Modeling the multi-category psychometric function markedly improves loudness estimation accuracy.

Evidence

Here, we demonstrate how a ML technique can be used with the catalog to estimate a novel listener’s MCPF from a low number (≤ 30) of CLS experimental trials. As this adaptive technique has a maximum of three presentations at each stimulus level, a histogram-estimation of the MCPF from this sparse data can be inaccurate. The 50 % intercepts of the ML MCPF category boundary curves may be used to estimate the CLS function (examples in Fig. 3 ). The average root mean squared error (RMSE) (fixed-level data used as baseline) for the median of the adaptive stimulus-level technique was 6.7 dB, and the average RMSE for the ML estimate, based on the same adaptive technique data, was 4.2 dB. The ML MCPF estimate improves the accuracy of the resulting CLS function and provides a probabilistic model of the listener’s loudness perception. — Dijk, Pim van, Physiology, Psychoacoustics and…, 2559

We have developed a “catalog” of MCPFs to be representative of both NH and HL listener CLS outcomes based on listener data. Each function in the catalog has the same form (Eq. 2 ) and is defined by a set of parameters, A PCA analysis decomposes the primary sources of variability in the parameters. Let the matrix of all listener parameters be where is the number of listeners, then the set of listener weightings, from the projections on the PCA eigenvectors is Permutations of the sampled weightings were used to reconstruct the MCPFs that comprise the catalog. — Dijk, Pim van, Physiology, Psychoacoustics and…, 2528

Categorization

• DEFINE: Clarifies the MCPF catalog method: PCA decomposes listener-parameter variability into eigenvectors whose sampled weightings reconstruct the function catalog.
• corroborated: 2 independent passages support this claim — Dijk, Pim van, Physiology, Psychoacoustics and…, 2559; Dijk, Pim van, Physiology, Psychoacoustics and…, 2528