Environmental sounds remain identifiable even when severely spectrally filtered

Evidence

For instance, in a first experiment (Gygi et al. 2007) 145 sounds were rated in 20 semantic dimensions, and the rating suggests that 90 % of variance was associated with four factors: harshness, size, complexity, and appeal. In a second experiment the sounds were submitted to a lowand high-pass filter to understand what spectral region was selected in the sound identification of a set of 70 environmental sounds (Table 5.2) using limited spectral regions(Gygi et al. 2004). Subjects were able to identify sounds also when they were severely filtered (70 % or more of correct identification). The spectral region requested for identification ranges from 1,200 to 2,400 Hz, similar to the spectrum related to speech recognition. This type of approach investigates the minimum code for sonic identification. — Almo Farina, Soundscape Ecology, 995

Categorization

• XREF: Pays off ch13’s finding that birds decode degraded songs down to a minimal core code — the same minimum-code principle holds for human sound identification. — connects to: Acoustic codes carry messages robustly enough to survive partial signal loss — Almo Farina, Soundscape Ecology: Principles,…, 804; Detecting sounds does not guarantee discriminating them, which also needs intact critical-band processing — David Howard, Jamie Angus, coustics and Psychoacoustics, 2352; Elevation localization rests on spectral recognition of familiar sounds, not spatial cues. — Toole, Floyd E, Sound reproduction- the acoustics and…, 748

Broader Connections

• Elevation localization rests on spectral recognition of familiar sounds, not spatial cues. #corroborates (0.78) — Toole, Floyd E, Sound reproduction- the acoustics and…, 748
• Detecting sounds does not guarantee discriminating them, which also needs intact critical-band processing #expands (0.70) — David Howard, Jamie Angus, coustics and Psychoacoustics, 2352