Exploratory Room-Level Acoustic Soundscape Monitoring of Cough-like Events Under Standard and Ventilation-Restricted...

Area of Science:

• Animal Science
• Agricultural Engineering
• Bioacoustics

Background:

• Precision pig farming relies on non-invasive monitoring for improved animal welfare and productivity.
• Limited data exists on calibrated, room-level respiratory sound monitoring under compromised air quality in swine housing.

Purpose of the Study:

• To explore the feasibility of room-level soundscape monitoring for assessing pig housing environments.
• To evaluate the performance of an Audio Spectrogram Transformer (AST) model for respiratory sound analysis in pigs under varying air quality.

Main Methods:

• A 28-day study monitored 52 pigs in two rooms (standard vs. ventilation-restricted) using microphones for room-level soundscape analysis.
• Airflow, sound playback calibration, and background noise analysis were conducted to minimize bias.
• An Audio Spectrogram Transformer (AST) pipeline was validated using cross-validation and temporal deployment.

Main Results:

• The AST model achieved high performance (test macro-F1 of 0.937) in classifying respiratory sounds.
• The ventilation-restricted room showed increased cough-like detections, vocalizations, and poorer air quality compared to the standard room.
• Audio-only analysis provided a median 34-minute lead time for detecting air quality deterioration, with cough-like events correlating with NH3, temperature, and CO2.

Conclusions:

• Calibrated AST-based monitoring can effectively summarize group-level acoustic changes related to degraded pig housing environments.
• The system shows practical utility as an early inspection cue for farmers regarding air quality.
• Further multi-room and multi-farm studies are needed for causal inference and broader generalization.