Page 32 - Volume 12, Issue 2 - Spring 2012

Fig. 2. Spectrum for the vowel /a/ and excitation patterns calculated for normal-hearing listeners and two listeners with hearing loss. Excitation patterns indicate good preser- vation of formant peaks in the normal system, but much impoverished representation of higher formants in the impaired systems.
bank of filters may differ between a normal cochlea and two different simulated severities of cochlear damage. The solid black line represents the normal excitation pattern in response to the vowel sound shown in the figure. The peaks at low frequencies show the response to the individual har- monics of the vowel sound because the filter bandwidths at lower frequencies are narrow, and only one or two harmon- ics of the vowel are processed through each filter. The overall peaks around 600 Hz, 2600 Hz, and 3300 Hz represent the first three formant or resonant frequencies that identify the sound. The dashed line displays an excitation pattern gener- ated by simulating a filter bank with twice normal bandwidth filters, and the dotted line shows three times normal filter bandwidths. Typically, for a moderate-to-severe hearing loss, auditory filters may be from two to four times broader than in a normal cochlea (Moore and Glasberg, 2004).
The excitation pattern for the normal-hearing system shows clear peaks and valleys that represent the formants that characterize this particular vowel sound. To the extent that these peaks and valleys are less precisely represented in the auditory system, the speech sounds will be less well identified
and there will be more ambiguity in the speech percept. This is demonstrated in the top two excitation patterns, showing representations of the vowel in the case of mild and moder- ate hearing losses. With each broadening in the filter band- widths, the excitation pattern becomes flatter, with less defi- nition of the peaks and valleys in the vowel spectrum. As the peaks are less well-defined the vowels are perceived with much greater ambiguity because a hearing-impaired listener may not accurately process the spectral detail (Molis and Leek, 2011).
In addition to a smearing of the peaks and valleys in the internal excitation pattern, the broader auditory filters in hearing-impaired listeners interfere with normal perception in yet another way. The excitation pattern calculation only uses the auditory spectrum, but the input to the ear is always a temporal waveform, not an amplitude spectrum. The con- sequences of separating a complex waveform into frequency channels, but preserving the temporal characteristics of the channels, rather than limiting the output to only the spec- trum, may be seen using a computer model of auditory pro- cessing developed by Patterson and his colleagues (1995)
Discover new heights in acoustics design
Room Acoustics Software
www.odeon.dk
Hearing Loss and Frequency Analysis of Complex Sounds 31