Simple Language Normalization Wins: Cross-Lingual Speaker Verification for the TidyVoice 2026 Challenge

简单语言归一化胜出：TidyVoice 2026 挑战赛中的跨语言说话人验证

Abstract: Cross-lingual mismatch remains a key source of overall degradation in modern speaker verification. The TidyVoice2026 Challenge targets this setting with text-independent verification, comprising 3,666 training and 808 development speakers in 40 languages and 2,200 evaluation speakers in 38 unseen languages, without language labels at test time.

摘要： 跨语言不匹配仍然是现代说话人验证性能下降的主要原因之一。TidyVoice 2026 挑战赛针对这一场景进行了与文本无关的验证测试，数据集包含 40 种语言下的 3,666 名训练说话人和 808 名开发说话人，以及 38 种未见语言下的 2,200 名评估说话人，且在测试阶段不提供语言标签。

Starting from the official SimAM-ResNet34 baseline pretrained on VoxBlink2 and VoxCeleb2 and fine-tuned on TidyVoice, we revisit Nuisance Attribute Projection (NAP) as a simple language-normalization step in the embedding space.

我们以在 VoxBlink2 和 VoxCeleb2 上预训练并在 TidyVoice 上微调的官方 SimAM-ResNet34 基线模型为起点，重新审视了干扰属性投影（Nuisance Attribute Projection, NAP），将其作为嵌入空间中一种简单的语言归一化步骤。

We estimate a compact language subspace from cross-language same-speaker differences and project embeddings onto its orthogonal complement before cosine scoring with Adaptive Symmetric score normalization.

我们通过跨语言的同一说话人差异来估计一个紧凑的语言子空间，并在使用自适应对称（Adaptive Symmetric）分数归一化进行余弦评分之前，将嵌入向量投影到该子空间的补空间上。

This reduces development EER from 2.97% with cosine and 2.70% with AS-Norm to 2.18% and yields a Codabench evaluation score of 8.40, showing that simple back-end language normalization can rival more complex systems.

这一方法将开发集的等错误率（EER）从余弦评分的 2.97% 和 AS-Norm 的 2.70% 降低到了 2.18%，并在 Codabench 评估中获得了 8.40 的分数，这表明简单的后端语言归一化方法完全可以与更复杂的系统相媲美。