A Machine Hearing System for Robust Cough Detection Based on a High-Level Representation of Band-Specific Audio Features

TL;DR

This work tackles robust cough detection from audio in real-life noisy settings using a two-tier approach: band-specific short-term spectral features and a high-level representation. Five frequency bands are used to compute 12+ features per band, followed by feature selection to a compact 29-feature set, then long-term representations (AvgSD and BoAW) feed SVM classifiers. The method achieves a protocol-averaged AUC of 90.69% with SEN 92.71% and SPE 88.58% on real patient data, outperforming several state-of-the-art detectors. The approach supports smartphone-based, non-invasive cough monitoring with potential benefits for patients, clinicians, and health systems, and demonstrates strong generalization via leave-one-patient-out validation.

Abstract

Cough is a protective reflex conveying information on the state of the respiratory system. Cough assessment has been limited so far to subjective measurement tools or uncomfortable (i.e., non-wearable) cough monitors. This limits the potential of real-time cough monitoring to improve respiratory care. Objective: This paper presents a machine hearing system for audio-based robust cough segmentation that can be easily deployed in mobile scenarios. Methods: Cough detection is performed in two steps. First, a short-term spectral feature set is separately computed in five predefined frequency bands: [0, 0.5), [0.5, 1), [1, 1.5), [1.5, 2), and [2, 5.5125] kHz. Feature selection and combination are then applied to make the short-term feature set robust enough in different noisy scenarios. Second, high-level data representation is achieved by computing the mean and standard deviation of short-term descriptors in 300 ms long-term frames. Finally, cough detection is carried out using a support vector machine trained with data from different noisy scenarios. The system is evaluated using a patient signal database which emulates three real-life scenarios in terms of noise content. Results: The system achieves 92.71% sensitivity, 88.58% specificity, and 90.69% Area Under Receiver Operating Characteristic (ROC) curve (AUC), outperforming state-of-the-art methods. Conclusion: Our research outcome paves the way to create a device for cough monitoring in real-life situations. Significance: Our proposal is aligned with a more comfortable and less disruptive patient monitoring, with benefits for patients (allows self-monitoring of cough symptoms), practitioners (e.g., assessment of treatments or better clinical understanding of cough patterns), and national health systems (by reducing hospitalizations).

Figures (11)

• Figure 1: Representation of different cough events and their spectrograms where the specific phases have been detailed: (I) explosive phase, (II) intermediate phase, and (III) voiced phase. Events: (a) strong intermediate phase; (b) absent vocal phase; (c) strong vocal; (d) weak intermediate and vocal.
• Figure 2: Patient-aggregated SNR distribution for each part of the protocol
• Figure 3: Processing pipeline of the proposed cough detection system with specific references to the sections describing each block.
• Figure 4: Sample average periodogram of the recorded cough events.
• Figure 5: Pipeline of the the feature selection process.