Acoustical correlates of performance on a dynamic range compression discrimination task

The perceptual limitations of troubleshooting hearing-aids based on patients' descriptions

OBJECTIVES

Hearing-aid frequency-gain responses are routinely adjusted by clinicians to patient preferences and descriptions. This study measured the minimum gain adjustments required to elicit preferences, and the assignment of descriptors to gain adjustments, to perceptually evaluate description-based troubleshooting.

DESIGN

Participants judged whether short sentences with ±0-12 dB gain adjustments in one of three frequency bands were "better", "worse" or "no different" from the same sentence at their individual real-ear or prescribed gain. If judged "better" or "worse", participants were then asked to assign one of the six common sound-quality descriptors to their preference.

STUDY SAMPLE

Thirty-two adults (aged 51-75 years) all with hearing-aid experience.

RESULTS

Median preference thresholds, the minimum gain adjustments to elicit "better" or "worse" judgments, ranged from 4 to 12 dB, increasing with frequency. There was some between-participant agreement in preferences: participants generally preferred greater low-frequency gain. Within-participant reliability for preferences was moderate. There was, however, little between-participant agreement in descriptor selection for gain adjustments. Furthermore, within-participant reliability for descriptor selection was lacking.

CONCLUSIONS

The scale of gain adjustments necessary to elicit preferences, along with the low agreement and reliability in descriptors for these adjustments questions the efficiency and efficacy of current description-based troubleshooting, especially with short speech stimuli.

Discrimination of Gain Increments in Speech-Shaped Noises

Frequency-dependent gain adjustments are routine in hearing-aid fittings, whether in matching to real-ear targets or fine-tuning to patient feedback. Patient feedback may be unreliable and fittings inefficient if adjustments are not discriminable. To examine what gain adjustments are discriminable, we measured the just-noticeable differences (JNDs) for level increments in speech-shaped noises processed with prescription gains. JNDs were measured in the better ears of 38 participants with hearing impairment using a fixed-level, same-different task. JNDs were measured for increments at six individual frequency-bands: a 0.25-kHz low-pass band; octave-wide bands at 0.5, 1, 2, and 4 kHz; and a 6-kHz high-pass band. JNDs for broadband increments were also measured. JNDs were estimated at d' of 1 for a minimally discriminable increment in optimal laboratory conditions. The JND for frequency-band increments was 2.8 dB excluding the 0.25-kHz low-pass band, for which the JND was 4.5 dB. The JND for broadband increments was 1.5 dB. Participants' median frequency-band and broadband JNDs were positively correlated. JNDs were mostly independent of age, pure-tone thresholds, and cognitive score. In consideration of self-fitting adjustments in noisier conditions, JNDs were additionally estimated at a more sensitive d' of 2. These JNDs were 6 dB for bands below 1 kHz, and 5 dB for bands at and above 1 kHz. Overall, the results suggest noticeable fine-tuning adjustments of 3 dB and self-fitting adjustments of 5 dB.

Object-based attention modulates the discrimination of level increments in stop-consonant noise bursts

This study tested the hypothesis that object-based attention modulates the discrimination of level increments in stop-consonant noise bursts. With consonant-vowel-consonant (CvC) words consisting of an ≈80-dB vowel (v), a pre-vocalic (Cv) and a post-vocalic (vC) stop-consonant noise burst (≈60-dB SPL), we measured discrimination thresholds (LDTs) for level increments (ΔL) in the noise bursts presented either in CvC context or in isolation. In the 2-interval 2-alternative forced-choice task, each observation interval presented a CvC word (e.g., /pæk/ /pæk/), and normal-hearing participants had to discern ΔL in the Cv or vC burst. Based on the linguistic word labels, the auditory events of each trial were perceived as two auditory objects (Cv-v-vC and Cv-v-vC) that group together the bursts and vowels, hindering selective attention to ΔL. To discern ΔL in Cv or vC, the events must be reorganized into three auditory objects: the to-be-attended pre-vocalic (Cv-Cv) or post-vocalic burst pair (vC-vC), and the to-be-ignored vowel pair (v-v). Our results suggest that instead of being automatic this reorganization requires training, in spite of using familiar CvC words. Relative to bursts in isolation, bursts in context always produced inferior ΔL discrimination accuracy (a context effect), which depended strongly on the acoustic separation between the bursts and the vowel, being much keener for the object apart from (post-vocalic) than for the object adjoining (pre-vocalic) the vowel (a temporal-position effect). Variability in CvC dimensions that did not alter the noise-burst perceptual grouping had minor effects on discrimination accuracy. In addition to being robust and persistent, these effects are relatively general, evincing in forced-choice tasks with one or two observation intervals, with or without variability in the temporal position of ΔL, and with either fixed or roving CvC standards. The results lend support to the hypothesis.

Comparison of the CAM2A and NAL-NL2 hearing-aid fitting methods for participants with a wide range of hearing losses

OBJECTIVE

To compare preferences for sounds processed via a simulated five-channel compression hearing aid fitted using CAM2A and NAL-NL2.

DESIGN

Within a trial, the same segment of sound was presented twice, once with CAM2A settings and once with NAL-NL2 settings, in random order. The participant indicated which one was preferred and by how much. Stimuli included female and male speech in quiet and four types of music. The compression speed was slow or fast and the input sound level was 50, 65, or 80 dB SPL.

STUDY SAMPLE

Sixteen experienced hearing-aid users with a wide range of sensorineural hearing losses.

RESULTS

For both compression speeds, CAM2A was slightly preferred over NAL-NL2 for input levels of 65 and 80 dB, but NAL-NL2 was slightly preferred at 50 dB SPL.

CONCLUSIONS

Preferences for CAM2A relative to NAL-NL2 vary with input level. The results suggest that preferences for CAM2A might be increased by using lower gains for high frequencies and low input levels.

Sentence intelligibility during segmental interruption and masking by speech-modulated noise: Effects of age and hearing loss

This study investigated how single-talker modulated noise impacts consonant and vowel cues to sentence intelligibility. Younger normal-hearing, older normal-hearing, and older hearing-impaired listeners completed speech recognition tests. All listeners received spectrally shaped speech matched to their individual audiometric thresholds to ensure sufficient audibility with the exception of a second younger listener group who received spectral shaping that matched the mean audiogram of the hearing-impaired listeners. Results demonstrated minimal declines in intelligibility for older listeners with normal hearing and more evident declines for older hearing-impaired listeners, possibly related to impaired temporal processing. A correlational analysis suggests a common underlying ability to process information during vowels that is predictive of speech-in-modulated noise abilities. Whereas, the ability to use consonant cues appears specific to the particular characteristics of the noise and interruption. Performance declines for older listeners were mostly confined to consonant conditions. Spectral shaping accounted for the primary contributions of audibility. However, comparison with the young spectral controls who received identical spectral shaping suggests that this procedure may reduce wideband temporal modulation cues due to frequency-specific amplification that affected high-frequency consonants more than low-frequency vowels. These spectral changes may impact speech intelligibility in certain modulation masking conditions.

Sentence intelligibility during segmental interruption and masking by speech-modulated noise: Effects of age and hearing loss

This study investigated how single-talker modulated noise impacts consonant and vowel cues to sentence intelligibility. Younger normal-hearing, older normal-hearing, and older hearing-impaired listeners completed speech recognition tests. All listeners received spectrally shaped speech matched to their individual audiometric thresholds to ensure sufficient audibility with the exception of a second younger listener group who received spectral shaping that matched the mean audiogram of the hearing-impaired listeners. Results demonstrated minimal declines in intelligibility for older listeners with normal hearing and more evident declines for older hearing-impaired listeners, possibly related to impaired temporal processing. A correlational analysis suggests a common underlying ability to process information during vowels that is predictive of speech-in-modulated noise abilities. Whereas, the ability to use consonant cues appears specific to the particular characteristics of the noise and interruption. Performance declines for older listeners were mostly confined to consonant conditions. Spectral shaping accounted for the primary contributions of audibility. However, comparison with the young spectral controls who received identical spectral shaping suggests that this procedure may reduce wideband temporal modulation cues due to frequency-specific amplification that affected high-frequency consonants more than low-frequency vowels. These spectral changes may impact speech intelligibility in certain modulation masking conditions.