A modified masking model predicts consonant recognition rates better than the original.

Evidence

For P (panel a), an underestimation of the recognition for many consonants was observed, indicated by the mismatch of the on-diagonal circles. For Pmod (panel b), a good consonant-specific recognition rate match was found. Both models hit most of the confusions, reflected in the proportion of gray off-diagonal circles matched with a black circle. However, there were also many “false alarms”, particularly in the P-based predictions (panel a). — Dijk, Pim van, Physiology, Psychoacoustics and…, 5616

Categorization

Category: Causal Relationships

Related

• Standard prediction functions underestimate heard consonant recognition while modified ones do not. (similarity 0.87)
• Both models overpredict consonant confusions that listeners do not actually make. (similarity 0.86)
• Modeling per-token variability captures consonant recognition rates that a fixed-rate model misses. (similarity 0.84)
• Modified model captures token-specific speech reception thresholds far better than baseline (similarity 0.81)
• Masking reduces to a comparison of excitation-level patterns over frequency and time (similarity 0.81)
• Perceptual distance between consonant tokens is highly predictable from internal-model responses. (similarity 0.80)
• Energy-based and modulation-based models predict consonant recognition but not confusions. (similarity 0.80)
• Both spectral sparsity and lack of masker fluctuations contribute to release from masking. (similarity 0.79)
• Masking release from modulated noise also improves speech perception. (similarity 0.79)
• Auditory modulation frequency selectivity allows inherent masker fluctuations to be separated from beating-induced ones. (similarity 0.79)
• Masking depends on stimulus fluctuations in ways steady-state models cannot predict. (similarity 0.79)
• Modeled acoustic distances predict perceived consonant distances with near-perfect correlation (similarity 0.79)
• Fluctuating maskers produce masking release that Fletcher’s power spectrum model cannot explain. (similarity 0.79)
• Masking is quantified by the shift in a listener’s hearing-threshold curve (similarity 0.79)
• Temporal masking alone masks brief consonants even without room reverberation. (similarity 0.78)