TIME-DOMAIN TO FREQUENCY-DOMAIN MAPPING OF TIME-SERIES NOISE LEVELS VIA FFT: IMPLICATIONS FOR LOW‐FREQUENCY NOISE ANNOYANCE ASSESSMENT

Environmental noise assessment
overwhelmingly relies on A-weighted equivalent
levels (LAeq), which aggregate time-varying
signals while suppressing low-frequency
content. This simplification masks spectral and
temporal features that are central to lowfrequency noise annoyance. Consequently, there
is a need for analytical methods that transform
high-resolution time-series measurements into
interpretable spectral signatures, particularly as
a function of distance from a noise source in
real environments. A stationary agglomeration
of industrial noise sources (dominant frequency
~300–350 Hz) was measured at radial distances
from 1 m to 50 m in a semi-open built
environment. At each location, 60-second
continuous time series were recorded (44.1 kHz
sampling, 1-second integration). Fast Fourier
Transform (FFT) analysis was performed using a
Hanning window, 8192-point segments, and
50% overlap to generate power spectral
densities. Temporal stability was assessed by
subsampling into 10 s, 30 s, and 60 s windows.
Key spectral metrics: dominant frequency,
spectral centroid, low-frequency energy
proportion (50–300 Hz vs. 300–2000 Hz), and
tonal prominence, were evaluated as functions
of distance. Results were compared with ISO
1996-2 annoyance criteria and established lowfrequency noise thresholds. The results indicate
that a 30-second window was sufficient forstable spectral extraction (coefficient of
variation <5%). At 1m, the source exhibited a
narrowband peak at 348 Hz with strong
harmonic structure. With distance, the
dominant frequency shifted to 305 Hz at 50 m,
while spectral bandwidth increased from 45 Hz
to 78 Hz. The spectral centroid decreased
consistently, indicating redistribution of energy
toward lower frequencies. Low-frequency
energy proportion increased from 22% at 1 m to
41% at 50 m, demonstrating progressive lowfrequency dominance. Tonal prominence
declined from clearly audible (12.5 dB) to
insignificant (2.1 dB). Although LAeq decreased
from 137.7 dB(A) to 43 dB(A), below WHO
guidelines beyond ~40 m, low-frequency
annoyance criteria remained exceeded up to 50
m due to spectral shifts. It was concluded that
FFT-based time-series analysis provides a robust
framework for capturing spatial spectral
evolution in environmental noise. Results show
that low-frequency energy increases with
distance despite overall A-weighted
attenuation, meaning annoyance potential
persists beyond compliance zones. Regulatory
frameworks should incorporate frequencysensitive indicators and distance-dependent
spectral metrics.