Add Sparse JOC decoding support.

README.en.md

@@ -288,7 +288,6 @@ See the [mathematical notes](docs/math.en.md) for the equations used by the deco
 ## Known limitations
 
 - Only the common contiguous EMDF transport is covered. Fragmented transport across multiple audio-block skip fields is not covered.
-- Dense JOC is the main path. The Sparse JOC branch should not be treated as supported.
 - The speaker and SOFA binaural paths currently cover ordinary point objects; extent, spread, diffuse, divergence, channel lock, and similar controls are outside the supported scope.
 - OAMD trim elements are boundary-checked and skipped; warp, balance, and trim parameters are not applied to raw object trajectories or speaker rendering.
 - Multi-data-point streams, uncommon band configurations, and unusual OAMD scheduling have less coverage than common 12-band, single-data-point material.

README.md

@@ -258,7 +258,6 @@ JustOneCacophony/
 ## 已知限制
 
 - 当前只覆盖常见 continuous EMDF transport；跨多个 audio-block skip field 的碎片化 transport 尚未覆盖。
-- Dense JOC 是当前主要路径；Sparse JOC 分支不应视为受支持能力。
 - 扬声器与 SOFA 双耳路径当前只覆盖普通点对象；extent、spread、diffuse、divergence、channel lock 等对象控制不在支持范围内。
 - OAMD trim element 会按声明边界校验并跳过；warp、balance 和 trim 参数不应用于当前原始对象轨迹或扬声器渲染。
 - 多数据点、少见参数带配置和特殊 OAMD 调度的覆盖度低于常见 12-band、单数据点素材。

docs/math.en.md

@@ -4,7 +4,7 @@
 
 This document covers only the signal model and formulas used in the JustOneCacophony research path: how JOC parameters combine with core PCM to reconstruct object signals, and how OAMD coordinates become speaker gains.
 
-The formulas describe the dense-JOC and ordinary point-object paths studied by the project. They are not a complete definition of every E-AC-3 JOC variant.
+The formulas describe the JOC matrix parameters (both the dense and the sparse differential syntax) and the ordinary point-object paths studied by the project. They are not a complete definition of every E-AC-3 JOC variant.
 
 ## 1. Overall path and notation
 
@@ -52,9 +52,11 @@ $$
 N_f=1536=24\times64.
 $$
 
-## 2. Dense-JOC matrix parameters
+## 2. JOC matrix parameters
 
-### 2.1 Differential reconstruction
+For every object and data point, the quantized matrix `joc_mix_mtx_q` is defined on $N_q$ quantization levels. The `b_joc_sparse` flag selects one of two differential syntaxes: dense sends one MTX difference per core channel, while sparse sends one active channel plus one coefficient difference per parameter band.
+
+### 2.1 Dense differential reconstruction
 
 Let `quant_idx` be $q_i\in\{0,1\}$. The number of quantization levels is
 
@@ -85,7 +87,49 @@ Q_{o,d,c,p}=
 \qquad p>0.
 $$
 
-### 2.2 Dequantization
+### 2.2 Sparse differential reconstruction
+
+Let $I_{o,d,p}$ be the `joc_channel_idx` symbol (IDX), $V_{o,d,p}$ the `joc_vec` symbol (VEC), and $N_c\in\{5,7\}$ the number of core channels. Each parameter band has exactly one active channel:
+
+$$
+A_{o,d,p}=
+\begin{cases}
+I_{o,d,0}, & p=0,\\[2pt]
+\left(A_{o,d,p-1}+I_{o,d,p}\right)\bmod N_c, & p>0,
+\end{cases}
+$$
+
+where $I_{o,d,0}$ is a 3-bit absolute channel index and every later IDX symbol is an increment relative to the previous **active channel**. The coefficient is a single accumulator running across parameter bands:
+
+$$
+\kappa_{o,d,-1}=O^{(s)}_q,\qquad
+\kappa_{o,d,p}=
+\left(\kappa_{o,d,p-1}+V_{o,d,p}\right)\bmod N_q,
+$$
+
+with a sparse starting point two quantization levels above the dense center offset:
+
+$$
+O^{(s)}_q=
+\begin{cases}
+50, & q_i=0,\\
+100, & q_i=1.
+\end{cases}
+$$
+
+The accumulator is **not** reset when the active channel changes. The complete matrix is
+
+$$
+Q_{o,d,c,p}=
+\begin{cases}
+\kappa_{o,d,p}, & c=A_{o,d,p},\\[2pt]
+\dfrac{N_q}{2}, & c\neq A_{o,d,p}.
+\end{cases}
+$$
+
+Non-active entries take $N_q/2$, which dequantizes to exactly 0.
+
+### 2.3 Dequantization
 
 The dequantized matrix coefficient is
 
@@ -97,7 +141,7 @@ $$
 
 The effective denominator is therefore 4096 in coarse mode and 8192 in fine mode.
 
-### 2.3 JOC clipgain
+### 2.4 JOC clipgain
 
 If the clipgain field consists of integer $x$ and mantissa $y$, then
 
@@ -557,7 +601,7 @@ $$
 
 ## 14. Scope of the formulas
 
-- The JOC matrix section describes dense JOC; Sparse JOC uses a different sparse coefficient/index path.
+- The JOC matrix section covers both the dense MTX and the sparse IDX/VEC differential syntax.
 - The speaker-panning section describes ordinary point objects; extent, spread, divergence, and similar modes require additional models.
 - Multiple OAMD position blocks must be scheduled in time order.
 - A limiter is separate post-processing and is not included in the mixing equations above.

docs/math.md

@@ -4,7 +4,7 @@
 
 本文只说明 JustOneCacophony 研究路径中使用的信号模型和公式：JOC 参数如何与核心 PCM 结合并重建对象信号，以及 OAMD 坐标如何转换为扬声器增益。
 
-这些公式描述项目当前研究的 dense JOC 与普通点对象路径，不代表对所有 E-AC-3 JOC 变体的完整定义。
+这些公式描述项目当前研究的 JOC 矩阵参数（dense 与 sparse 两条差分语法）与普通点对象路径，不代表对所有 E-AC-3 JOC 变体的完整定义。
 
 ## 1. 总体路径与记号
 
@@ -52,9 +52,11 @@ $$
 N_f=1536=24\times64.
 $$
 
-## 2. Dense JOC 矩阵参数
+## 2. JOC 矩阵参数
 
-### 2.1 差分还原
+每个对象、每个数据点的量化矩阵 `joc_mix_mtx_q` 都定义在 $N_q$ 个量化级上。标志位 `b_joc_sparse` 选择两条差分语法之一：dense 为每个核心声道各送一路 MTX 差分，sparse 每参数带只送一个 active 声道与一路系数差分。
+
+### 2.1 Dense 差分还原
 
 令 `quant_idx` 为 $q_i\in\{0,1\}$，量化级数为
 
@@ -85,7 +87,49 @@ Q_{o,d,c,p}=
 \qquad p>0.
 $$
 
-### 2.2 去量化
+### 2.2 Sparse 差分还原
+
+令 $I_{o,d,p}$ 为 `joc_channel_idx` 符号（IDX），$V_{o,d,p}$ 为 `joc_vec` 符号（VEC），$N_c\in\{5,7\}$ 为核心声道数。每参数带只有一个 active 声道
+
+$$
+A_{o,d,p}=
+\begin{cases}
+I_{o,d,0}, & p=0,\\[2pt]
+\left(A_{o,d,p-1}+I_{o,d,p}\right)\bmod N_c, & p>0,
+\end{cases}
+$$
+
+其中 $I_{o,d,0}$ 是 3 bit 绝对声道号，其余 IDX 符号是相对上一个 **active 声道**的增量。系数是一个跨参数带连续的单累加器
+
+$$
+\kappa_{o,d,-1}=O^{(s)}_q,\qquad
+\kappa_{o,d,p}=
+\left(\kappa_{o,d,p-1}+V_{o,d,p}\right)\bmod N_q,
+$$
+
+sparse 起点比 dense 的中心偏移高两个量化级：
+
+$$
+O^{(s)}_q=
+\begin{cases}
+50, & q_i=0,\\
+100, & q_i=1.
+\end{cases}
+$$
+
+active 声道切换时累加器**不**重置。完整矩阵为
+
+$$
+Q_{o,d,c,p}=
+\begin{cases}
+\kappa_{o,d,p}, & c=A_{o,d,p},\\[2pt]
+\dfrac{N_q}{2}, & c\neq A_{o,d,p}.
+\end{cases}
+$$
+
+非 active 项取 $N_q/2$，即去量化后恰为 0。
+
+### 2.3 去量化
 
 矩阵系数的去量化值为
 
@@ -97,7 +141,7 @@ $$
 
 因此 coarse 模式的有效分母为 4096，fine 模式为 8192。
 
-### 2.3 JOC clipgain
+### 2.4 JOC clipgain
 
 若 clipgain 字段由整数 $x$ 和尾数 $y$ 组成，则
 
@@ -557,7 +601,7 @@ $$
 
 ## 14. 公式适用范围
 
-- JOC 矩阵部分描述 dense JOC；Sparse JOC 使用不同的稀疏系数/索引路径。
+- JOC 矩阵部分同时描述 dense MTX 与 sparse IDX/VEC 两条差分语法。
 - 扬声器声像部分描述普通点对象；extent、spread、divergence 等模式需要额外模型。
 - 多个 OAMD position block 必须按其时间顺序调度。
 - limiter 属于独立后处理，不包含在上述混音公式中。

docs/native.en.md

@@ -42,7 +42,7 @@ int ejoc_renderer_process(
     float* output16_planar);       /* [16][1536] */
 ```
 
-Python performs dense-JOC Huffman decoding, differential reconstruction, and dequantization before the call. Sparse JOC is not silently passed to the dense native path.
+Python performs JOC Huffman decoding, differential reconstruction, and dequantization before the call, with the dense and sparse syntaxes sharing one entry point. The native core consumes the already dequantized `dq` in double precision, and both syntaxes have the same layout at that ABI.
 
 Thread control is exposed as:
 

docs/native.md

@@ -42,7 +42,7 @@ int ejoc_renderer_process(
     float* output16_planar);       /* [16][1536] */
 ```
 
-Dense JOC 的 Huffman 解码、差分还原和去量化先在 Python 中完成。Sparse JOC 不会被静默送入 dense 原生路径。
+JOC 的 Huffman 解码、差分还原和去量化先在 Python 中完成，dense 与 sparse 两条语法共用同一条入口。原生核心消费已去量化的 `dq`（double），两条语法在该 ABI 上布局一致。
 
 线程接口为：
 

native/include/eac3joc_core.h

@@ -53,8 +53,9 @@ Fixed array layouts used by ejoc_renderer_process():
   output16     [16][1536]
 
 Only objects selected by object_mask are read from the descriptor arrays.
-Sparse JOC must be rejected by the caller; this ABI accepts already dequantized
+dq carries already dequantized matrix coefficients in double precision; the
-dense matrix coefficients.
+caller performs the JOC bitstream differential decoding for both dense and
+sparse objects, so this ABI is identical for both syntaxes.
 */
 
 EJOC_API uint32_t EJOC_CALL ejoc_abi_version(void);

src/joc_decode.py

@@ -2,11 +2,11 @@
 
 范围：
   - EMDF ID14 的 joc_header、joc_info 和 Huffman joc_data；
-  - 差分解码得到 joc_mix_mtx_q；
+  - 差分还原得到 joc_mix_mtx_q（dense MTX 与 sparse IDX/VEC 两条语法）；
   - 去量化得到 joc_mix_mtx_dq；
   - 位流自洽验证（joc_data 后剩余 = padding_bits 0..7 + 可能 joc_ext_data）
 后续的时间插值、QMF/时域重建和 ``joc_clipgain`` 位于 ``renderer.py``。
-Sparse 分支仍缺少实际样本验证。
+公式与符号定义见 ``docs/math.md`` 第 2 节。
 """
 from pathlib import Path
 
@@ -23,6 +23,7 @@ _HUFF_NAMES = (
     "joc_huff_code_7ch_pos_index_sparse",
 )
 
+
 def _load_huff_tables():
     with np.load(_TABLES_PATH) as tables:
         return {
@@ -30,18 +31,24 @@ def _load_huff_tables():
             for name in _HUFF_NAMES
         }
 
+
 H = _load_huff_tables()
 
 JOC_NUM_CHANNELS = {0: 5, 1: 7, 2: 7, 3: 5, 4: 7}          # Table 33
 JOC_NUM_BANDS = {0: 1, 1: 3, 2: 5, 3: 7, 4: 9, 5: 12, 6: 15, 7: 23}  # Table 35
-_PUBLIC_TABLE39_EXCERPT_UNUSED = {  # 仅保留作表格差异说明；渲染映射在 joc_qmf.py。
+JOC_NUM_QUANT = {0: 96, 1: 192}                            # Table 51
-    23: [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22,22],
+# dense 的量化零点就是 nquant/2；sparse 的递推起点比它高 2 个量化步。
-}
+JOC_DENSE_OFFSET = {0: 48, 1: 96}
+JOC_SPARSE_OFFSET = {0: 50, 1: 100}
+
 
 class BR:
+    """MSB-first 位读取器；位置以载荷内的 bit offset 表示。"""
+
     def __init__(self, data, pos=0):
         self.d = data
         self.p = pos
+
     def bits(self, n):
         v = 0
         for _ in range(n):
@@ -53,18 +60,19 @@ class BR:
 def huff_decode(tree, br):
     node = 0
     while node >= 0:
-        b = br.bits(1)
+        node = tree[node][br.bits(1)]
-        node = tree[node][b]
     return -node - 1
 
 
 def get_huff_code(mode, typ, nch):
     if typ == "IDX":
-        return H["joc_huff_code_5ch_pos_index_sparse" if nch == 5 else "joc_huff_code_7ch_pos_index_sparse"]
+        return H["joc_huff_code_5ch_pos_index_sparse" if nch == 5
+                 else "joc_huff_code_7ch_pos_index_sparse"]
     if typ == "VEC":
-        return H["joc_huff_code_coarse_coeff_sparse" if mode == 0 else "joc_huff_code_fine_coeff_sparse"]
+        return H["joc_huff_code_coarse_coeff_sparse" if mode == 0
-    # MTX
+                 else "joc_huff_code_fine_coeff_sparse"]
-    return H["joc_huff_code_coarse_generic" if mode == 0 else "joc_huff_code_fine_generic"]
+    return H["joc_huff_code_coarse_generic" if mode == 0
+             else "joc_huff_code_fine_generic"]           # MTX
 
 
 def parse_joc(payload):
@@ -76,6 +84,8 @@ def parse_joc(payload):
     out["ext_config_idx"] = br.bits(3)
     n_objects = out["num_objects_bits"] + 1
     n_channels = JOC_NUM_CHANNELS.get(out["dmx_config_idx"])
+    if n_channels is None:
+        raise ValueError(f"未知的 JOC downmix 配置 {out['dmx_config_idx']}")
     out["n_objects"], out["n_channels"] = n_objects, n_channels
     out["clipgain_x_bits"] = br.bits(3)
     out["clipgain_y_bits"] = br.bits(5)
@@ -98,78 +108,120 @@ def parse_joc(payload):
                 o["offset_ts"] = [br.bits(5) + 1 for _ in range(o["n_dpoints"])]
         objs.append(o)
     out["objs"] = objs
-    # joc_data（Huffman）
+    # joc_data（Huffman）：dense 逐声道逐带读 MTX；sparse 读 IDX 后读 VEC。
     for obj, o in enumerate(objs):
         if not o["present"]:
             continue
-        nquant = 96 if o["quant_idx"] == 0 else 192
         o["channel_idx"] = []
         o["vec"] = []
         o["mtx"] = []
         for dp in range(o["n_dpoints"]):
             if o["sparse"] == 1:
-                # Sparse JOC 使用 VEC/IDX Huffman 树；此分支尚无真实码流验证。
+                tree = get_huff_code(o["quant_idx"], "IDX", n_channels)
-                ci0 = br.bits(3)
+                idx = [br.bits(3)]
-                tree = get_huff_code(n_channels, "IDX", n_channels)
+                idx += [huff_decode(tree, br) for _ in range(o["n_bands"] - 1)]
-                ci = [ci0] + [huff_decode(tree, br) for _ in range(o["n_bands"] - 1)]
-                o["channel_idx"].append(ci)
                 tree = get_huff_code(o["quant_idx"], "VEC", n_channels)
-                vec = [huff_decode(tree, br) for _ in range(o["n_bands"])]
+                o["channel_idx"].append(idx)
-                o["vec"].append(vec)
+                o["vec"].append([huff_decode(tree, br) for _ in range(o["n_bands"])])
+                o["mtx"].append(None)
             else:
                 tree = get_huff_code(o["quant_idx"], "MTX", n_channels)
                 mtx = [[huff_decode(tree, br) for _ in range(o["n_bands"])]
                        for _ in range(n_channels)]
                 o["mtx"].append(mtx)
+                o["channel_idx"].append(None)
+                o["vec"].append(None)
     out["data_end_bits"] = br.p
     out["remaining_bits"] = len(payload) * 8 - br.p
     out["tail_bytes"] = payload[br.p // 8:]
     return out
 
 
+def reconstruct_dense(o, dp, n_ch):
+    """Dense 差分还原 → 量化矩阵 ``[ch][pb]``。
+
+    每个核心声道各自从 ``nquant/2``（去量化 0）出发，沿参数带累加 MTX 符号。
+    """
+    nquant = JOC_NUM_QUANT[o["quant_idx"]]
+    offset = JOC_DENSE_OFFSET[o["quant_idx"]]
+    mtx = o["mtx"][dp]
+    if mtx is None:
+        raise ValueError("Dense JOC 对象缺少 MTX 符号")
+    q = np.zeros((n_ch, o["n_bands"]), dtype=np.int64)
+    for ch in range(n_ch):
+        q[ch][0] = (offset + mtx[ch][0]) % nquant
+        for pb in range(1, o["n_bands"]):
+            q[ch][pb] = (q[ch][pb - 1] + mtx[ch][pb]) % nquant
+    return q
+
+
+def reconstruct_sparse(o, dp, n_ch):
+    """Sparse 差分还原 → 量化矩阵 ``[ch][pb]``。
+
+    每个参数带只有一个 active 声道：
+      - ``active[0]`` 是 3 bit 绝对声道号，其后由 IDX 符号累加得到，
+        因此递推锚点是**已重建**的 active 声道，而不是编码器送的符号本身；
+      - 系数是一个跨参数带连续的单累加器（active 声道切换时**不**重置），
+        起点为 sparse offset，增量为 VEC 符号；
+      - 非 active 项取 ``nquant/2``，即去量化后的 0。
+
+    IDX 符号取值恒为 ``0..n_ch-1``（Huffman 叶数即声道数），
+    故 ``(active + idx) % n_ch`` 与单次条件减等价。
+    """
+    if n_ch not in (5, 7):
+        raise ValueError(f"Sparse JOC 需要 5 或 7 个核心声道，实际 {n_ch}")
+    nquant = JOC_NUM_QUANT[o["quant_idx"]]
+    offset = JOC_SPARSE_OFFSET[o["quant_idx"]]
+    n_bands = o["n_bands"]
+    idx = o["channel_idx"][dp]
+    vec = o["vec"][dp]
+    if idx is None or vec is None:
+        raise ValueError("Sparse JOC 对象缺少 channel_idx/vec 符号")
+    if len(idx) != n_bands or len(vec) != n_bands:
+        raise ValueError(
+            f"Sparse JOC 维度不符: bands={n_bands}, idx={len(idx)}, vec={len(vec)}")
+    if not 0 <= idx[0] < n_ch:
+        raise ValueError(f"Sparse JOC 初始声道 {idx[0]} 超出 {n_ch} 个声道")
+
+    q = np.full((n_ch, n_bands), nquant // 2, dtype=np.int64)
+    active = idx[0]
+    coefficient = offset
+    for pb in range(n_bands):
+        if pb:
+            active = (active + idx[pb]) % n_ch
+        coefficient = (coefficient + vec[pb]) % nquant
+        q[active][pb] = coefficient
+    return q
+
+
 def diff_decode(out):
-    """6.6.2：差分解码 → joc_mix_mtx_q[obj][dp][ch][pb]。"""
+    """差分还原 → joc_mix_mtx_q[obj][dp][ch][pb]。"""
     mix_q = {}
     n_ch = out["n_channels"]
     for obj, o in enumerate(out["objs"]):
         if not o["present"]:
             continue
-        nquant = 96 if o["quant_idx"] == 0 else 192
         q = np.zeros((o["n_dpoints"], n_ch, o["n_bands"]), dtype=np.int64)
         for dp in range(o["n_dpoints"]):
             if o["sparse"] == 1:
-                # Sparse 差分路径尚无真实码流验证。
+                q[dp] = reconstruct_sparse(o, dp, n_ch)
-                offset = 50 if o["quant_idx"] == 0 else 100
-                ci = o["channel_idx"][dp]
-                vec = o["vec"][dp]
-                for pb in range(o["n_bands"]):
-                    ci_mod = ci[0] if pb == 0 else (ci[pb - 1] + ci[pb]) % n_ch
-                    for ch in range(n_ch):
-                        if ch == ci_mod:
-                            if pb == 0:
-                                q[dp][ch][pb] = (offset + vec[pb]) % nquant
-                            else:
-                                q[dp][ch][pb] = (q[dp][ch][pb - 1] + vec[pb]) % nquant
-                        else:
-                            q[dp][ch][pb] = offset
             else:
-                offset = 48 if o["quant_idx"] == 0 else 96
+                q[dp] = reconstruct_dense(o, dp, n_ch)
-                mtx = o["mtx"][dp]
-                for ch in range(n_ch):
-                    q[dp][ch][0] = (offset + mtx[ch][0]) % nquant
-                    for pb in range(1, o["n_bands"]):
-                        q[dp][ch][pb] = (q[dp][ch][pb - 1] + mtx[ch][pb]) % nquant
         mix_q[obj] = q
     return mix_q
 
 
 def dequantize(out, mix_q):
-    """6.6.4：去量化 → joc_mix_mtx_dq。"""
+    """去量化 → joc_mix_mtx_dq。
+
+    Sparse 的非 active 项在 ``joc_mix_mtx_q`` 中取 ``nquant/2``，
+    因此与 dense 共用同一条去量化公式即得到 0。
+    """
     mix_dq = {}
     for obj, o in enumerate(out["objs"]):
         if not o["present"]:
             continue
-        nquant = 96 if o["quant_idx"] == 0 else 192
+        nquant = JOC_NUM_QUANT[o["quant_idx"]]
         q = mix_q[obj]
         dq = (q.astype(np.float64) - nquant / 2) * 820 / (4096 * (1 + o["quant_idx"]))
         mix_dq[obj] = dq

src/metadata.py

@@ -246,20 +246,6 @@ def inspect(index, limit=None, print_frames=False):
                     "parser_error": str(exc),
                     "repair_hint": "检查 JOC header、对象数、参数带、Huffman 或扩展字段",
                 }) from exc
-        sparse = [i for i, obj in enumerate(parsed["objs"]) if obj["present"] and obj["sparse"]]
-        if sparse:
-            raise UnsupportedVariantError(
-                "joc", "sparse_joc",
-                "发现尚未验证的 Sparse JOC 帧",
-                frame=frame_number,
-                details={
-                    "sparse_objects": sparse,
-                    "downmix_config": parsed["dmx_config_idx"],
-                    "extension_config": parsed["ext_config_idx"],
-                    "objects": parsed["n_objects"],
-                    "payload": bytes_descriptor(subs[14]),
-                    "repair_hint": "需要 Sparse JOC 实际样本及对应输出建立回归后再启用",
-                })
         if parsed["n_channels"] != 5 or parsed["n_objects"] > 15:
             raise UnsupportedVariantError(
                 "joc", "unsupported_configuration",

src/native_renderer.py

@@ -208,8 +208,6 @@ class NativeJocRenderer:
         for object_index, info in enumerate(out["objs"]):
             if not info["present"]:
                 continue
-            if info["sparse"]:
-                raise ValueError("native core does not accept unvalidated Sparse JOC")
             bands = int(info["n_bands"])
             points = int(info["n_dpoints"])
             if bands > MAX_BANDS or points > MAX_DPOINTS: