Short-term acoustic forecasting via artificial neural networks for neonatal intensive care units

J Acoust Soc Am. 2012 Nov;132(5):3234-9. doi: 10.1121/1.4754556.

Abstract

Noise levels in hospitals, especially neonatal intensive care units (NICUs), have become of great concern for hospital designers. This paper details an artificial neural network (ANN) approach to forecasting the sound loads in NICUs. The ANN is used to learn the relationship between past, present, and future noise levels. By training the ANN with data specific to the location and device used to measure the sound, the ANN is able to produce reasonable predictions of noise levels in the NICU. Best case results show average absolute errors of 5.06 ± 4.04% when used to predict the noise levels one hour ahead, which correspond to 2.53 dBA ± 2.02 dBA. The ANN has the tendency to overpredict during periods of stability and underpredict during large transients. This forecasting algorithm could be of use in any application where prediction and prevention of harmful noise levels are of the utmost concern.

MeSH terms

  • Acoustics* / instrumentation
  • Algorithms
  • Environmental Exposure* / adverse effects
  • Environmental Exposure* / prevention & control
  • Environmental Monitoring / methods
  • Forecasting
  • Hospital Design and Construction / methods*
  • Humans
  • Infant, Newborn
  • Intensive Care Units, Neonatal*
  • Neural Networks, Computer*
  • Noise* / adverse effects
  • Noise* / prevention & control
  • Pressure
  • Sound Spectrography
  • Time Factors
  • Transducers, Pressure