Add JOC downmix configurations 0..4, cross-layer checks, and drop the native core

This commit is contained in:
TheM14 committed 2026-10-16 19:46:24 +08:00
1 parent 428772eb87
commit a1853f2dc5
37 files changed
+926 -6420

No files matched your search

+193 -13
View File
@@ -73,6 +73,9 @@ class EmdfContainer:
raw: bytes
payloads: dict
sample_offsets: dict
# payload_id -> group_id(未声明 groupid 的载荷为 None)。规格 §8.2/Table 56:
# 同一帧的 OAMD 与 JOC 载荷属于同一 group,凭此把一对载荷关联起来。
group_ids: dict = None
def _parse_at(data, start_bit):
@@ -99,6 +102,7 @@ def _parse_at(data, start_bit):
payloads = {}
sample_offsets = {}
group_ids = {}
terminated = False
while reader.position + 5 <= body_end:
payload_id = reader.read(5)
@@ -115,7 +119,9 @@ def _parse_at(data, start_bit):
if reader.read(1):
variable_bits(reader, 11) # duration
if reader.read(1):
variable_bits(reader, 2) # group id
group_ids[payload_id] = variable_bits(reader, 2) # group id
else:
group_ids[payload_id] = None
if reader.read(1):
reader.skip(8) # codec data
@@ -140,7 +146,7 @@ def _parse_at(data, start_bit):
total_bytes = 4 + length
raw_reader = BitReader(data, start_bit, start_bit + total_bytes * 8)
raw = raw_reader.read_bytes(total_bytes)
return EmdfContainer(start_bit, raw, payloads, sample_offsets)
return EmdfContainer(start_bit, raw, payloads, sample_offsets, group_ids)
def _marker_offsets(data):
@@ -162,13 +168,18 @@ def _marker_offsets(data):
return sorted(offsets)
def find_joc_emdf(frame):
def find_joc_emdf(frame, usable=None, diagnostics=None, frame_number=None):
"""返回同步帧中唯一、顶层连续且包含 ID11/ID14 的 JOC EMDF 容器。
EMDF payload 是不透明字节串,其中可能自然出现另一个 ``0x5838``。若从这个
内嵌 marker 开始的后续随机位恰好也能通过容器语法探测,它仍不是一个独立的
transport 容器。因此,候选的起点一旦落在较早 JOC 容器的声明范围内,就只把
它记作内嵌伪候选,不参与“多个容器”的判定。
一个同步帧里可能出现多个顶层容器(同一帧可能同时带两套 ID11/ID14)。
``usable`` 给出“这个容器能否作为本帧元数据”的分值(0 = 不是,越大越可信):
只有唯一最高分时用它,其余记入 ``diagnostics``;没有候选或最高分并列仍按变体
报错,不猜。
"""
matches = []
offsets = _marker_offsets(frame)
@@ -216,6 +227,22 @@ def find_joc_emdf(frame):
})
if len(top_level_matches) != 1:
starts = [item.start_bit for item in top_level_matches]
if usable is not None:
scored = [(usable(item), item) for item in top_level_matches]
best = max(score for score, _item in scored)
winners = [item for score, item in scored if score == best]
if best > 0 and len(winners) == 1:
if diagnostics is not None:
diagnostics.append({
"stage": "emdf",
"variant": "extra_container_ignored",
"frame": frame_number,
"message": "同步帧内有多个含 ID11/ID14 的 EMDF 容器,只有一个能作为本帧元数据",
"details": {"joc_container_start_bits": starts,
"used_start_bit": winners[0].start_bit,
"pair_score": best},
})
return winners[0]
raise UnsupportedVariantError(
"emdf_transport", "multiple_joc_containers",
"同步帧中存在多个可用 JOC EMDF,当前无法自动选择",
@@ -235,31 +262,184 @@ def parse_container(data):
return container
def iter_eac3_frames(data):
"""按 E-AC-3 ``frmsiz`` 遍历同步帧,拒绝静默重同步。"""
# AC-3 同步帧长度表(16-bit word),索引为 fscod(0=48 kHz、1=44.1 kHz、2=32 kHz)
# 再取 frmsizecod;与 A/52 的比特率表等价。
_AC3_FRAME_WORDS = (
(64, 64, 80, 80, 96, 96, 112, 112, 128, 128, 160, 160, 192, 192, 224, 224,
256, 256, 320, 320, 384, 384, 448, 448, 512, 512, 640, 640, 768, 768,
896, 896, 1024, 1024, 1152, 1152, 1280, 1280),
(69, 70, 87, 88, 104, 105, 121, 122, 139, 140, 174, 175, 208, 209, 243, 244,
278, 278, 348, 348, 417, 418, 487, 487, 557, 557, 696, 696, 835, 835,
975, 975, 1114, 1114, 1253, 1253, 1393, 1393),
(96, 96, 120, 120, 144, 144, 168, 168, 192, 192, 240, 240, 288, 288, 336, 336,
384, 384, 480, 480, 576, 576, 672, 672, 768, 768, 960, 960, 1152, 1152,
1344, 1344, 1536, 1536, 1728, 1728, 1920, 1920),
)
# E-AC-3 chanmap(TS 102 366 Annex E Table E2.5):字段的最高位对应 bit 0。
CHANMAP_LOCATIONS = (
"L", "C", "R", "Ls", "Rs", "Lc/Rc", "Lrs/Rrs", "Cs", "Ts", "Lsd/Rsd",
"Lw/Rw", "Vhl/Vhr", "Vhc", "Lts/Rts", "LFE2", "LFE",
)
# 单个 bit 展开为两个相邻声道的成对位置;展开顺序为先左后右。
CHANMAP_PAIR_BITS = frozenset((5, 6, 9, 10, 11, 13))
# A/52 acmod 对应的全频段声道位置(不含 LFE,按解码顺序)。
ACMOD_LOCATIONS = {
0: ("C",), # 1+1 双单声道
1: ("C",),
2: ("L", "R"),
3: ("L", "C", "R"),
4: ("L", "R", "Cs"),
5: ("L", "C", "R", "Cs"),
6: ("L", "R", "Ls", "Rs"),
7: ("L", "C", "R", "Ls", "Rs"),
}
def expand_chanmap(chanmap):
"""按 Table E2.5 把 16 bit chanmap 展开为有序声道位置列表。"""
locations = []
for bit in range(16):
if not (int(chanmap) >> (15 - bit)) & 1:
continue
locations.extend(CHANMAP_LOCATIONS[bit].split("/"))
return locations
@dataclass(frozen=True)
class SyncFrame:
"""一个同步帧及其传输层(L2)字段。"""
offset: int
size: int
codec: str # "ac3" 或 "eac3"
bsid: int
strmtyp: int # AC-3 没有子流类型,固定记为 0
substreamid: int
acmod: int
lfeon: int
chanmap: object # 依赖子流的 16 bit chanmap;未声明为 None
@property
def key(self):
"""同一音频帧内唯一的子流标识。"""
return (self.codec, self.strmtyp, self.substreamid)
def _read_bits(data, start, count):
value = 0
for index in range(start, start + count):
value = (value << 1) | ((data[index >> 3] >> (7 - (index & 7))) & 1)
return value
def sync_frame_at(data, offset):
"""解析 offset 处同步帧的传输层头。
区分 AC-3 与 E-AC-3 的依据是 bsid(bit 40 起 5 bit):bsid <= 8 为 AC-3,
其帧长由 fscod/frmsizecod 的比特率表给出;bsid >= 16 为 E-AC-3,帧长由
frmsiz 给出。依赖子流(strmtyp=1)在这里顺带解析出 chanmap。
"""
if offset + 6 > len(data) or data[offset] != 0x0B or data[offset + 1] != 0x77:
return None
base = offset * 8
bsid = _read_bits(data, base + 40, 5)
if bsid <= 8:
fscod = _read_bits(data, base + 32, 2)
frmsizecod = _read_bits(data, base + 34, 6)
if fscod > 2 or frmsizecod >= 38:
return None
size = _AC3_FRAME_WORDS[fscod][frmsizecod] * 2
acmod = _read_bits(data, base + 48, 3)
bit = 51
if acmod & 0x1 and acmod != 0x1:
bit += 2 # cmixlev
if acmod & 0x4:
bit += 2 # surmixlev
if acmod == 0x2:
bit += 2 # dsurmod
lfeon = _read_bits(data, base + bit, 1)
return SyncFrame(offset, size, "ac3", bsid, 0, 0, acmod, lfeon, None)
strmtyp = _read_bits(data, base + 16, 2)
substreamid = _read_bits(data, base + 18, 3)
frmsiz = _read_bits(data, base + 21, 11)
# fscod==3 时的 fscod2 同样是 2 bit,因此 acmod 恒在 bit 36。
acmod = _read_bits(data, base + 36, 3)
lfeon = _read_bits(data, base + 39, 1)
bit = 45 # bsid 之后
bit += 5 # dialnorm
compre = _read_bits(data, base + bit, 1)
bit += 1
if compre:
bit += 8 # compr
if acmod == 0:
bit += 5 # dialnorm2
compr2 = _read_bits(data, base + bit, 1)
bit += 1
if compr2:
bit += 8
chanmap = None
if strmtyp == 1:
if _read_bits(data, base + bit, 1):
chanmap = _read_bits(data, base + bit + 1, 16)
return SyncFrame(offset, (frmsiz + 1) * 2, "eac3", bsid, strmtyp, substreamid,
acmod, lfeon, chanmap)
def iter_syncframes(data):
"""按同步字长度遍历输入中的 AC-3 与 E-AC-3 同步帧。"""
pos = 0
while pos < len(data):
if pos + 4 > len(data) or data[pos:pos + 2] != b"\x0b\x77":
frame = sync_frame_at(data, pos)
if frame is None:
raise UnsupportedVariantError(
"eac3_transport", "syncframe_header",
"E-AC-3 同步帧头无效或出现了未处理的子流排列",
"同步帧头无效或出现了未处理的子流排列",
details={
"byte_offset": pos,
"remaining_bytes": len(data) - pos,
"next_16_bytes_hex": data[pos:pos + 16].hex(),
})
size = ((((data[pos + 2] & 7) << 8) | data[pos + 3]) + 1) * 2
if pos + size > len(data):
if pos + frame.size > len(data):
raise UnsupportedVariantError(
"eac3_transport", "truncated_syncframe",
"E-AC-3 末帧长度超过输入剩余数据",
"末帧长度超过输入剩余数据",
details={
"byte_offset": pos,
"declared_frame_bytes": size,
"codec": frame.codec,
"declared_frame_bytes": frame.size,
"remaining_bytes": len(data) - pos,
})
yield data[pos:pos + size]
pos += size
yield frame
pos += frame.size
def audio_frame_groups(frames):
"""把同步帧序列按音频帧分组。
一个音频帧内每个 (编码, strmtyp, substreamid) 组合只出现一次,因此再次遇到
已出现过的组合即表示下一个音频帧开始。独立子流缺席(只有依赖子流)或核心是
AC-3 时同样成立。
"""
group = []
keys = set()
for frame in frames:
if frame.key in keys:
yield group
group = []
keys = set()
keys.add(frame.key)
group.append(frame)
if group:
yield group
def iter_eac3_frames(data):
"""兼容入口:只产出同步帧字节。"""
for frame in iter_syncframes(data):
yield data[frame.offset:frame.offset + frame.size]
def extract_index(eac3_path, output_dir, max_frames=None):