Acoustic Comfort Prediction: Integrating Sound Event Detection and Noise Levels from a Wireless Acoustic Sensor...

Area of Science:

• Environmental Science
• Acoustics
• Urban Planning

Background:

• Accurate evaluation of urban soundscapes is crucial for citizens' acoustic comfort.
• Existing methods often require specialized equipment or extensive data collection.
• Citizen science projects offer valuable data for acoustic research.

Purpose of the Study:

• To develop an enhanced predictor for dwelling acoustic comfort using cost-effective data.
• To incorporate objective noise levels from public sensor networks to improve prediction accuracy.
• To evaluate the predictor using binary and Acoustic Comfort Index (ACI) rating systems.

Main Methods:

• Utilized 30-s audio clips and location information for predictor training.
• Employed gammatone cepstral coefficients and convolutional neural networks for sound event detection.
• Integrated objective noise level data from IoT-based wireless acoustic sensor networks.

Main Results:

• The proposed predictor achieved up to 85% accuracy in acoustic comfort prediction.
• Incorporating public noise levels significantly enhanced prediction accuracy for both rating systems.
• Gammatone cepstral coefficients effectively characterized sound events for analysis.

Conclusions:

• The enhanced predictor offers a cost-effective solution for assessing acoustic comfort.
• Publicly available noise data from sensor networks is valuable for improving urban acoustic assessments.
• This approach supports better urban planning and citizen well-being.