Raise ACR robustness with retrieval-first structure and music-aware inputs

Shift the prototype toward music-retrieval behavior by documenting dataset contracts, upgrading the frontend to 128-bin Mel plus band splitting, and adding retrieval evaluation plus harder confusion-oriented augmentation.

Constraint: The previous pipeline mixed train splits with the searchable catalog and hid real retrieval quality
Rejected: Keep classification-centric validation and whole-song averaged references | it masked structural accuracy failures
Confidence: medium
Scope-risk: moderate
Directive: Next iterations should target humming/confused top1 with specialized melody-aware retrieval and stronger real-data calibration
Tested: synthetic_v2 generation; 3-epoch CPU training; index build; evaluate.py top1=0.65 top5=0.95 on test split
Not-tested: external open-dataset ingestion; foundation-model baselines; production latency

@@ -5,9 +5,11 @@ model:
   se_channels: 128
   se_channels: 128
   res2net_scale: 8
   res2net_scale: 8
   num_blocks: 3
   num_blocks: 3
-  n_mels: 80
+  n_mels: 128
   aam_m: 0.3
   aam_m: 0.3
   aam_s: 30.0
   aam_s: 30.0
+  use_band_split: true
+  band_split_channels: 128
 
 
 data:
 data:
   sample_rate: 16000
   sample_rate: 16000
@@ -39,3 +41,8 @@ engine:
     chroma_weight: 0.3
     chroma_weight: 0.3
     ecapa_weight: 0.7
     ecapa_weight: 0.7
     reject_threshold: 0.4
     reject_threshold: 0.4
+
+augmentation:
+  pro_wgan_balance: true
+  minority_noise_scale: 0.35
+  minority_pitch_shift: 8

@@ -229,9 +229,12 @@ class SongPairDataset(Dataset):
             y = self._load_clip(sample)
             y = self._load_clip(sample)
             if self.augment:
             if self.augment:
                 from src.utils.augment import AugmentPipeline
                 from src.utils.augment import AugmentPipeline
-                y = AugmentPipeline(self.sr)(y)
+                y = AugmentPipeline(self.sr, aggressive=sample.get("type") in {"confused", "humming_like"})(y)
             wavs.append(self._to_mel(y))
             wavs.append(self._to_mel(y))
 
 
+        max_t = max(w.shape[1] for w in wavs)
+        wavs = [torch.nn.functional.pad(w, (0, max_t - w.shape[1])) if w.shape[1] < max_t else w for w in wavs]
+
         label = self.song_to_idx[song_id]
         label = self.song_to_idx[song_id]
         return {
         return {
             "mel": torch.stack(wavs, dim=0),
             "mel": torch.stack(wavs, dim=0),

@@ -28,14 +28,20 @@ class ECAPAEmbedder:
         state = torch.load(model_path, map_location="cpu", weights_only=True)
         state = torch.load(model_path, map_location="cpu", weights_only=True)
         cfg = state.get("config", {})
         cfg = state.get("config", {})
         model_cfg = cfg.get("model", {})
         model_cfg = cfg.get("model", {})
+        data_cfg = cfg.get("data", {})
+        self.n_mels = model_cfg.get("n_mels", n_mels)
+        self.n_fft = data_cfg.get("n_fft", n_fft)
+        self.hop_length = data_cfg.get("hop_length", hop_length)
         self.model = ECAPA_ACR(
         self.model = ECAPA_ACR(
-            n_mels=model_cfg.get("n_mels", n_mels),
+            n_mels=self.n_mels,
             embed_dim=model_cfg.get("embed_dim", 192),
             embed_dim=model_cfg.get("embed_dim", 192),
             channels=model_cfg.get("channels", 512),
             channels=model_cfg.get("channels", 512),
             se_channels=model_cfg.get("se_channels", 128),
             se_channels=model_cfg.get("se_channels", 128),
             res2net_scale=model_cfg.get("res2net_scale", 8),
             res2net_scale=model_cfg.get("res2net_scale", 8),
             num_blocks=model_cfg.get("num_blocks", 3),
             num_blocks=model_cfg.get("num_blocks", 3),
             num_classes=None,
             num_classes=None,
+            use_band_split=model_cfg.get("use_band_split", True),
+            band_split_channels=model_cfg.get("band_split_channels", 128),
         )
         )
         missing = self.model.load_state_dict(state["model_state_dict"], strict=False)
         missing = self.model.load_state_dict(state["model_state_dict"], strict=False)
         if missing.unexpected_keys:
         if missing.unexpected_keys:

 import torch
 import torch
 import torch.nn as nn
 import torch.nn as nn
 import torch.nn.functional as F
 import torch.nn.functional as F
-from typing import Optional, Tuple
+from typing import Optional, Tuple, List
 
 
 
 
 class SEModule(nn.Module):
 class SEModule(nn.Module):
@@ -19,13 +19,43 @@ class SEModule(nn.Module):
         return x * self.se(x)
         return x * self.se(x)
 
 
 
 
+class BandSplitBlock(nn.Module):
+    def __init__(self, n_mels: int, split_points: Optional[List[int]] = None, out_channels: int = 128):
+        super().__init__()
+        self.split_points = split_points or [16, 32, 64, 96, n_mels]
+        starts = [0] + self.split_points[:-1]
+        widths = [end - start for start, end in zip(starts, self.split_points)]
+        self.band_projs = nn.ModuleList(
+            [
+                nn.Sequential(
+                    nn.Conv1d(width, out_channels, kernel_size=1),
+                    nn.ReLU(),
+                    nn.BatchNorm1d(out_channels),
+                )
+                for width in widths
+            ]
+        )
+        self.fuse = nn.Sequential(
+            nn.Conv1d(out_channels * len(widths), out_channels * len(widths), kernel_size=1),
+            nn.ReLU(),
+            nn.BatchNorm1d(out_channels * len(widths)),
+        )
+
+    def forward(self, x):
+        starts = [0] + self.split_points[:-1]
+        bands = []
+        for proj, start, end in zip(self.band_projs, starts, self.split_points):
+            bands.append(proj(x[:, start:end, :]))
+        return self.fuse(torch.cat(bands, dim=1))
+
+
 class Res2Block(nn.Module):
 class Res2Block(nn.Module):
     def __init__(self, channels, kernel_size=3, dilation=1, scale=8, se_channels=128):
     def __init__(self, channels, kernel_size=3, dilation=1, scale=8, se_channels=128):
         super().__init__()
         super().__init__()
         self.width = channels // scale
         self.width = channels // scale
         self.num_split = scale
         self.num_split = scale
         self.convs = nn.ModuleList()
         self.convs = nn.ModuleList()
-        for i in range(self.num_split):
+        for _ in range(self.num_split):
             self.convs.append(
             self.convs.append(
                 nn.Sequential(
                 nn.Sequential(
                     nn.Conv1d(
                     nn.Conv1d(
@@ -54,7 +84,7 @@ class Res2Block(nn.Module):
             if i == 0:
             if i == 0:
                 out.append(conv(part))
                 out.append(conv(part))
             else:
             else:
-                out.append(conv(out[-1] if len(out) else part + part))
+                out.append(conv(part + out[-1]))
         x = torch.cat(out, dim=1)
         x = torch.cat(out, dim=1)
         x = self.conv1x1(x)
         x = self.conv1x1(x)
         x = self.se(x)
         x = self.se(x)
@@ -96,7 +126,7 @@ class AAMSoftmax(nn.Module):
 class ECAPA_ACR(nn.Module):
 class ECAPA_ACR(nn.Module):
     def __init__(
     def __init__(
         self,
         self,
-        n_mels: int = 80,
+        n_mels: int = 128,
         embed_dim: int = 192,
         embed_dim: int = 192,
         channels: int = 512,
         channels: int = 512,
         se_channels: int = 128,
         se_channels: int = 128,
@@ -105,20 +135,23 @@ class ECAPA_ACR(nn.Module):
         num_classes: Optional[int] = None,
         num_classes: Optional[int] = None,
         aam_m: float = 0.3,
         aam_m: float = 0.3,
         aam_s: float = 30.0,
         aam_s: float = 30.0,
+        use_band_split: bool = True,
+        band_split_channels: int = 128,
     ):
     ):
         super().__init__()
         super().__init__()
         self.embed_dim = embed_dim
         self.embed_dim = embed_dim
+        front_channels = band_split_channels * 5 if use_band_split else channels
+        self.band_split = BandSplitBlock(n_mels=n_mels, out_channels=band_split_channels) if use_band_split else None
 
 
         self.conv1 = nn.Sequential(
         self.conv1 = nn.Sequential(
-            nn.Conv1d(n_mels, channels, kernel_size=5, stride=1, padding=2),
+            nn.Conv1d(front_channels, channels, kernel_size=5, stride=1, padding=2),
             nn.ReLU(),
             nn.ReLU(),
             nn.BatchNorm1d(channels),
             nn.BatchNorm1d(channels),
         )
         )
 
 
-        dilations = [1, 2, 3] if num_blocks == 3 else [d * 1 for d in range(1, num_blocks + 1)]
-        self.blocks = nn.ModuleList()
-        for d in dilations[:num_blocks]:
-            self.blocks.append(
+        dilations = [1, 2, 3] if num_blocks == 3 else [d for d in range(1, num_blocks + 1)]
+        self.blocks = nn.ModuleList(
+            [
                 Res2Block(
                 Res2Block(
                     channels=channels,
                     channels=channels,
                     kernel_size=3,
                     kernel_size=3,
@@ -126,7 +159,9 @@ class ECAPA_ACR(nn.Module):
                     scale=res2net_scale,
                     scale=res2net_scale,
                     se_channels=se_channels,
                     se_channels=se_channels,
                 )
                 )
-            )
+                for d in dilations[:num_blocks]
+            ]
+        )
 
 
         in_channels = channels * num_blocks
         in_channels = channels * num_blocks
         self.mfa = nn.Sequential(
         self.mfa = nn.Sequential(
@@ -134,34 +169,25 @@ class ECAPA_ACR(nn.Module):
             nn.ReLU(),
             nn.ReLU(),
             nn.BatchNorm1d(channels * 3),
             nn.BatchNorm1d(channels * 3),
         )
         )
-
         self.pooling = StatisticsPooling()
         self.pooling = StatisticsPooling()
         self.fc = nn.Linear(channels * 3 * 2, embed_dim)
         self.fc = nn.Linear(channels * 3 * 2, embed_dim)
         self.bn = nn.BatchNorm1d(embed_dim, affine=False)
         self.bn = nn.BatchNorm1d(embed_dim, affine=False)
+        self.aam = AAMSoftmax(embed_dim, num_classes, m=aam_m, s=aam_s) if num_classes is not None else None
 
 
-        if num_classes is not None:
-            self.aam = AAMSoftmax(embed_dim, num_classes, m=aam_m, s=aam_s)
-        else:
-            self.aam = None
-
-    def forward(
-        self, mel: torch.Tensor, labels: Optional[torch.Tensor] = None
-    ) -> Tuple[torch.Tensor, Optional[torch.Tensor]]:
-        x = self.conv1(mel)
+    def forward(self, mel: torch.Tensor, labels: Optional[torch.Tensor] = None) -> Tuple[torch.Tensor, Optional[torch.Tensor]]:
+        x = self.band_split(mel) if self.band_split is not None else mel
+        x = self.conv1(x)
         block_outputs = []
         block_outputs = []
         for block in self.blocks:
         for block in self.blocks:
             x = block(x)
             x = block(x)
             block_outputs.append(x)
             block_outputs.append(x)
-
         x = torch.cat(block_outputs, dim=1)
         x = torch.cat(block_outputs, dim=1)
         x = self.mfa(x)
         x = self.mfa(x)
         x = self.pooling(x)
         x = self.pooling(x)
         x = self.fc(x)
         x = self.fc(x)
         x = self.bn(x)
         x = self.bn(x)
         embedding = F.normalize(x, p=2, dim=1)
         embedding = F.normalize(x, p=2, dim=1)
-
         if labels is not None and self.aam is not None:
         if labels is not None and self.aam is not None:
             logits = self.aam(embedding, labels)
             logits = self.aam(embedding, labels)
             return embedding, logits
             return embedding, logits
-
         return embedding, None
         return embedding, None

@@ -4,12 +4,13 @@ from typing import Optional, Tuple
 
 
 
 
 class AugmentPipeline:
 class AugmentPipeline:
-    def __init__(self, sr: int = 16000):
+    def __init__(self, sr: int = 16000, aggressive: bool = False):
         self.sr = sr
         self.sr = sr
         self.noise_snr_range = (5, 30)
         self.noise_snr_range = (5, 30)
         self.pitch_shift_range = (-6, 6)
         self.pitch_shift_range = (-6, 6)
         self.time_stretch_range = (0.85, 1.15)
         self.time_stretch_range = (0.85, 1.15)
         self.mp3_bitrate_range = (32, 128)
         self.mp3_bitrate_range = (32, 128)
+        self.aggressive = aggressive
 
 
     def add_noise(self, y: np.ndarray, snr_db: Optional[float] = None) -> np.ndarray:
     def add_noise(self, y: np.ndarray, snr_db: Optional[float] = None) -> np.ndarray:
         if snr_db is None:
         if snr_db is None:
@@ -57,14 +58,18 @@ class AugmentPipeline:
         return mel
         return mel
 
 
     def __call__(self, y: np.ndarray) -> np.ndarray:
     def __call__(self, y: np.ndarray) -> np.ndarray:
-        if random.random() < 0.5:
-            y = self.add_noise(y)
-        if random.random() < 0.3:
-            y = self.time_stretch(y)
-        if random.random() < 0.3:
-            y = self.pitch_shift(y)
-        if random.random() < 0.2:
-            y = self.add_reverb(y)
+        noise_p = 0.75 if self.aggressive else 0.5
+        stretch_p = 0.55 if self.aggressive else 0.3
+        pitch_p = 0.55 if self.aggressive else 0.3
+        reverb_p = 0.35 if self.aggressive else 0.2
+        if random.random() < noise_p:
+            y = self.add_noise(y, snr_db=random.uniform(0, 18) if self.aggressive else None)
+        if random.random() < stretch_p:
+            y = self.time_stretch(y, rate=random.uniform(0.8, 1.2) if self.aggressive else None)
+        if random.random() < pitch_p:
+            y = self.pitch_shift(y, semitones=random.uniform(-8, 8) if self.aggressive else None)
+        if random.random() < reverb_p:
+            y = self.add_reverb(y, decay=random.uniform(0.2, 0.6))
         return y
         return y
 
 
 
 

@@ -187,6 +187,8 @@ def main():
         num_classes=num_classes,
         num_classes=num_classes,
         aam_m=cfg["model"]["aam_m"],
         aam_m=cfg["model"]["aam_m"],
         aam_s=cfg["model"]["aam_s"],
         aam_s=cfg["model"]["aam_s"],
+        use_band_split=cfg["model"].get("use_band_split", True),
+        band_split_channels=cfg["model"].get("band_split_channels", 128),
     ).to(device)
     ).to(device)
 
 
     criterion = CombinedLoss(
     criterion = CombinedLoss(

@@ -21,3 +21,35 @@
 - 已完成 1 epoch CPU 训练并生成 `best_model.pt`
 - 已完成 1 epoch CPU 训练并生成 `best_model.pt`
 - 已完成指纹索引与 embedding 索引构建
 - 已完成指纹索引与 embedding 索引构建
 - 已完成识别命令并输出 JSON 候选结果
 - 已完成识别命令并输出 JSON 候选结果
+
+## 2026-06-02
+
+### Stage: 准确率优化 v2（128 Mel / band-split / retrieval 评测 / dataset 规范 / SOTA 调研）
+
+完成项：
+- 补充 dataset / 输入输出规范：`docs/dataset-spec.md`
+- 补充开源数据集接入计划：`docs/open-dataset-plan.md`
+- 补充 2026 SOTA 研究说明：`docs/sota-research-2026.md`
+- 输入特征从低维说话人风格配置改为 `128 Mel`
+- 新增频带分割模块 `BandSplitBlock`
+- 引入 pro-WGAN 风格工程近似平衡策略（针对困难样本的更强增广）
+- 合成数据新增 `confused` / `humming_like` 样本类型
+- 引入 `catalog.json` 作为可搜索 reference 清单
+- 索引从整曲单向量改为 window-level embedding index
+- 新增 `evaluate.py` 做 retrieval 评测
+- 训练逻辑改为更 retrieval-oriented 的 song-pair 训练输入
+
+验证结果：
+- synthetic_v2 端到端重新跑通
+- build-index 成功
+- evaluate 成功
+- test split 指标：top1=0.65, top5=0.95
+- 分类型指标：
+  - clean top1=1.00
+  - augmented top1=0.75
+  - humming_like top1=0.25
+  - confused top1=0.25
+
+结论：
+- 结构性错误（catalog/index/fusion/评测缺失）已明显改善
+- 当前主要剩余短板是 humming_like / confused 的鲁棒识别

+# ACR Dataset / 输入输出规范
+
+> 更新：2026-06-02
+
+## 1. 目标
+
+定义本项目数据集规范、输入输出处理流程、catalog/query 划分方式，以及训练/评测所需的 manifest 结构。
+
+## 2. 数据层对象
+
+### 2.1 Reference / Catalog
+可检索曲库中的标准参考音频。
+
+字段：
+
+```json
+{
+  "song_id": "song_0001",
+  "audio_path": "songs/song_0001.wav",
+  "duration": 20.0,
+  "base_freq": 261.63,
+  "type": "reference"
+}
+```
+
+用途：
+- 建立 chromaprint 索引
+- 建立 embedding window 索引
+- 作为检索目标集合
+
+### 2.2 Query Segment
+待识别片段。
+
+字段：
+
+```json
+{
+  "song_id": "song_0001",
+  "audio_path": "segments/song_0001_seg_02_confused.wav",
+  "duration": 5.0,
+  "type": "confused",
+  "offset": 8.3,
+  "segment_type": "mid"
+}
+```
+
+用途：
+- 训练片段对
+- top-k 检索评测
+- 鲁棒性测试
+
+## 3. Manifest 文件
+
+| 文件 | 用途 |
+|---|---|
+| `train.json` | 训练查询片段 + 训练 reference |
+| `val.json` | 验证查询片段 + 验证 reference |
+| `test.json` | 测试查询片段 + 测试 reference |
+| `catalog.json` | 可搜索 reference 总表 |
+
+注意：
+- `catalog.json` 是**检索索引输入**
+- `train/val/test.json` 是**实验 split**
+- 不再把 “模型训练 split” 和 “可搜索曲库” 混为一谈
+
+## 4. 输入特征规范
+
+### 4.1 输入音频
+- 默认采样率：`16 kHz`
+- 通道：`mono`
+- 训练/query 窗长：`5s`
+- 滑窗步长：`2.5s`
+
+### 4.2 声学特征
+当前改为：
+- **128维 Mel 频谱**
+
+不再采用传统说话人任务常见的 40 维 MFCC 作为主输入，因为：
+- 音乐任务更依赖频带结构与谐波信息
+- Mel 频谱对音乐 timbre / harmony / texture 表达更自然
+- 便于 band-split 模块对频带进行分块建模
+
+## 5. 输出规范
+
+### 5.1 训练输出
+模型输出：
+- `embedding: [B, D]`
+- `logits: [B, num_classes]`（辅助分类头）
+
+主要目标：
+- retrieval embedding 学得稳定
+- 同 song 片段彼此接近
+- 不同 song 分离
+
+### 5.2 推理输出
+识别输出：
+
+```json
+{
+  "candidates": [
+    {
+      "song_id": "song_0001",
+      "confidence": 0.93,
+      "chromaprint_score": 0.88,
+      "ecapa_score": 0.96,
+      "accepted": true,
+      "metadata": {}
+    }
+  ],
+  "processing_time_ms": 120.4,
+  "num_candidates": 5
+}
+```
+
+## 6. Query 类型定义
+
+| type | 含义 |
+|---|---|
+| `clean` | 原始干净片段 |
+| `augmented` | 常规增强片段 |
+| `confused` | 强混淆/干扰片段 |
+| `humming_like` | 哼唱风格近似片段 |
+| `reference` | 标准参考整曲 |
+
+## 7. pro-WGAN 平衡策略（工程近似版）
+
+当前仓库先实现的是**pro-WGAN 风格的数据平衡近似策略**，不是完整生成式 GAN 训练：
+
+- 对难样本类型（`confused`, `humming_like`）增加更强增广概率
+- 通过 harder augmentation 近似 minority/hard-case oversampling
+- 保持 manifest 结构兼容，后续可替换成真正的生成式平衡器
+
+后续若接入完整 GAN 平衡器，可把它作为：
+- 离线样本扩增器
+- 困难类别样本生成器
+- catalog/query domain adaptation 工具
+
+## 8. 频带分割模块
+
+输入层新增 `BandSplitBlock`：
+- 将 128 Mel bins 分割为多个子频带
+- 每个子带做独立投影
+- 再拼接进入主干网络
+
+目的：
+- 强化低频节奏 / 中频和声 / 高频音色的分带建模
+- 更符合音乐频谱结构
+- 为后续更复杂 band-aware retrieval 打基础

+# ACR / Music Retrieval SOTA Research (截至 2026-06-02)
+
+## 结论摘要
+
+到 2025-2026，这个方向相比传统“从零训练一个小型 ECAPA embedding”已经明显前进了。
+
+当前更强的方向主要有三类：
+
+1. **Neural Audio Fingerprinting 的鲁棒训练增强**
+2. **Music Foundation Model 作为 backbone / teacher**
+3. **Band-split / band-aware 结构用于音乐频谱建模**
+
+## 1. Neural AFP 的更强实践
+
+### Enhancing Neural Audio Fingerprint Robustness to Audio Degradation for Music Identification (2025)
+- arXiv: https://arxiv.org/abs/2506.22661
+
+关键信息：
+- 指出很多 neural AFP 工作对真实退化模拟不够真实
+- 系统比较 metric learning 方法
+- 发现自监督 triplet loss 变体在该任务中更优
+- 强调多个 positive samples 对不同 loss 的影响不同
+
+对本项目的启发：
+- 不应只依赖当前简单 SupCon + CE
+- 应增加更真实的退化增强
+- 应明确做 retrieval 指标选择，而非只看分类头
+
+## 2. Music Foundation Model Backbones
+
+### Robust Neural Audio Fingerprinting using Music Foundation Models (2025)
+- arXiv: https://arxiv.org/abs/2511.05399
+
+关键信息：
+- 使用预训练 music foundation model（例如 MuQ、MERT）作为 neural fingerprinting backbone
+- 在 distorted / compressed / manipulated 音频条件下优于从零训练模型
+- 还能更好做 segment-level localization
+
+### MERT (2023)
+- arXiv: https://arxiv.org/abs/2306.00107
+
+关键信息：
+- 大规模自监督 music understanding 模型
+- 在多个 music understanding 任务上达到强表现
+
+### MuQ (2025)
+- arXiv: https://arxiv.org/abs/2501.01108
+
+关键信息：
+- 面向音乐的自监督表征学习模型
+- 使用 Mel-RVQ 目标
+- 在多种下游任务上优于更早工作
+
+对本项目的启发：
+- 2026 继续只用小模型从零训，不太可能是最佳路线
+- 更合理路线：
+  - 当前仓库保留轻量自训 baseline
+  - 下一阶段增加 MERT / MuQ frozen encoder 或 adapter fine-tune 版本
+
+## 3. Band-split / band-aware 结构
+
+### Music Source Separation with Band-split RNN (2022)
+- arXiv: https://arxiv.org/abs/2209.15174
+
+关键信息：
+- 显式把频谱切成多个频带再建模
+- 对音乐任务优于直接照搬通用音频结构
+
+虽然该文主要做 source separation，不是 ACR，但它对“音乐频带先验”很有启发。
+
+对本项目的启发：
+- 输入层加入 band-split 是合理工程方向
+- 未来可继续发展成：
+  - band-aware attention
+  - multi-band retrieval heads
+  - harmonic/rhythm 双塔结构
+
+## 4. 数据平衡与生成增强
+
+### BAGAN: Data Augmentation with Balancing GAN (2018)
+- arXiv: https://arxiv.org/abs/1803.09655
+
+严格说你提到的 `pro-WGAN` 我这次没有找到一个明确、权威、在该任务里广泛标准化的同名主文献；当前更接近、且有明确权威来源的是 **BAGAN / balancing GAN** 这一类面向不平衡数据增强的方法。
+
+因此本次实现里我采用的是：
+- **pro-WGAN 风格的工程近似平衡策略**
+- 不是声称已经复现某篇明确的 `pro-WGAN` SOTA 论文
+
+如果你之后指定了准确论文或仓库，我可以按那一版精确对齐实现。
+
+## 5. 2026 年是否已经有更好的方案？
+
+有，结论是：**有明显更好的路线**。
+
+最值得参考的是：
+1. 用 **music foundation model** 做 backbone
+2. 用 **更真实退化模拟 + retrieval-first metric learning**
+3. 用 **segment-level / window-level indexing**，而不是整曲平均 embedding
+4. 对哼唱任务增加 **melody/pitch contour 专门支路**
+
+## 6. 对本项目的建议排序
+
+### 当前阶段（已开始）
+- 128 Mel 替换低维说话人风格输入
+- band-split 输入层
+- 更强混淆增强
+- retrieval-first 评测
+
+### 下一阶段
+- MERT / MuQ frozen feature baseline
+- triplet / multi-positive metric learning 对比 SupCon
+- window-level index aggregation
+- FMA / Jamendo 小规模真实数据验证
+
+### 更后阶段
+- humming 专门 melody tower
+- foundation model + lightweight fingerprint head
+- ANN + reranker 两阶段工业化检索
+
+## Sources
+- Araz et al., 2025, Enhancing Neural Audio Fingerprint Robustness to Audio Degradation for Music Identification: https://arxiv.org/abs/2506.22661
+- Singh et al., 2025, Robust Neural Audio Fingerprinting using Music Foundation Models: https://arxiv.org/abs/2511.05399
+- Li et al., 2023, MERT: Acoustic Music Understanding Model with Large-Scale Self-supervised Training: https://arxiv.org/abs/2306.00107
+- Zhu et al., 2025, MuQ: Self-Supervised Music Representation Learning with Mel Residual Vector Quantization: https://arxiv.org/abs/2501.01108
+- Luo & Yu, 2022, Music Source Separation with Band-split RNN: https://arxiv.org/abs/2209.15174
+- Mariani et al., 2018, BAGAN: Data Augmentation with Balancing GAN: https://arxiv.org/abs/1803.09655