Classification of Complex Sounds.
Abstract
We have identified at least two factors which characterize listeners' abilities to discriminate complex spectra. One is the ability to integrate information across different auditory channels. Using COSS analysis we have found that good listeners show spectral weights that are close to optimal, whereas poor listeners show weights which deviate from optimal weights. A second factor is internal noise, that is, the inherent variability of listeners. We have developed a model which incorporates both two factors. An important aspect of this model is that it has no free parameters. Estimates of spectral weights and internal noise are completely determined by a trial-by-trial analysis of the data. Predictions are very consistent with thresholds obtained in a profile analysis task.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 31, 1992
- Accession Number
- ADA258405
Entities
People
- Bruce G. Berg
Organizations
- University of California, Irvine