Speech intelligibility models need nonlinear SNR-importance functions for filtered sources

Evidence

If that were not enough, he also made substantial contributions to our knowledge of speech perception and developed a tool (originally called the Articulation Index, now revised to the Speech Intelligibility Index [SII]) that allows one to calculate the amount of speech information available in different frequency bands. The SII is able to robustly predict intelligibility scores for certain speech materials and acoustic environments (ANSI- 3.5, 2017) and is now used in industrial applications and to assess the impact of hearing loss on speech perception. — Jennifer Lentz, Psychoacoustics- Perception of Normal…, 90 (#complicates, via linked claim)

Categorization

• XREF: connects to: Speech intelligibility saturates beyond certain signal-to-noise ratio bounds. — Dijk, Pim van, Physiology, Psychoacoustics and…, 3538; Sharply filtered speech violates standard speech-intelligibility-index SNR assumptions; The range of SNRs contributing to speech intelligibility is wider than SII standards assume (+2 more)

Broader Connections

• Frequency importance functions weight speech audibility by band significance #corroborates (0.80)
• Speech intelligibility indexes average signal-to-noise ratio across critical bands #complicates (0.75) — Jennifer Lentz, Psychoacoustics- Perception of Normal…, 90