Experimental Configuration for Pertussis Cough Recognition with 178 Recordings

178 pertussis cough recordings from three Kaggle datasets were used to evaluate an interpretable deep learning framework. The proposed method achieved AUC scores up to 0.996 and demonstrated noise robustness with performance degradation below 3%. This 5.5 KB XLS file, authored by Siheng Zhang and last updated in May 2026, contains the experimental configuration details under a CC-BY-4.0 license.

Use Cases

• Benchmarking interpretable deep learning models for audio classification based on the described adaptive time–frequency fusion and medically guided attention mechanisms.
• Studying noise robustness in clinical audio recognition models based on the reported <3% performance degradation in experiments.
• Replicating the experimental setup for pertussis cough sound recognition based on the documented 8:2 training-testing split and data leakage prevention.
• Analyzing the alignment of model attention with medically relevant acoustic features based on the described Medically-Guided Interpretable Attention Mapping module.

Strengths

• Experimental results are quantitatively reported, including AUC scores of 0.994, 0.984, and 0.996 on three datasets.
• The methodology is described in detail, including specific components like the Adaptive Time–Frequency Fusion Transformer and strict data leakage prevention.
• Noise robustness is validated with a concrete metric of performance degradation remaining below 3%.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is very small at 5.5 KB, indicating it likely contains configuration or summary data, not the raw audio recordings.

Provenance

Source

figshare, authored by Siheng Zhang.

Collection Method

Experiments conducted on three publicly available pertussis cough sound datasets from Kaggle.

Time Range

null

Freshness

Last updated 2026-05-04 17:47:16; freshness should be verified.

Geography

null