Feature Scaling Beats Architecture Choice in Offline Respiratory Screening for Low-Resource Clinics

Deep models are the usual choice for respiratory sound classification, but in clinics without GPUs or connectivity that choice carries a cost. We compare classical and deep classifiers on a five-class respiratory-sound corpus of 1,211 clips (asthma, COPD, pneumonia, Bronchial, healthy) from 201 patient groups, with a 3.86:1 class imbalance, using pre-registered patient-grouped cross-validation (25 runs per configuration).

深度模型通常是呼吸音分类的首选，但在没有 GPU 或网络连接的诊所中，这种选择会带来额外的成本。我们对比了经典分类器和深度分类器在包含 1,211 个片段（哮喘、慢阻肺、肺炎、支气管炎、健康）的五类呼吸音语料库上的表现，这些片段来自 201 个患者组，类别不平衡比为 3.86:1，并使用了预注册的患者分组交叉验证（每种配置运行 25 次）。

An SVM over 30 summary features (MFCC, ZCR and RMS statistics) reaches macro recall 0.78 and 0.71 recall on the rarest class. A 1.13M-parameter convolutional recurrent network over log-Mel spectrograms reaches 0.75 to 0.77, and no difference between model families is resolved. The largest single effect we measure is a preprocessing effect rather than an architectural one.

一个基于 30 个汇总特征（MFCC、ZCR 和 RMS 统计量）的 SVM 模型达到了 0.78 的宏观召回率，在最稀有类别上的召回率为 0.71。一个基于对数梅尔频谱图、拥有 113 万参数的卷积循环网络达到了 0.75 到 0.77 的召回率，且模型系列之间未见显著差异。我们测得的最大单一影响因素是预处理，而非架构选择。

Standardizing the SVM’s features, a routine step for RBF kernels, moves its macro recall from 0.51 to 0.78, a 27-point swing wider than any gap between model families here. Augmentation did not help: SMOTE, waveform augmentation and SpecAugment have no measurable effect. Latency does not separate the models (51 to 74 ms per clip on a CPU-only laptop), but the classical models avoid a 449 MB PyTorch runtime.

对 SVM 特征进行标准化（这是 RBF 核函数的常规步骤）使其宏观召回率从 0.51 提升至 0.78，这一 27 个百分点的波动幅度超过了此处模型系列间的任何差距。数据增强并无帮助：SMOTE、波形增强和 SpecAugment 均无明显效果。延迟并不能区分这些模型（在仅使用 CPU 的笔记本电脑上，每个片段的延迟为 51 到 74 毫秒），但经典模型避免了 449 MB 的 PyTorch 运行时开销。

Clip-level splits, which let a patient’s recordings fall on both sides, inflate macro recall by 7.5 to 9.8 points. Recording source is strongly tied to label (Cramer’s V = 0.65 between native sampling rate and class), and in an exploratory analysis clips recorded at a class’s minority rate are mostly assigned to classes typical of that rate; we state this confound as the paper’s main limitation.

片段级划分（即允许同一患者的录音出现在训练集和测试集两侧）会将宏观召回率虚高 7.5 到 9.8 个百分点。录音来源与标签高度相关（原始采样率与类别之间的 Cramer’s V 系数为 0.65），在探索性分析中，以某类别的少数采样率录制的片段大多被分配给了该采样率的典型类别；我们将此混淆因素列为本文的主要局限性。