Frequency-weighting functions for broadband speech as estimated by a correlational method

Abstract

The relative contributions of various regions of the frequency spectrum to speech recognition were assessed with a correlational method [K. A. Doherty and C. W. Turner, J. Acoust. Soc. Am. 100, 3769-3773 (1996)]. The speech materials employed were the 258-item set of the Nonsense Syllable Test. The speech was filtered into four frequency bands and a random level of noise was added to each band on each trial. A point biserial correlation was computed between the signal-to-noise ratio in each band on the trials and the listener's responses, and these correlations were then taken as estimates of the relative weights for each frequency band. When the four bands were presented separately, the correlations for each band were approximately equal; however, when the four bands were presented in combination, the correlations were quite different from one another, implying that in the broadband case listeners relied much more on some bands than on others. It is hypothesized that these differences reflect the way in which listeners combine and attend to speech information across various frequency regions. The frequency-weighting functions as determined by this method were highly similar across all subjects, suggesting that normal-hearing listeners use similar frequency-weighting strategies in recognizing speech.

Use of a compound approach to derive auditory-filter-wide frequency-importance functions for vowels and consonants

“…First, the compound technique was specifically designed to reflect much of the synergetic and redundant interactions that exist across frequency. As suggested by Turner et al (1998), these interactions may significantly affect the shape of the BIFs. For instance, the authors showed that when four broad speech bands having the same intelligibility when presented in isolation are presented together, their relative weight is not equal.…”

“…Therefore, the traditional hole approach may not be suited for estimating the weight of auditory-filter outputs. A second technique is based on the use of a correlational procedure (Doherty and Turner, 1996;Turner et al, 1998;Apoux and Bacon, 2004;Calandruccio and Doherty, 2007). To determine the weights applied to various frequency regions of speech, the information in each speech band is independently and randomly degraded by a given amount on each trial by the addition of noise.…”

The role of temporal and dynamic signal components in the perception of syllable-final stop voicing by children and adults

“…The separation between functions at the phoneme boundaries (for each level of the noncontinuous property) is taken as an indication of the weight assigned to that noncontinuous property: the greater the separation, the greater the weight that was assigned. More recently, some investigators (e.g., Turner et al, 1998) have started computing partial correlation coefficients (partial rs) for each of the acoustic properties manipulated in the experiment and the proportion of one response option given (i.e., looking at how well each acoustic property predicts responses). Both kinds of metrics were computed in this experiment.…”

Use of an adaptive-bandwidth protocol to measure importance functions for simulated cochlear implant frequency channels

“…One promising approach for measuring individual electrode contributions is the correlational method (Lutfi, 1995;Doherty and Turner, 1996;Turner et al, 1998;Calandruccio and Doherty, 2007;. In this approach, speech is filtered into several bands, each of which is corrupted by additive noise at a wide range of signal-tonoise ratios (SNRs) and presented to subjects in a speech recognition task.…”