Skip to content

Mechanical & transport

Engines, machinery, and rail/road traffic are broadband but low-frequency-weighted, and temporally steady or quasi-periodic (engine firing, bogies over rail joints, a compressor duty cycle). Before, the pieces were scattered across rhythm, spatial, states, and compare; mechanical gives the domain a first-class home. It reads the cached features (no audio pass) and its descriptors ride in the analyze summary.

What it measures

  • mech_lowfreq_fraction — share of octave energy below 250 Hz. Traffic and machinery pile energy low; a train is typically >0.9.
  • mech_rumble_db — mean level in the 31.5–125 Hz rumble band.
  • mech_periodicity_hz / mech_periodicity_strength — the peak of the 50 Hz broadband envelope's modulation spectrum in 0.3–12 Hz (engine / bogie / duty-cycle rhythm) and how strongly it stands out.
  • mechanical_index — low-frequency weight scaled by periodicity prominence, in [0, 1].

Usage

ambiscape analyze <session>      # mechanical_* keys appear in the summary
ambiscape mechanical <session>   # detail: writes mechanical.json

from ambiscape import features, mechanical
F = features.load_features(sorted((out / "features").glob("*.npz")))
mechanical.summarize_mechanical(F)

Caveats

These are acoustic-structure proxies, not detections. Indoor HVAC rumble and a passing lorry look alike here, and a steady tonal machine may also light up Mains hum & ENF. For directional confirmation use the pass-by view in Spatial analysis; for pitched machinery, the strike periodicity in Strike-level rhythm.