Speech recognition performance at loudness discomfort level

Die Verständlichkeitskurve für den Freiburger Einsilbertest im Störschall mit einem Signal-Rausch-Abstand (SNR) von 5 dB

Hintergrund

Der Freiburger Einsilbertest (FBE) ist seit Langem einer der wesentlichen deutschen Sprachhörtests. Seit einiger Zeit wird er auch zur Messung der Hörverbesserung im Rahmen einer Hörgeräteversorgung im Störschall verwendet (FBE-S). Mit dieser Studie soll eine Verständlichkeitskurve für den Freiburger Einsilbertest im Störschall mit fixem Signal-Störschall-Verhältnis von 5 dB bei verschiedenen Schalldruckpegeln bestimmt werden.

Methode

Im Herbst 2018 wurden 60 normalhörende, muttersprachlich Deutsch sprechende Probanden im Alter zwischen 18 und 31 Jahren in der Klinik für HNO-Heilkunde des Bundeswehrkrankenhauses in Hamburg untersucht. Das Sprachverstehen wurde mit einer Testliste des FBE mit einem Schalldruckpegel von 15 bis 90 dB und einem 5 dB niedrigeren Störschall in 5‑dB-Schritten gemessen. Anschließend wurden von den Verständlichkeitsquoten die Mittelwerte gebildet und die 95%-Konfidenzintervalle (95%-KI) bestimmt.

Ergebnisse

Es nahmen 29 weibliche und 31 männliche Probanden an der Untersuchung teil. Das mittlere Alter aller Probanden betrug 24,32 Jahre (±3,39 Jahre). In der Varianzanalyse für fixe Effekte mit der Verständlichkeit als abhängiger Variablen zeigte sich ein hochsignifikanter Zusammenhang zwischen den Schalldruckpegeln des Sprach- und Störschalls auf der einen Seite und der Verständlichkeit auf der anderen Seite (p < 0,001). Die gemittelten Verständlichkeitsquoten und zugehörigen 95%-KI sowie die Häufigkeitsverteilungen wurden tabellarisch und grafisch dargestellt.

Schlussfolgerung

Die Verständlichkeitskurve des FBE‑S ist im Vergleich zur Normkurve des FBE nach rechts verschoben. Die gemittelte Kurve des FBE‑S erreicht bei einem Schalldruckpegel von 70 dB und einem Störschalldruckpegel von 65 dB den Sättigungsbereich von 90 % Verständlichkeit (zum Vergleich: Der Sättigungspunkt von 100 % Verständlichkeit liegt beim FBE ohne Störschall bei 55 dB). Durch die hier vorgestellte Verständlichkeitskurve lassen sich individuelle Messergebnisse ohne und mit Hörgeräten besser interpretieren. Künftig sollte der Hörgewinn für Hörgeräte möglicherweise bei niedrigeren Schalldruckpegeln erfolgen als heute.

A demonstration of improved precision of word recognition scores

PURPOSE The purpose of this study was to demonstrate improved precision of word recognition scores (WRSs) by increasing list length and analyzing phonemic errors. METHOD Pure-tone thresholds (frequencies between 0.25 and 8.0 kHz) and WRSs were measured in 3 levels of speech-shaped noise (50, 52, and 54 dB HL) for 24 listeners with normal hearing. WRSs were obtained for half-lists and full lists of Northwestern University Test No. 6 (Tillman & Carhart, 1966) words presented at 48 dB HL. A resampling procedure was used to derive dimensionless effect sizes for identifying a change in hearing using the data. This allowed the direct comparison of the magnitude of shifts in WRS (%) and in the average pure-tone threshold (dB), which provided a context for interpreting the WRS. RESULTS WRSs based on a 50-word list analyzed by the percentage of correct phonemes were significantly more sensitive for identifying a change in hearing than the WRSs based on 25-word lists analyzed by percentage of correct words. CONCLUSION Increasing the number of items that contribute to a WRS significantly increased the test's ability to identify a change in hearing. Clinical and research applications could potentially benefit from a more precise word recognition test, the only basic audiologic measure that estimates directly the distortion component of hearing loss and its effect on communication.

A comparison of presentation levels to maximize word recognition scores

BACKGROUND

While testing suprathreshold word recognition at multiple levels is considered best practice, studies on practice patterns do not suggest that this is common practice. Audiologists often test at a presentation level intended to maximize recognition scores, but methods for selecting this level are not well established for a wide range of hearing losses.

PURPOSE

To determine the presentation level methods that resulted in maximum suprathreshold phoneme-recognition scores while avoiding loudness discomfort.

RESEARCH DESIGN

Performance-intensity functions were obtained for 40 participants with sensorineural hearing loss using the Computer-Assisted Speech Perception Assessment. Participants had either gradually sloping (mild, moderate, moderately severe/severe) or steeply sloping losses. Performance-intensity functions were obtained at presentation levels ranging from 10 dB above the SRT to 5 dB below the UCL (uncomfortable level). In addition, categorical loudness ratings were obtained across a range of intensities using speech stimuli. Scores obtained at UCL - 5 dB (maximum level below loudness discomfort) were compared to four alternative presentation-level methods. The alternative presentation-level methods included sensation level (SL; 2 kHz reference, SRT reference), a fixed-level (95 dB SPL) method, and the most comfortable loudness level (MCL). For the SL methods, scores used in the analysis were selected separately for the SRT and 2 kHz references based on several criteria. The general goal was to choose levels that represented asymptotic performance while avoiding loudness discomfort. The selection of SLs varied across the range of hearing losses.

RESULTS

Scores obtained using the different presentation-level methods were compared to scores obtained using UCL - 5 dB. For the mild hearing loss group, the mean phoneme scores were similar for all presentation levels. For the moderately severe/severe group, the highest mean score was obtained using UCL - 5 dB. For the moderate and steeply sloping groups, the mean scores obtained using 2 kHz SL were equivalent to UCL - 5 dB, whereas scores obtained using the SRT SL were significantly lower than those obtained using UCL - 5 dB. The mean scores corresponding to MCL and 95 dB SPL were significantly lower than scores for UCL - 5 dB for the moderate and the moderately severe/severe group.

CONCLUSIONS

For participants with mild to moderate gradually sloping losses and for those with steeply sloping losses, the UCL - 5 dB and the 2 kHz SL methods resulted in the highest scores without exceeding listeners' UCLs. For participants with moderately severe/severe losses, the UCL - 5 dB method resulted in the highest phoneme recognition scores.

Most comfortable and uncomfortable loudness levels: six decades of research

This article critically reviews the influence of such factors as psychophysical testing method, stimulus type, and instructional set on most comfortable loudness (MCL) and uncomfortable loudness (UCL) levels. Generally, research indicates that test methods and instructions strongly affect both MCL and UCL while stimulus conditions affect them less substantially. Overall, the data suggest lower reliability for MCL than for UCL and lower reliability for pure-tone MCLs than for speech MCLs. Lower MCLs are typically obtained when measured by an ascending approach, in contrast to a descending approach. Results suggest that audiological efforts should be directed toward the development of a standardized test procedure that yields adequately reliable and valid MCLs and UCLs for routine clinical use.

Speech recognition performance of patients with sensorineural hearing loss under unaided and aided conditions using linear and compression hearing AIDS

OBJECTIVES

This study compared speech recognition performance on the Northwestern University Auditory Test No. 6 (NU-6) and the Connected Speech Test (CST) for three hearing aid circuits (peak clipping [PC], compression limiting [CL], and wide dynamic range compression [WDRC]) in adults with symmetrical sensorineural hearing loss. The study also questioned whether or not hearing aid benefit for the three circuits was dependent upon the speech level and the signal-to-babble ratio (S/B) and upon the degree and slope of hearing loss.

DESIGN

Unaided speech recognition performance for NU-6 and CST materials presented from a loudspeaker at 0 degrees was measured during Visit 1, and both unaided and aided performance was measured at 3-mo intervals during Visits 2 to 4. The NU-6 was presented in quiet at a conversational speech level of 62 dB SPL. The CST was presented in 10 listening conditions-three S/B (-3, 0, and 3 dB) at each of three speech levels (soft speech at 52 dB SPL, conversational speech at 62 dB SPL, and loud speech at 74 dB SPL) and in quiet at 74 dB SPL. Uncorrelated multi-talker babble was presented from two loudspeakers at 45 degrees on each side of the main speaker. Hearing aid benefit was examined for 360 subjects divided into four groups of hearing loss, pure tone average <40 dB HL and slope <10 dB/octave or >10 dB/octave and hearing loss >40 dB HL for the two slope categories.

RESULTS

Hearing aid benefit (aided minus unaided performance) measured on the NU-6 in quiet exceeded 31 rau for all three circuits. Although small statistical advantages were found for the WDRC, the differences were approximately 2% and are not considered clinically relevant. Unaided CST performance showed a complex relationship between presentation level and signal-to-babble ratio that was further confounded by the degree of hearing loss. For the two mild hearing loss groups and for each of the three nominal signal-to-babble ratios, CST performance decreased by 20 rau for the -3 dB S/B to 6 rau for the 3 dB S/B as speech level increased from 52 to 74 dB SPL. In contrast, unaided performance increased by 32 to 13 rau with signal level for all signal-to-babble ratios for the two >40 dB hearing loss groups. Overall, aided CST performance exceeded unaided performance for all 10 conditions. As expected, hearing aid benefit was greatest (27 rau) for soft speech and smallest for loud speech (6 rau). Differences among the hearing aid circuits were small with only one significant difference; the WDRC at 62/0 was poorer by 3 rau than the other two circuits. When the CST data were analyzed as a function of hearing loss, five pair-wise comparisons were significant. In contrast to the unaided performance, aided performance for all hearing loss groups decreased as presentation level increased, even though the signal-to-babble ratio was constant.

CONCLUSIONS

All three hearing aids circuits provided benefit over the unaided condition in both quiet and noise. The greatest benefit was measured for soft speech in the more severe hearing loss groups. Although only small differences were measured among the three hearing aid circuits, significant differences favored the PC and CL circuits over the WDRC in the mild hearing loss groups and favored the WDRC over the PC in the more severe, sloping hearing loss group. An interesting interaction between speech level, signal-to-babble ratio, degree of hearing loss, and amplification was found. For a constant signal-to-babble ratio, recognition performance decreased as speech level increased from 52 to 74 dB SPL. The effect was most marked in the milder hearing loss groups and in the aided conditions, and occurred at even the lowest speech levels.

Frequency and temporal resolution in elderly listeners with good and poor word recognition

There is a subgroup of elderly listeners with hearing loss who can be characterized by exceptionally poor speech understanding. This study examined the hypothesis that the poor speech-understanding performance of some elderly listeners is associated with disproportionate deficits in temporal resolution and frequency resolution, especially for complex signals. Temporal resolution, as measured by gap detection, and frequency resolution, as measured by the critical ratio, were examined in older listeners with normal hearing, older listeners with hearing loss and good speech-recognition performance, and older listeners with hearing loss and poor speech-recognition performance. Listener performance was evaluated for simple and complex stimuli and for tasks of added complexity. In addition, syllable recognition was assessed in quiet and noise. The principal findings were that older listeners with hearing loss and poor word-recognition performance did not perform differently from older listeners with hearing loss and good word recognition on the temporal resolution measures nor on the spectral resolution measures for relatively simple stimuli. However, frequency resolution was compromised for listeners with poor word-recognition abilities when targets were presented in the context of complex signals. Group differences observed for syllable recognition in quiet were eliminated in the noise condition. Taken together, the findings support the hypothesis that unusual deficits in word-recognition performance among elderly listeners were associated with poor spectral resolution for complex signals.

Monosyllabic word recognition at higher-than-normal speech and noise levels

The effects of intensity on monosyllabic word recognition were studied in adults with normal hearing and mild-to-moderate sensorineural hearing loss. The stimuli were bandlimited NU#6 word lists presented in quiet and talker-spectrum-matched noise. Speech levels ranged from 64 to 99 dB SPL and S/N ratios from 28 to -4 dB. In quiet, the performance of normal-hearing subjects remained essentially constant in noise, at a fixed S/N ratio, it decreased as a linear function of speech level. Hearing-impaired subjects performed like normal-hearing subjects tested in noise when the data were corrected for the effects of audibility loss. From these and other results, it was concluded that: (1) speech intelligibility in noise decreases when speech levels exceed 69 dB SPL and the S/N ratio remains constant; (2) the effects of speech and noise level are synergistic; (3) the deterioration in intelligibility can be modeled as a relative increase in the effective masking level; (4) normal-hearing and hearing-impaired subjects are affected similarly by increased signal level when differences in speech audibility are considered; (5) the negative effects of increasing speech and noise levels on speech recognition are similar for all adult subjects, at least up to 80 years; and (6) the effective dynamic range of speech may be larger than the commonly assumed value of 30 dB.

Normal and hearing-impaired word recognition scores for monosyllabic words in quiet and noise

The effects of noise on word recognition scores were assessed with normal-hearing and hearing-impaired subjects. Fifty-one normal-hearing subjects were tested at 50 dB HL using signal-to-noise ratios (S/Ns) of 5, 10, and 15 dB. Thirty subjects with mild-to-moderate sensorineural hearing losses were tested in quiet and in noise at S/Ns of 10 dB and 15 dB. Monosyllabic words in a Multitalker Noise were selected for testing. Mean scores for the normal-hearing subjects were 45% at the 5 dB S/N, 74% at the 10 dB S/N, and 87% at the 15 dB S/N. For the hearing-impaired subjects, scores were 85% in quiet, 60% at the 15 dB S/N, and 40% at the 10 dB S/N. These results suggest that background noise which is mildly disruptive for normal hearing subjects can be highly disruptive to hearing-impaired subjects. Moreover, these findings indicate that subjects with mild-to-moderate sensorineural hearing loss require a more favourable S/N than normal listeners to achieve comparable word recognition scores. Test-retest differences for word recognition scores revealed variability that agreed closely with predictions based on the binomial distribution for both groups of subjects. Speech-in-noise abilities must be measured directly because regression equations revealed that speech-in-noise scores cannot be predicted accurately from either puretone thresholds or speech-in-quiet scores. Word recognition functions are presented from several hearing-impaired subjects and demonstrate the value of testing in noise.

A Revised Speech Intelligibility Rating (RSIR) test: listeners with normal hearing

Two experiments were conducted to examine the intelligibility of 72 passages of connected discourse prepared by Cox and McDaniel in their development of the Speech Intelligibility Rating (SIR) test. Intelligibility was assessed with a method-of-adjustment (MOA) procedure in which listeners adjusted the level of a multi-talker babble until they could just understand 50% of a passage; the measure of intelligibility was the signal-to-babble ratio, dB S/B. The objective was to develop a Revised Speech Intelligibility Rating (RSIR) test that would comprise a large number of equivalent passages that produce reliable intelligibility measures. In experiment 1, the S/B ratio was based on the overall root-mean-square (rms) levels of speech and babble, as represented by the average level of frequent peaks observed on a VU meter. Across all 72 passages, mean intelligibility was -1.43 dB S/B, and the measure of intelligibility for 42 passages was within +/- 0.5 dB of the overall mean for all 72 passages. In experiment 2, the S/B ratio was based on long-term rms levels of speech and babble measured in 16 one-third-octave bands, with center frequencies from 160 to 5000 Hz. In an effort to achieve greater equivalence in intelligibility among passages, the overall rms level of each passage was attenuated by the difference between SB16-band for an individual passage and S/B16-band for a reference passage. Mean intelligibility across all 72 passages was -8.06 dB, and the measure of intelligibility was within +/- 0.5 dB of the overall mean for 64 of the 72 passages. For those 64 passages, the 95% critical difference for five MOAs was 0.72 dB, which corresponds to an estimated percentage critical difference of 10.8%.

Loudness discomfort level and reliability as a function of instructional set

Loudness discomfort levels (LDLs) for a 12-speaker babble were measured in 20 subjects with normal hearing sensitivity for two instructional sets, listening in the sound field. One instructional set asked the listener to indicate when they would choose 'not to listen for any period of time'. The second instructional set asked the listeners to indicate when they 'would choose not to listen for 15 minutes or longer'. A simple up-down adaptive procedure was used to measure LDLs. Data were gathered on both instructional sets for all listeners. LDLs for each instructional set were repeated on four separate occasions within a range of 1-3 weeks. The average LDLs for the instructions utilizing the criterion of choosing not to listen for any period of time were 9 dB higher than the criterion of choosing not to listen for 15 minutes or longer. The reliability of LDLs for both instructional sets was high, ranging from correlations of 0.90-0.97 for the criterion of choosing not to listen for any period of time, and ranging from correlations of 0.62-0.89 for the criterion of listening for 15 minutes or longer. The criterion of choosing not to listen for any period of time resulted in slightly more reliable measures. These results support the importance of considering the instructions given to a person to measure LDLs, and hence for where to set the saturation sound pressure level of a hearing aid. Also, with appropriate methods, LDLs can be reliably measured in the sound field.