转：TS流的解析
TS即是"Transport Stream"的缩写。他是分包发送的，每一个包长为188字节。在TS流里可以填入很多类型的数据，如视频、音频、自定义信息等。他的包的结构为，包头为4个字节，负载为184个字节（这184个字节不一定都是有效数据，有一些可能为填充数据）。
工作形式：
因为在TS流里可以填入很多种东西，所以有必要有一种机制来确定怎么来标识这些数据。制定TS流标准的机构就规定了一些数据结构来定义。比如: PSI（Program Specific Information）表，所以解析起来就像这样: 先接收一个负载里为PAT的数据包，在整个数据包里找到一个PMT包的ID。然后再接收一个含有PMT的数据包，在这个数据包里找到有关填入数据类型的ID。之后就在接收到的TS包里找含有这个ID的负载内容，这个内容就是填入的信息。根据填入的数据类型的ID的不同，在TS流复合多种信息是可行的。关键就是找到标识的ID号。
因为在TS流里可以填入很多种东西，所以有必要有一种机制来确定怎么来标识这些数据。制定TS流标准的机构就规定了一些数据结构来定义。比如: PSI（Program Specific Information）表，所以解析起来就像这样: 先接收一个负载里为PAT的数据包，在整个数据包里找到一个PMT包的ID。然后再接收一个含有PMT的数据包，在这个数据包里找到有关填入数据类型的ID。之后就在接收到的TS包里找含有这个ID的负载内容，这个内容就是填入的信息。根据填入的数据类型的ID的不同，在TS流复合多种信息是可行的。关键就是找到标识的ID号。
现在以一个例子来说明具体的操作：
在开始之前先给出一片实际TS流例子：
0000f32ch: 47 40 00 17 00 00 B0 0D 00 01 C1 00 00 00 01 E0 ; G@....?..?...?
0000f33ch: 20 A2 C3 29 41 FF FF FF FF FF FF FF FF FF FF FF ; ⒚)A
0000f34ch: FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ;
0000f35ch: FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ;
0000f36ch: FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ;
0000f37ch: FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ;
0000f38ch: FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ;
0000f39ch: FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ;
0000f3ach: FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ;
0000f3bch: FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ;
0000f3cch: FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ;
0000f3dch: FF FF FF FF FF FF FF FF FF FF FF FF 47 40 20 17 ; G@ .
0000f3ech: 00 02 B0 1B 00 01 C1 00 00 E0 21 F0 00 1B E0 21 ; ..?..?.??.?
0000f3fch: F0 04 2A 02 7E 1F 03 E0 22 F0 00 5D 16 BD 48 ; ?*.~..??].紿
0000f32ch: 47 40 00 17 00 00 B0 0D 00 01 C1 00 00 00 01 E0 ; G@....?..?...?
0000f33ch: 20 A2 C3 29 41 FF FF FF FF FF FF FF FF FF FF FF ; ⒚)A
0000f34ch: FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ;
0000f35ch: FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ;
0000f36ch: FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ;
0000f37ch: FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ;
0000f38ch: FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ;
0000f39ch: FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ;
0000f3ach: FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ;
0000f3bch: FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ;
0000f3cch: FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF FF ;
0000f3dch: FF FF FF FF FF FF FF FF FF FF FF FF 47 40 20 17 ; G@ .
0000f3ech: 00 02 B0 1B 00 01 C1 00 00 E0 21 F0 00 1B E0 21 ; ..?..?.??.?
0000f3fch: F0 04 2A 02 7E 1F 03 E0 22 F0 00 5D 16 BD 48 ; ?*.~..??].紿
具体的分析就以这个例子来分析。
// Adjust TS packet header
void adjust_TS_packet_header(TS_packet_header* pheader)
{
unsigned char buf[4];
memcpy(buf, pheader, 4);
pheader->transport_error_indicator = buf[1] >> 7;
pheader->payload_unit_start_indicator = buf[1] >> 6 & 0x01;
pheader->transport_priority = buf[1] >> 5 & 0x01;
pheader->PID = (buf[1] & 0x1F) << 8 | buf[2];
pheader->transport_scrambling_control = buf[3] >> 6;
pheader->adaption_field_control = buf[3] >> 4 & 0x03;
pheader->continuity_counter = buf[3] & 0x03;
}
void adjust_TS_packet_header(TS_packet_header* pheader)
{
unsigned char buf[4];
memcpy(buf, pheader, 4);
pheader->transport_error_indicator = buf[1] >> 7;
pheader->payload_unit_start_indicator = buf[1] >> 6 & 0x01;
pheader->transport_priority = buf[1] >> 5 & 0x01;
pheader->PID = (buf[1] & 0x1F) << 8 | buf[2];
pheader->transport_scrambling_control = buf[3] >> 6;
pheader->adaption_field_control = buf[3] >> 4 & 0x03;
pheader->continuity_counter = buf[3] & 0x03;
}
这是一个调整TS流数据包头的函数，这里牵扯到位段调整的问题。现在看一下TS流数据包头的结构的定义：
// Transport packet header
typedef struct TS_packet_header
{
unsigned sync_byte : 8;
unsigned transport_error_indicator : 1;
unsigned payload_unit_start_indicator : 1;
unsigned transport_priority : 1;
unsigned PID : 13;
unsigned transport_scrambling_control : 2;
unsigned adaption_field_control : 2;
unsigned continuity_counter : 4;
} TS_packet_header;
typedef struct TS_packet_header
{
unsigned sync_byte : 8;
unsigned transport_error_indicator : 1;
unsigned payload_unit_start_indicator : 1;
unsigned transport_priority : 1;
unsigned PID : 13;
unsigned transport_scrambling_control : 2;
unsigned adaption_field_control : 2;
unsigned continuity_counter : 4;
} TS_packet_header;
下面我们来分析，在ISO/IEC 13818-1里有说明，PAT(Program Association Table)的PID值为0x00，TS包的标识(即sync_byte)为0x47，并且为了确保这个TS包里的数据有效，所以我们一开始查找47 40 00这三组16进制数，为什么这样？具体的奥秘在TS包的结构上，前面已经说了sync_byte固定为0x47。现在往下看transport_error_indicator、payload_unit_start_indicator、transport_priority和PID这四个元素，PID为0x00，这是PAT的标识。transport_error_indicator为0，transport_priority为0。把他们看成是两组8位16进制数就是：40 00。现在看看我们的TS流片断例子，看来正好是47 40 00开头的，一个TS流的头部占据了4个字节。剩下的负载部分的内容由PID来决定，例子看来就是一个PAT表。在这里有个地方需要注意一下，payload_unit_start_indicator为1时，在前4个字节之后会有一个调整字节，它的数值决定了负载内容的具体开始位置。现在看例子中的数据47 40 00 17 00第五个字节是00，说明紧跟着00之后就是具体的负载内容。
下面给出PAT表的结构体：
// PAT table
// Programm Association Table
typedef struct TS_PAT
{
unsigned table_id : 8;
unsigned section_syntax_indicator : 1;
unsigned zero : 1;
unsigned reserved_1 : 2;
unsigned section_length : 12;
unsigned transport_stream_id : 16;
unsigned reserved_2 : 2;
unsigned version_number : 5;
unsigned current_next_indicator : 1;
unsigned section_number : 8;
unsigned last_section_number : 8;
unsigned program_number : 16;
unsigned reserved_3 : 3;
unsigned network_PID : 13;
unsigned program_map_PID : 13;
unsigned CRC_32 : 32;
} TS_PAT;
// Programm Association Table
typedef struct TS_PAT
{
unsigned table_id : 8;
unsigned section_syntax_indicator : 1;
unsigned zero : 1;
unsigned reserved_1 : 2;
unsigned section_length : 12;
unsigned transport_stream_id : 16;
unsigned reserved_2 : 2;
unsigned version_number : 5;
unsigned current_next_indicator : 1;
unsigned section_number : 8;
unsigned last_section_number : 8;
unsigned program_number : 16;
unsigned reserved_3 : 3;
unsigned network_PID : 13;
unsigned program_map_PID : 13;
unsigned CRC_32 : 32;
} TS_PAT;
再给出PAT表字段调整函数：
// Adjust PAT table
void adjust_PAT_table ( TS_PAT * packet, char * buffer )
{
int n = 0, i = 0;
int len = 0;
packet->table_id = buffer[0];
packet->section_syntax_indicator = buffer[1] >> 7;
packet->zero = buffer[1] >> 6 & 0x1;
packet->reserved_1 = buffer[1] >> 4 & 0x3;
packet->section_length = (buffer[1] & 0x0F) << 8 | buffer[2];
packet->transport_stream_id = buffer[3] << 8 | buffer[4];
packet->reserved_2 = buffer[5] >> 6;
packet->version_number = buffer[5] >> 1 & 0x1F;
packet->current_next_indicator = (buffer[5] << 7) >> 7;
packet->section_number = buffer[6];
packet->last_section_number = buffer[7];
// Get CRC_32
len = 3 + packet->section_length;
packet->CRC_32 = (buffer[len-4] & 0x000000FF) << 24
| (buffer[len-3] & 0x000000FF) << 16
| (buffer[len-2] & 0x000000FF) << 8
| (buffer[len-1] & 0x000000FF);
// Parse network_PID or program_map_PID
for ( n = 0; n < packet->section_length - 4; n ++ )
{
packet->program_number = buffer[8] << 8 | buffer[9];
packet->reserved_3 = buffer[10] >> 5;
if ( packet->program_number == 0x0 )
packet->network_PID = (buffer[10] << 3) << 5 | buffer[11];
else
{
packet->program_map_PID = (buffer[10] << 3) << 5 | buffer[11];
}
n += 5;
}
}
void adjust_PAT_table ( TS_PAT * packet, char * buffer )
{
int n = 0, i = 0;
int len = 0;
packet->table_id = buffer[0];
packet->section_syntax_indicator = buffer[1] >> 7;
packet->zero = buffer[1] >> 6 & 0x1;
packet->reserved_1 = buffer[1] >> 4 & 0x3;
packet->section_length = (buffer[1] & 0x0F) << 8 | buffer[2];
packet->transport_stream_id = buffer[3] << 8 | buffer[4];
packet->reserved_2 = buffer[5] >> 6;
packet->version_number = buffer[5] >> 1 & 0x1F;
packet->current_next_indicator = (buffer[5] << 7) >> 7;
packet->section_number = buffer[6];
packet->last_section_number = buffer[7];
// Get CRC_32
len = 3 + packet->section_length;
packet->CRC_32 = (buffer[len-4] & 0x000000FF) << 24
| (buffer[len-3] & 0x000000FF) << 16
| (buffer[len-2] & 0x000000FF) << 8
| (buffer[len-1] & 0x000000FF);
// Parse network_PID or program_map_PID
for ( n = 0; n < packet->section_length - 4; n ++ )
{
packet->program_number = buffer[8] << 8 | buffer[9];
packet->reserved_3 = buffer[10] >> 5;
if ( packet->program_number == 0x0 )
packet->network_PID = (buffer[10] << 3) << 5 | buffer[11];
else
{
packet->program_map_PID = (buffer[10] << 3) << 5 | buffer[11];
}
n += 5;
}
}
通过上面的分析，例子中的数据00 B0 0D 00 01 C1 00 00 00 01 E0 20 A2 C3 29 41就是具体的PAT表的内容，然后根据PAT结构体来具体分析PAT表。但是我们需要注意的是在PAT表里有program_number、network_PID的元素不只有一个，这两个元素是通过循环来确定的。循环的次数通过section_length元素的确定。在这个例子中program_map_PID为20，所以下面来PMT分析时，就是查找47 40 20的开头的TS包。
下面来分析PMT表，先给出PMT(Program Map Table)的结构体：
// PMT table
// Program Map Table
typedef struct TS_PMT
{
unsigned table_id : 8;
unsigned section_syntax_indicator : 1;
unsigned zero : 1;
unsigned reserved_1 : 2;
unsigned section_length : 12;
unsigned program_number : 16;
unsigned reserved_2 : 2;
unsigned version_number : 5;
unsigned current_next_indicator : 1;
unsigned section_number : 8;
unsigned last_section_number : 8;
unsigned reserved_3 : 3;
unsigned PCR_PID : 13;
unsigned reserved_4 : 4;
unsigned program_info_length : 12;
unsigned stream_type : 8;
unsigned reserved_5 : 3;
unsigned elementary_PID : 13;
unsigned reserved_6 : 4;
unsigned ES_info_length : 12;
unsigned CRC_32 : 32;
} TS_PMT;
// Program Map Table
typedef struct TS_PMT
{
unsigned table_id : 8;
unsigned section_syntax_indicator : 1;
unsigned zero : 1;
unsigned reserved_1 : 2;
unsigned section_length : 12;
unsigned program_number : 16;
unsigned reserved_2 : 2;
unsigned version_number : 5;
unsigned current_next_indicator : 1;
unsigned section_number : 8;
unsigned last_section_number : 8;
unsigned reserved_3 : 3;
unsigned PCR_PID : 13;
unsigned reserved_4 : 4;
unsigned program_info_length : 12;
unsigned stream_type : 8;
unsigned reserved_5 : 3;
unsigned elementary_PID : 13;
unsigned reserved_6 : 4;
unsigned ES_info_length : 12;
unsigned CRC_32 : 32;
} TS_PMT;
在给出调整字段函数：
// Adjust PMT table
void adjust_PMT_table ( TS_PMT * packet, char * buffer )
{
int pos = 12, len = 0;
int i = 0;
packet->table_id = buffer[0];
packet->section_syntax_indicator = buffer[1] >> 7;
packet->zero = buffer[1] >> 6;
packet->reserved_1 = buffer[1] >> 4;
packet->section_length = (buffer[1] & 0x0F) << 8 | buffer[2];
packet->program_number = buffer[3] << 8 | buffer[4];
packet->reserved_2 = buffer[5] >> 6;
packet->version_number = buffer[5] >> 1 & 0x1F;
packet->current_next_indicator = (buffer[5] << 7) >> 7;
packet->section_number = buffer[6];
packet->last_section_number = buffer[7];
packet->reserved_3 = buffer[8] >> 5;
packet->PCR_PID = ((buffer[8] << 8) | buffer[9]) & 0x1FFF;
packet->reserved_4 = buffer[10] >> 4;
packet->program_info_length = (buffer[10] & 0x0F) << 8 | buffer[11];
// Get CRC_32
len = packet->section_length + 3;
packet->CRC_32 = (buffer[len-4] & 0x000000FF) << 24
| (buffer[len-3] & 0x000000FF) << 16
| (buffer[len-2] & 0x000000FF) << 8
| (buffer[len-1] & 0x000000FF);
// program info descriptor
if ( packet->program_info_length != 0 )
pos += packet->program_info_length;
// Get stream type and PID
for ( ; pos <= (packet->section_length + 2 ) - 4; )
{
packet->stream_type = buffer[pos];
packet->reserved_5 = buffer[pos+1] >> 5;
packet->elementary_PID = ((buffer[pos+1] << 8) | buffer[pos+2]) & 0x1FFF;
packet->reserved_6 = buffer[pos+3] >> 4;
packet->ES_info_length = (buffer[pos+3] & 0x0F) << 8 | buffer[pos+4];
// Store in es
es[i].type = packet->stream_type;
es[i].pid = packet->elementary_PID;
if ( packet->ES_info_length != 0 )
{
pos = pos+5;
pos += packet->ES_info_length;
}
else
{
pos += 5;
}
i++;
}
}
void adjust_PMT_table ( TS_PMT * packet, char * buffer )
{
int pos = 12, len = 0;
int i = 0;
packet->table_id = buffer[0];
packet->section_syntax_indicator = buffer[1] >> 7;
packet->zero = buffer[1] >> 6;
packet->reserved_1 = buffer[1] >> 4;
packet->section_length = (buffer[1] & 0x0F) << 8 | buffer[2];
packet->program_number = buffer[3] << 8 | buffer[4];
packet->reserved_2 = buffer[5] >> 6;
packet->version_number = buffer[5] >> 1 & 0x1F;
packet->current_next_indicator = (buffer[5] << 7) >> 7;
packet->section_number = buffer[6];
packet->last_section_number = buffer[7];
packet->reserved_3 = buffer[8] >> 5;
packet->PCR_PID = ((buffer[8] << 8) | buffer[9]) & 0x1FFF;
packet->reserved_4 = buffer[10] >> 4;
packet->program_info_length = (buffer[10] & 0x0F) << 8 | buffer[11];
// Get CRC_32
len = packet->section_length + 3;
packet->CRC_32 = (buffer[len-4] & 0x000000FF) << 24
| (buffer[len-3] & 0x000000FF) << 16
| (buffer[len-2] & 0x000000FF) << 8
| (buffer[len-1] & 0x000000FF);
// program info descriptor
if ( packet->program_info_length != 0 )
pos += packet->program_info_length;
// Get stream type and PID
for ( ; pos <= (packet->section_length + 2 ) - 4; )
{
packet->stream_type = buffer[pos];
packet->reserved_5 = buffer[pos+1] >> 5;
packet->elementary_PID = ((buffer[pos+1] << 8) | buffer[pos+2]) & 0x1FFF;
packet->reserved_6 = buffer[pos+3] >> 4;
packet->ES_info_length = (buffer[pos+3] & 0x0F) << 8 | buffer[pos+4];
// Store in es
es[i].type = packet->stream_type;
es[i].pid = packet->elementary_PID;
if ( packet->ES_info_length != 0 )
{
pos = pos+5;
pos += packet->ES_info_length;
}
else
{
pos += 5;
}
i++;
}
}
TS流可以复合很多的节目的视频和音频，但是解码器是怎么来区分的呢？答案就在PMT表里，如其名节目映射表。他就是来解决这个问题的。现在看PMT结构体里的stream_type、elementary_PID这两个元素，前一个用来确定后一个作为标识PID的内容具体是什么，音频或视频等。还有要注意他们不只有一个，所以他们是通过循环读取来确保所有的值都被读取了，当然循环也是有规定的(具体看调整函数上)。从例子上来看，我们在倒数第三行找到了上面分析来的PMT表的PID为0x20的TS包。然后就可以把数据是用调整函数填入结构中。然后得到具体节目的PID为视频0x21, 音频0x22。
PS. 文章里的PID是用来判断具体TS包是什么包的。分析每个包得到的PID值，都可以复合在TS头部结构体的PID里。
