Behavioral estimates of human frequency selectivity at low frequencies

A fundamental property of our hearing organ is its ability to break down sound into different spectral components, allowing us to make use of the richness in natural sound phenomena. Auditory filters, which conceptualize this property of the ear, however, have not been appropriately described at low sound frequencies. As a consequence of our lack of knowledge, we cannot accurately model our perception of complex low-frequency sound (such as that emitted by wind turbines or industrial processes, which can easily produce annoyance) nor make meaningful predictions of our perception based on physical sound measurements.

In this PhD thesis a detailed description of frequency selectivity at low frequencies is given. Different experiments have been performed to determine the properties of human auditory filters. Besides, loudness perception of low-frequency sinusoidal signals has been evaluated. In the analysis of results, factors that may influence our perception of low-frequency sound have been considered.

It was found that the relative sharpness of auditory filters is poor at low frequencies. However, their bandwidth decreases with decreasing frequency down to about 80 Hz. Below this, a worsening in tuning was observed, together with marked asymmetries, where steeper slopes below the center of the filter were observed. These effects could be explained as a probable consequence of the filtering produced by the middle-ear and the effect of the helicotrema on cochlear mechanics. Objective estimates of the effects of the latter factors were also found to highly influence the perception of loudness for low-frequency sinusoids.

A corollary of these results is that, when modelling our perception of complex sounds, the relative effect of tonal components is likely to be rated much higher (than e.g. at higher frequencies or considering larger bandwidths), due to the much narrower absolute bandwidths observed at low frequencies. Besides, observed increases in filter shape variability suggest that individual differences in perception are likely to increase at very low frequencies.