A direct approach for the design of chirp stimuli used for the recording of auditory brainstem responses

Abstract

A recent study evaluates auditory brainstem responses (ABRs) evoked by chirps of different durations (sweeping rates) [Elberling et al. (2010). J. Acoust. Soc. Am. 128, 215-223]. The study demonstrates that shorter chirps are most efficient at higher levels of stimulation whereas longer chirps are most efficient at lower levels. Mechanisms other than the traveling wave delay, in particular, upward spread of excitation and changes in cochlear-neural delay with level, are suggested to be responsible for these findings. As a consequence, delay models based on estimates of the traveling wave delay are insufficient for the design of chirp stimuli, and another delay model based on a direct approach is therefore proposed. The direct approach uses ABR-latencies from normal-hearing subjects in response to octave-band chirps over a wide range of levels. The octave-band chirps are constructed by decomposing a broad-band chirp, and constitute a subset of the chirp. The delay compensations of the proposed model are similar to those found in the previous experimental study, which thus verifies the results of the proposed model.

Figure 1

Amplitude-frequency characteristics of the filters used for the design of the broad-band CE-Chirp and the four octave-band chirps. The basic filter characteristics follow the specification for octave-band filters (class 1) given in IEC 61260 (1995). The amplitude spectra of the electrical waveforms of the stimuli are similar to the filter characteristics, whereas the acoustical spectra will be modified by the characteristics of the chosen earphone and the ear simulator in which the acoustic output is measured. See the text for further explanation.

Figure 2

The waveform and envelope of the broad-band CE-Chirp and the four octave-band chirps are shown. All five stimuli are plotted using the same (arbitrary) amplitude scale (in digital units, d.u.). The individual levels (in dB nHL), by which each of the octave-band stimuli contributes to the broad-band chirp at 60 dB nHL, are indicated to the right (see also Table 1). The zero point (0 ms) corresponds to the temporal location of the 10 000 Hz component of the CE-Chirp. See the text for further details.

Figure 3

Graphs show the ABR-latency as a function of stimulus level for each of the four octave-band chirps. (a) The x-axis refers to the level of the individual octave-band chirp in dB nHL (from Table 2). The broken parts of the curves for 500 and 1000 Hz indicate where estimated values are applied. (b) The x-axis refers to the level of the broad-band CE-Chirp in dB nHL (from Table 3). The differences between the longest and shortest latency are plotted for 30, 40, 50, and 60 dB nHL (Table 3) together with a continuous estimate (broken curve) of the latency difference as a function of level.

Figure 4

Final new delay models corresponding to the chirp levels 20, 40, 60, and 80 dB nHL. The power functions are vertically offset to have a zero delay at 10 000 Hz. For comparison, the delay model that was used to design the CE-Chirp is also shown. For each model function the change of the delay with frequency (between 5700 and 710 Hz) is indicated.

Figure 5

The waveforms of the chirps designed from the new delay models in Fig. 4 are shown. For comparison the CE-Chirp is also shown. All five chirps are plotted with an amplitude scale (in digital units, d.u.), which is three times larger than used in Fig. 2.

Figure 6

The waveforms of the octave-band chirps at 500, 1000, and 2000 Hz for the CE-Chirp (broken lines) and the Chirp 80 dB nHL (full line) are shown. All six stimuli are plotted with an amplitude scale (in digital units, d.u.), which is the same as used in Fig. 2. The zero point (0 ms) corresponds to the temporal location of the 10 000 Hz component of the CE-Chirp. See the text for further details.

Figure 7

Estimated response latencies from ten ears to the octave-band chirps at the levels corresponding to the broad-band chirp model at 80 and 40 dB nHL. Each individual data set (i.e., the response latencies at 500, 1000, 2000 and 4000 Hz) are fitted to Eq. 1 and the fitted power functions are shown in the figure together with the new delay model at 80 dB nHL (left) and 40 dB nHL (right). The figure shows the variability across the ten ears, as well as the significant effect of level. In contrast to Fig. 4, the power functions here are not vertically offset to provide a zero delay at 10 000 Hz.