Low-frequency sound: Method to identify tones masked by high background noise

Montano W.A.; Gonzalez A.E.; Gushiken E. · Proceedings of Meetings on Acoustics (2025) · DOI: 10.1121/2.0002221

Resumen

Identifying low-frequency tones is challenging when measurements are masked by high ambient noise. Standard professional practice often involves non-ideal measurements where the specific sound under study (HVAC, pumps, or compressors) is masked by urban noise. While long measurement periods can capture brief moments where the steady noise is free of urban noise, a reliable record of the source can be attained. The authors present a heuristic statistical technique based on the Percentile-Based Method (PBM) for outliers’ removal to accurately estimate the sound pressure levels and spectrum of a specific sound containing tonal characteristics. The PBM uses a percentile value, like LA90, as a threshold to remove atypical sound level events (outliers) from the total sound time-series, which is crucial for non-normally distributed or skewed data sets. Unlike simply using a single statistical value like LA90 to appraise the specific sound, the PBM removes outliers and retains the one-third octave band data associated with the remaining sound levels. This allows for the restoration of the specific sound’s actual spectrum, which is critical for detecting and characterizing low-frequency tonal components. Real-world case studies demonstrate that the PBM provides a more realistic and comprehensive characterization of the specific sound compared to traditional methods.