Effect Of Frequency Uncertainty Research Articles

A level of stimulus representation model for auditory detection and attention.

A model is offered here to address an asymmetry of cueing in signal detection [Hafter et al. (1992)] where the effect of frequency uncertainty on the detection of a randomly chosen tone was ameliorated by cueing with a sequence of its harmonics, but detection of a randomly chosen sequence of harmonics was not improved by cueing with their fundamental. The model proposes that signal detection can be based on various levels of neural representation that, for the case at hand, refer to levels organized either by frequency or by complex pitch. Experiments offered to test the model used three-tone complexes for both cues and signals. These stimuli consisted of either three randomly chosen frequencies or three randomly chosen harmonics (from the set 2 f1 to 7 f1) of a randomly chosen fundamental. Support for the idea of cueing and detection at different levels of representation was found in higher performance with uncued detection of harmonic complexes relative to that found with complexes of unrelated tones and by successful cueing of each type of information with cues created to remove uncertainty about the relevant information. A final comparison suggests independence of performance (presumably of the limiting noise) at each of the putative levels of representation.

Listening bandwidths and frequency uncertainty in pure-tone signal detection

The effect of frequency uncertainty on the detection of tonal signals in noise was studied using a modified probe-signal method. Widths of the listening bands used during detection were measured directly, allowing for an analysis that separates the effects of having to monitor multiple independent bands from those due to limited frequency resolution. Uncertainty was varied by beginning each trial with a cue consisting of one, two, or four randomly chosen, simultaneously presented tones. An expected signal, whose frequency matched one of the components in a cue, was presented on a majority of trials. However, on remaining trials, the signal was a probe, which meant that its frequency differed from one of the components in the cue by a constant ratio. Performance as measured in percent correct declined for probes at increasingly distant ratios from the expected values. The results were converted to dB using individual psychometric functions for expected signals and listening bands were fitted using the rounded exponential filter of Patterson et al. [J. Acoust. Soc. Am. 72, 1788-1803 (1982)]. The obtained bandwidths are comparable to those reported using notched-noise maskers, but there is a small but consistent increase in bandwidth with increased numbers of components in the cues. The primary results is that the effects due to uncertainty are well described by a 1-of-M orthogonal band model, which takes into consideration limitations of the detector, including the widths of the listening bands.

Performance of a Square Law Pseudonoise Ranging Time-of-Arrival Estimator

The performance of a square law time-of-arrival (TOA) estimator that has been proposed for use in ASTRO-DABS, part of a possible satellite-based fourth generation air traffic control system is considered. The transmitted message consists of a pulse amplitude modulated (PAM) ranging sequence that, due to transmitter characteristics, is corrupted by an unknown frequency offset. The optimum TOA estimator, for the case of no frequency uncertainty, is first presented, together with a lower bound on the variance of the estimate generated. This is followed by the consideration of a suboptimum TOA estimator for which a high signal-to-noise ratio (SNR) performance analysis is carried out; here, the effects of frequency uncertainty are included. Next, the zero-crossing properties of the derivative of the (suboptimum) estimation statistic are presented and the results used to derive an upper bound to the TOA estimate variance that is valid for all SNR values. This latter result is significant because it displays the system threshold effect and complements performance lower bounds that may be derived via other methods. In addition, the method presented here may be applied to other optimum and suboptimum systems where a discrete set of parameters is to be estimated.

Detection of a Pulsed Sinusoid of Uncertain Frequency

Two conditions were used in the experiments. In one condition the observer knew that a particular signal frequency would be used on all the trials of a sequence. In the other condition the observer knew only that the signal would always be some frequency in a given range. In this second condition the particular signal frequency used on any trial of the sequence was determined by a random process. The observer was therefore uncertain as to the signal frequency on any one trial but did know the range of possible signal frequencies. In both conditions the signals were adjusted in amplitude so that their detectability would be the same at all frequencies. The effect of frequency uncertainty was very small. In the most extreme condition, in which the signal ranged from 500 to 4000 cps, the decrement in detectability was approximately 3 db when compared with the condition where the signal frequency was fixed. The results are inconsistent with several models which have been proposed to explain the effects of signal uncertainty. The data also raise questions about the interpretation of the “critical-band” experiments.