Effect of garment obstruction/muffling on cough detection

Garment type can systematically influence the performance of wrist-worn acoustic cough detection systems because clothing alters the transmission, attenuation, and contamination of body-borne acoustic energy. Clothing disrupts this pathway in several well-defined ways.

1. Garment-Generated Acoustic Artefacts

Sleeves in motion generate their own contact-acoustic noise, including frictional rubbing, fabric snapping, and rustling.

• These garment-borne vibrations can be transmitted to the device through direct contact or through transient pressure changes.
• Such artefacts can occupy the same frequency ranges as body-borne cough vibrations, increasing the risk of both false positives (artefacts mistaken for coughs) and false negatives (true coughs masked by noise).

2. Changes in Sensor Coupling and Fit

Although the sensing mechanism is acoustic, the quality of mechanical–acoustic coupling between the device and wrist strongly determines signal fidelity.

• Different garment types (tight cuffs, elastic sleeves, bulky sweaters) can slightly reposition the device or vary the contact pressure against the skin.
• Variability in coupling leads to inconsistent acoustic transmission paths and reduced reproducibility of cough signatures across conditions.

3. Muffling / Obstruction

Garments can form a physical barrier between which obstructs the microphone, obstructing the signal.

Hyfe’s Experiments to Quantify Garment-Related Acoustic Interference

Hyfe is planning a structured set of three experiments, all focused on quantifying how clothing affects acoustic pathways and algorithmic performance. These experiments will take place during Q4 2025 and Q1 2026.

Experiment 1: Bench-Top Acoustic Transmission Study

Objective: Characterize how different fabrics and sleeve configurations modify the acoustic transmission of cough-like signals from a simulated wrist.

Design Overview:

• Use an instrumented wrist phantom or rigid fixture that receives standardised cough-like acoustic inputs (e.g., via a mini-shaker or speaker generating a canonical cough waveform).

• Apply controlled garment conditions:

  • Fabric type (cotton, wool, fleece, polyester, denim)
  • Sleeve tightness (tight, regular, loose)
  • Layer count (bare, single layer, multi-layer).

• Record waveform attenuation, spectral distortion, and noise generated by fabric movement.

Rationale: This isolates the acoustic effects of clothing without human variability, establishing baseline attenuation and noise profiles for different garment types.

Experiment 2: Controlled Human Study With Standardised Garments

Objective: Quantitatively assess how clothing impacts acoustic cough detection performance in a controlled human environment.

Design Overview:

• Participants wear the Hyfe wrist-worn acoustic cough monitor under several garment conditions:

  • Bare wrist
  • Light long sleeve (tight and loose)
  • Sweater/hoodie
  • Jacket/outerwear

• Each participant performs:

  • Standardised voluntary coughs (for precisely labelled acoustic ground truth)
  • Controlled movements (sitting, walking, arm swings) to elicit garment-noise artefacts.

• Ground truth provided by synchronised high-quality audio or expert annotation.

• Evaluate:

  • Sensitivity/specificity under each garment condition
  • Changes in acoustic feature distributions (e.g., band-limited energy, spectral centroid, waveform envelope).

Rationale: This links garment-induced acoustic interference to actual algorithmic degradation in a controlled, reproducible setting.

Experiment 3: Free-Living Field Study With Garment Logging

Objective: Determine the real-world impact of garment type on wrist-based acoustic cough monitoring under natural behaviour and environments.

Design Overview:

• Participants wear the device continuously for multiple days/weeks.

• Garment state is logged through a smartphone interface (bare wrist, light sleeve, heavy sleeve, jacket, etc.), with optional time-stamped photographs to enhance accuracy.

• A subset may wear a second, chest-mounted acoustic cough monitor or provide audio-based ground truth for periodic validation.

• Analyse:

  • Algorithmic metrics stratified by garment type
  • How daily contexts interact with clothing to affect acoustic SNR
  • Whether garment-aware models or preprocessing restore performance.

Rationale: This assesses ecological validity: it quantifies how much acoustic interference clothing actually produces in everyday use and informs mitigation strategies.