITS+SCALEVLCBITS-1)/SCALEVLCBITS) |
Definition at line 118 of file wmaprodec.c.
| #define SCALERLMAXDEPTH ((HUFF_SCALE_RL_MAXBITS+VLCBITS-1)/VLCBITS) |
Definition at line 119 of file wmaprodec.c.
| #define SCALEVLCBITS 8 |
Definition at line 114 of file wmaprodec.c.
| #define VEC1MAXDEPTH ((HUFF_VEC1_MAXBITS+VLCBITS-1)/VLCBITS) |
Definition at line 117 of file wmaprodec.c.
| #define VEC2MAXDEPTH ((HUFF_VEC2_MAXBITS+VLCBITS-1)/VLCBITS) |
Definition at line 116 of file wmaprodec.c.
| #define VEC4MAXDEPTH ((HUFF_VEC4_MAXBITS+VLCBITS-1)/VLCBITS) |
Definition at line 115 of file wmaprodec.c.
| #define VLCBITS 9 |
Definition at line 113 of file wmaprodec.c.
| #define WMAPRO_BLOCK_MAX_BITS 12 |
| #define WMAPRO_BLOCK_MAX_SIZE (1 << WMAPRO_BLOCK_MAX_BITS) |
| #define WMAPRO_BLOCK_MIN_BITS 6 |
log2 of min block size
Definition at line 107 of file wmaprodec.c.
Referenced by decode_init(), decode_subframe(), and wmapro_window().
| #define WMAPRO_BLOCK_SIZES (WMAPRO_BLOCK_MAX_BITS - WMAPRO_BLOCK_MIN_BITS + 1) |
possible block sizes
Definition at line 110 of file wmaprodec.c.
Referenced by decode_end(), and decode_init().
| #define WMAPRO_MAX_CHANNELS 8 |
current decoder limitations
max number of handled channels
Definition at line 102 of file wmaprodec.c.
Referenced by decode_decorrelation_matrix(), decode_frame(), decode_init(), decode_tilehdr(), and inverse_channel_transform().
| static int decode_channel_transform | ( | WMAProDecodeCtx * | s | ) | [static] |
Decode channel transformation parameters.
| s | codec context |
in the one channel case channel transforms are pointless
decode channel mask
decode transform type
cos(pi/4)
FIXME: more than 6 coupled channels not supported
decode transform on / off
transform can be enabled for individual bands
Definition at line 665 of file wmaprodec.c.
Referenced by decode_subframe().
| static int decode_coeffs | ( | WMAProDecodeCtx * | s, | |
| int | c | |||
| ) | [static] |
Extract the coefficients from the bitstream.
| s | codec context | |
| c | current channel number |
decode vector coefficients (consumes up to 167 bits per iteration for 4 vector coded large values)
decode sign
switch to run level mode when subframe_len / 128 zeros were found in a row
decode run level coded coefficients
Definition at line 778 of file wmaprodec.c.
Referenced by decode_subframe().
| static void decode_decorrelation_matrix | ( | WMAProDecodeCtx * | s, | |
| WMAProChannelGrp * | chgroup | |||
| ) | [static] |
Calculate a decorrelation matrix from the bitstream parameters.
| s | codec context | |
| chgroup | channel group for which the matrix needs to be calculated |
Definition at line 615 of file wmaprodec.c.
Referenced by decode_channel_transform().
| static av_cold int decode_end | ( | AVCodecContext * | avctx | ) | [static] |
Uninitialize the decoder and free all resources.
| avctx | codec context |
Definition at line 258 of file wmaprodec.c.
| static int decode_frame | ( | WMAProDecodeCtx * | s, | |
| int * | got_frame_ptr | |||
| ) | [static] |
Decode one WMA frame.
| s | codec context |
get frame length
decode tile information
read postproc transform
read drc info
no idea what these are for, might be the number of samples that need to be skipped at the beginning or end of a stream
usually true for the first frame
sometimes true for the last frame
reset subframe states
decode all subframes
interleave samples and write them to the output buffer
reuse second half of the IMDCT output for the next frame
FIXME: not sure if this is always an error
skip the rest of the frame data
decode trailer bit
Definition at line 1296 of file wmaprodec.c.
| static av_cold int decode_init | ( | AVCodecContext * | avctx | ) | [static] |
Initialize the decoder.
| avctx | codec context |
dump the extradata
generic init
frame info
get frame len
subframe info
init previous block len
extract lfe channel position
calculate number of scale factor bands and their offsets for every possible block size
Scale factors can be shared between blocks of different size as every block has a different scale factor band layout. The matrix sf_offsets is needed to find the correct scale factor.
init MDCT, FIXME: only init needed sizes
init MDCT windows: simple sinus window
calculate subwoofer cutoff values
calculate sine values for the decorrelation matrix
Definition at line 274 of file wmaprodec.c.
| static int decode_packet | ( | AVCodecContext * | avctx, | |
| void * | data, | |||
| int * | got_frame_ptr, | |||
| AVPacket * | avpkt | |||
| ) | [static] |
Decode a single WMA packet.
| avctx | codec context | |
| data | the output buffer | |
| data_size | number of bytes that were written to the output buffer | |
| avpkt | input packet |
sanity check for the buffer length
parse packet header
get number of bits that need to be added to the previous frame
check for packet loss
append the previous frame data to the remaining data from the previous packet to create a full frame
decode the cross packet frame if it is valid
reset number of saved bits so that the decoder does not start to decode incomplete frames in the s->len_prefix == 0 case
when the frames do not have a length prefix, we don't know the compressed length of the individual frames however, we know what part of a new packet belongs to the previous frame therefore we save the incoming packet first, then we append the "previous frame" data from the next packet so that we get a buffer that only contains full frames
save the rest of the data so that it can be decoded with the next packet
Definition at line 1494 of file wmaprodec.c.
| static int decode_scale_factors | ( | WMAProDecodeCtx * | s | ) | [static] |
Extract scale factors from the bitstream.
| s | codec context |
should never consume more than 5344 bits MAX_CHANNELS * (1 + MAX_BANDS * 23)
resample scale factors for the new block size as the scale factors might need to be resampled several times before some new values are transmitted, a backup of the last transmitted scale factors is kept in saved_scale_factors
decode DPCM coded scale factors
run level decode differences to the resampled factors
swap buffers
calculate new scale factor maximum
Definition at line 881 of file wmaprodec.c.
Referenced by decode_subframe().
| static int decode_subframe | ( | WMAProDecodeCtx * | s | ) | [static] |
Decode a single subframe (block).
| s | codec context |
reset channel context and find the next block offset and size == the next block of the channel with the smallest number of decoded samples
get a list of all channels that contain the estimated block
substract already processed samples
and count if there are multiple subframes that match our profile
check if the frame will be complete after processing the estimated block
calculate number of scale factor bands and their offsets
configure the decoder for the current subframe
skip extended header if any
no idea for what the following bit is used
decode number of vector coded coefficients
decode quantization step
decode quantization step modifiers for every channel
decode scale factors
parse coefficients
reconstruct the per channel data
inverse quantization and rescaling
apply imdct (imdct_half == DCTIV with reverse)
window and overlapp-add
handled one subframe
Definition at line 1059 of file wmaprodec.c.
| static int decode_subframe_length | ( | WMAProDecodeCtx * | s, | |
| int | offset | |||
| ) | [static] |
Decode the subframe length.
| s | context | |
| offset | sample offset in the frame |
no need to read from the bitstream when only one length is possible
1 bit indicates if the subframe is of maximum length
sanity check the length
Definition at line 474 of file wmaprodec.c.
| static int decode_tilehdr | ( | WMAProDecodeCtx * | s | ) | [static] |
Decode how the data in the frame is split into subframes.
Every WMA frame contains the encoded data for a fixed number of samples per channel. The data for every channel might be split into several subframes. This function will reconstruct the list of subframes for every channel.
If the subframes are not evenly split, the algorithm estimates the channels with the lowest number of total samples. Afterwards, for each of these channels a bit is read from the bitstream that indicates if the channel contains a subframe with the next subframe size that is going to be read from the bitstream or not. If a channel contains such a subframe, the subframe size gets added to the channel's subframe list. The algorithm repeats these steps until the frame is properly divided between the individual channels.
| s | context |
< sum of samples for all currently known subframes of a channel
< flag indicating if a channel contains the current subframe
< number of channels that contain the current subframe
< flag indicating that all channels use the same subframe offsets and sizes
< smallest sum of samples (channels with this length will be processed first)
reset tiling information
loop until the frame data is split between the subframes
check which channels contain the subframe
get subframe length, subframe_len == 0 is not allowed
add subframes to the individual channels and find new min_channel_len
Definition at line 522 of file wmaprodec.c.
| static void av_cold dump_context | ( | WMAProDecodeCtx * | s | ) | [static] |
helper function to print the most important members of the context
| s | context |
Definition at line 239 of file wmaprodec.c.
| static void flush | ( | AVCodecContext * | avctx | ) | [static] |
Clear decoder buffers (for seeking).
| avctx | codec context |
reset output buffer as a part of it is used during the windowing of a new frame
Definition at line 1610 of file wmaprodec.c.
| static void inverse_channel_transform | ( | WMAProDecodeCtx * | s | ) | [static] |
Reconstruct the individual channel data.
| s | codec context |
multichannel decorrelation
multiply values with the decorrelation_matrix
Definition at line 974 of file wmaprodec.c.
Referenced by decode_subframe().
| static int remaining_bits | ( | WMAProDecodeCtx * | s, | |
| GetBitContext * | gb | |||
| ) | [static] |
Calculate remaining input buffer length.
| s | codec context | |
| gb | bitstream reader context |
Definition at line 1429 of file wmaprodec.c.
| static void save_bits | ( | WMAProDecodeCtx * | s, | |
| GetBitContext * | gb, | |||
| int | len, | |||
| int | append | |||
| ) | [static] |
Fill the bit reservoir with a (partial) frame.
| s | codec context | |
| gb | bitstream reader context | |
| len | length of the partial frame | |
| append | decides whether to reset the buffer or not |
when the frame data does not need to be concatenated, the input buffer is resetted and additional bits from the previous frame are copyed and skipped later so that a fast byte copy is possible
Definition at line 1441 of file wmaprodec.c.
| static void wmapro_window | ( | WMAProDecodeCtx * | s | ) | [static] |
Apply sine window and reconstruct the output buffer.
| s | codec context |
Definition at line 1029 of file wmaprodec.c.
Referenced by decode_subframe().
Initial value:
{
.name = "wmapro",
.type = AVMEDIA_TYPE_AUDIO,
.id = CODEC_ID_WMAPRO,
.priv_data_size = sizeof(WMAProDecodeCtx),
.init = decode_init,
.close = decode_end,
.decode = decode_packet,
.capabilities = CODEC_CAP_SUBFRAMES | CODEC_CAP_DR1,
.flush= flush,
.long_name = NULL_IF_CONFIG_SMALL("Windows Media Audio 9 Professional"),
}
Definition at line 1626 of file wmaprodec.c.
float sin64[33] [static] |
1.5.8