Loudness scaling revisited.

The present work was undertaken in an attempt to evaluate whether it is reasonable to expect that categorical loudness scaling can provide useful information for nonlinear hearing aid fitting. Normative data from seven scaling procedures show that the individual procedures relate the perceptual categories differently to sound level and with a substantial between-subject variance. Hearing-impaired data from four studies demonstrate that the inverse slope of the loudness function varies linearly with hearing loss and with a constant variance. In relation to hearing aid fitting, the slope can, in most cases, be predicted from the hearing loss with an accuracy within the range of a normal finetuning. For the fitting of nonlinear hearing aids, the statistical properties of both normal and impaired loudness functions are equally important. The present analysis strongly suggests that categorical loudness scaling cannot, in general, provide significant information for the fitting process.

Non-linear signal processing in digital hearing aids.

Three different non-linear digital signal processing algorithms were developed; LinEar, DynEar and RangeEar. All three provided individual frequency shaping via a seven-band low-power filterbank and compression in two channels. RangeEar and DynEar used wide dynamic range syllabic compression in the low-frequency (LF) channel, while LinEar used compression limiting. In the high-frequency (HF) channel, RangeEar used a slow-acting automatic volume control, while DynEar and LinEar used compression limiting. Wearable digital signal processing-based experimental instruments were used to evaluate the fitting algorithms under real world conditions with experienced hearing aid users. Evaluation included laboratory testing of speech recognition in noise and questionnaires on sound quality ratings. Results did not indicate one general good-for-all algorithm, but different algorithms resulting in preference and performance depending on the hearing loss configuration. Preference for any of the new algorithms could be predicted based on auditory dynamic range measurements. It was hypothesized that the different preferences were affected by different susceptibility to masking of HF sounds by amplified LF sounds.

A digital filterbank hearing aid: three digital signal processing algorithms--user preference and performance.

OBJECTIVE: Three digital signal processing algorithms named RangeEar, DynEar, and LinEar were compared with regard to user preference and performance when a wearable digital filterbank hearing aid was used. All three algorithms provided individual frequency shaping via a seven-band filterbank. Compression was used in a low-frequency (LF) and a high-frequency (HF) channel. RangeEar and DynEar used wide dynamic range syllabic compression in the LF channel, whereas LinEar used compression limiting. In the HF channel, RangeEar used a slow acting automatic volume control, whereas DynEar and LinEar used compression limiting. The subjects had access to a manual volume control when using the LinEar or DynEar options. DESIGN: The study included 13 hearing aid users with symmetrical sensorineural losses. In a 1 mo long blind field test, the RangeEar algorithm was compared with the preferred algorithm from an earlier study, DynEar or LinEar. A data logger function was included for objective recording of the total time each algorithm was used and how the volume controls were used. The preference was based on the time used for each algorithm and from subjective statements. Threshold signal-to-noise ratio (S/N-threshold) for speech was tested, and sound quality ratings were obtained through a questionnaire. RESULTS: Of the 13 subjects, six preferred the RangeEar fitting and another four preferred the DynEar fitting. Two subjects preferred the LinEar fitting and one had equal preference for RangeEar and LinEar. The results from the questionnaire showed that the preferred fittings were rated higher concerning overall impression of sound quality and clearness, whereas the S/N for the speech test did not show any differences. Preferences, where stated, could be predicted from auditory dynamic range measurements in the LF and HF frequency ranges. The mean dynamic range was broader for low and narrower for high frequencies for those who preferred the RangeEar or DynEar fitting as compared with those who preferred the LinEar fitting. The preference between RangeEar and DynEar was predicted by differences in the HF range, with the narrower dynamic range for the DynEar preference subjects. CONCLUSION: Most subjects preferred the option of having a wide dynamic range syllabic compressor in the LF channel and having the overall gain in the HF channel adjustable, either manually (DynEar) or automatically (RangeEar).

A digital filterbank hearing aid: predicting user preference and performance for two signal processing algorithms.

OBJECTIVE: In a series of experiments with a wearable binaural digital hearing aid, two hearing aid processing algorithms were compared. Both algorithms provided individual frequency shaping via a seven-band filterbank with compression limiting in the high-frequency channel. They differed in the processing of the low-frequency channel, using dynamic range compression for one (DynEar) and linear processing with compression limiting for the other (LinEar). In a pilot field test we found that LinEar/ DynEar preference based on use time could be predicted from auditory dynamic range data. For the subjects who preferred DynEar, the mean dynamic range was broader for low and mid frequencies and narrower for high frequencies, as compared with the LinEar preference subjects. These groupings were tested as predictors of user preference and performance in a main field test. DESIGN: The main study included 26 hearing aid users with symmetrical sensorineural losses. The algorithms were compared in a one-mo-long blind field test. A data logger function was included for objective recording of the total time each algorithm was used and how the volume controls were used. The preference was based on the time used for each algorithm and on subjective statements. Threshold signal-to-noise ratio (S/N-threshold) for speech was tested, and sound quality ratings were obtained through a questionnaire. We also tested the S/N-thresholds for the subjects' conventional (own) aids. RESULTS: The preference was correctly predicted by the dynamic range data on 12 out of 15 new cases. S/N-thresholds were lower for the preferred fittings compared with the nonpreferred fittings and with the subjects' own aids. In the questionnaire the preferred fittings were rated significantly higher in terms of overall impression and clearness. Because of the systematic way the DynEar-preference subjects adjusted the high-frequency DynEar gain, we speculate that upward spread of masking may have been a factor in preference and performance. Additionally, LinEar-preference subjects' preference and performance might have been influenced by excessive compression ratios with the DynEar processing in these cases. CONCLUSIONS: 1. Preference for DynEar versus LinEar depends on the auditory dynamic range. 2. S/N-thresholds for speech were better for the preferred fittings, which also were rated higher in terms of overall impression of sound quality and clearness.

A digital filterbank hearing aid. Improving a prescriptive fitting with subjective adjustments.

Two fitting algorithms for linear hearing aids were compared using a wearable digital hearing aid in a one-month blind field test: a prescriptive method (POGO II) and a new algorithm, LinEar. Both used seven bands for frequency shaping, and two channel compression limiting. When fitting LinEar, the subjects individually adjusted the frequency response according to specified criteria. LinEar used a lower compression threshold setting than prescribed by POGO II. Eight subjects tested the two algorithms in a one-months blind field test as well as in the laboratory. The individual LF- and HF-gain adjustments of the frequency response in LinEar showed rather large variations compared to the POGO II prescription. Measures of S/N for speech did not show any significant differences between LinEar and POGO II, while overall sound quality ratings in laboratory and field test showed that LinEar was rated significantly higher than POGO II.

Use of quantitative measures of auditory brain-stem response peak amplitude and residual background noise in the decision to stop averaging.

An objective quantitative approach to the decision of when to stop averaging sweeps in auditory brain-stem response (ABR) testing is presented. This decision is based on (1) the knowledge of the amplitude distributions of wave V in the ABRs of normal hearing individuals for varying stimulus levels, (2) calculated estimates of the residual background noise in the average, and (3) use of a quantitative statistical detector of an evoked potential. Several reasons for terminating an average are presented along with a specific protocol for each of the reasons. These protocols provide a general but consistent framework to address the issue of when to stop averaging and should improve the efficiency of ABR testing. Furthermore, it is quite possible to automate the procedure and the decision process.

Evaluating residual background noise in human auditory brain-stem responses.

The nature of the residual background noise in ABR averages was empirically examined in normal hearing objects. The residual noise in the average was estimated with use of the technique described by Elberling and Don [Scand. Audiol. 13, 187-197 (1984)]. Low-level click stimuli were presented in 2-dB steps spanning the range from 30 to 48 dB p-p.e. SPL. For each stimulus level, 10,000 sweeps were acquired and stored for analysis. Shortcomings of the use of artifact rejection and standard averaging are demonstrated. It is further demonstrated how application of the Bayesian estimation technique of Elberling and Wahlgreen [Scand. Audiol. 14, 89-96 (1985)] to form weighted averages can help minimize these shortcomings. Finally, the effects of smaller sweep block sizes on the Bayesian technique's ability to control the destructive effects of nonstationary noise are analyzed. Minimizing the destructive effects increases the value of statistical techniques used to detect objectively or to control the quality of ABR recordings. In all, these techniques in combination improve not only the accuracy of test interpretation but also the efficiency of clinical test time, which is becoming important for the control of medical costs.

Evoked acoustic emissions from the human ear. VI. Findings in cochlear hearing impairment.

Click-evoked otoacoustic emissions (CEOAE) were recorded in 28 subjects with mild to moderate flat or steeply sloping cochlear hearing loss. We used the same equipment and recording technique as previously employed in the testing of newborns. A rescaling and subtraction procedure was implemented in an attempt to eliminate the tail of the stimulus artifact. However, in some cases the method also seems to eliminate a true response. In ears with flat losses and with identified CEOAE, no one had a hearing loss exceeding 40 dB HL in the mid-frequency region (0.5, 1 and 2 kHz). Conversely, in ears with flat losses and without CEOAE, no one had a hearing loss less than 30 dB HL in this frequency region. In ears with sloping hearing losses the thresholds at 1 and 2 kHz were most important for the generation of the CEOAE and a significant correlation between the emission amplitude at 70aud (approximately dB p.e. SPL) and the threshold at 1 kHz was found.

The design and testing of a noise reduction algorithm based on spectral subtraction.

Three noise reduction algorithms based on amplitude subtraction were designed and used to process speech mixed with babble noise in two signal-to-noise ratios. The estimation of the noise-magnitude spectrum was performed with a novel synchro method, which exploits specific characteristics of the speech signal. The unprocessed and processed signals were evaluated psychoacoustically by means of a four-alternative-forced choice test with monosyllabic words (minimal pairs) in carrier phrases. The testing was carried out on groups of normally hearing and hearing-impaired subjects and the long-term power spectra of the processed signals were shaped to be essentially identical with those from the corresponding unprocessed signals. For the hearing-impaired subjects all signals were spectrally shaped according to the POGO-fitting rule. None of the algorithms improved speech intelligibility for any group of listeners and no change in the overall pattern of confusion was observed.

Evaluation of a noise reduction method--comparison between observed scores and scores predicted from STI.

Speech was mixed with different noise signals and then processed according to the well-known noise reduction method of 'spectral subtraction'. Three different algorithms were examined. The speech signals were subjected to a four alternative forced choice (4AFC) test. Both the processed and unprocessed signals were evaluated psycho-acoustically and objectively. Speech intelligibility was measured with the 4AFC test by presenting the signals via headphones to a group of normal-hearing and to a group of hearing-impaired listeners. The intelligibility scores were compared with the intelligibility scores predicted from a modified version of the Speech Transmission Index (STI). It appeared that although the noise reduction algorithms reduced the noise level, they did not improve the measured speech intelligibility, either for normal-hearing or for hearing-impaired listeners. This, however, was inconsistent with the scores estimated from STI, which erroneously predicted a significant improvement in intelligibility due to the noise reduction processing.

Hearing disorders in patients with insulin-dependent diabetes mellitus.

The cochlear and retrocochlear hearing function was evaluated in patients with long- and short-term insulin-dependent diabetes mellitus (IDDM) by means of psychoacoustic testing and auditory brain stem responses (ABR). Twenty patients with diabetic microangiopathy (median age 41 years, range 25-66 years) were examined. The median duration of their diabetes was 26 years (range 13-46 years). In addition, 19 patients without microangiopathy (median age 27 years, range 17-42 years) and with a median duration of the diabetes of 2 years (range 0-6 years) were examined. The metabolic control estimated by blood glucose concentration and glycosylated haemoglobin was identical in the two groups of IDDM patients. After correction for age and sex, no significant differences in hearing thresholds or discrimination scores were present between the two diabetic groups, or between the diabetic patients and an age- and sex-matched normal background population. In the patients with long-term IDDM, ABR produced abnormal responses in 40%, indicating the presence of diabetic encephalopathy, whereas ABR were abnormal in only 5% of the patients with short-term IDDM.

Prediction of intelligibility of non-linearly processed speech.

Speech in various noise backgrounds was processed through four different non-linear devices and the intelligibility of the processed signals was predicted from the Speech Transmission Index (STI). A novel calculation method was applied in order to avoid artifacts. Running speech was used as input signal and STI was calculated from the envelopes of the squared, noise-free speech signal and of the processed, squared, noisy signal in 23 critical bands. In situations with linearly processed speech and a stationary background noise, this calculation method gives results identical with the procedure described by Steeneken & Houtgast (6). However, in a number of situations with non-linearly processed speech, or a time varying background noise level, the calculation method used here is preferable. The predictions were evaluated in a psycho-acoustic listening test and the predictions agreed well with the listening test results.

DANTALE: a new Danish speech material.

A new Danish speech material (DANTALE) for clinical and experimental speech audiometry is digitally recorded on compact disc (CD). The speech material is designed to meet present audiological requirements at Danish hearing centres. One channel of the CD contains the speech signals and the other a masking noise. The CD also contains various calibration signals recorded on both channels at the end of the CD. The speech material compromises: 1) Digit triplets for the measurement of speech reception threshold (SRT). 2) Lists of monosyllabic words for the measurement of discrimination score (DS) for adults, children and small children. The word lists for the adults are equalized with regard to important phonetic and "visual" elements and the word lists for the children consist of minimal pairs. 3) Continuous speech for the measurement of the most comfortable loudness level (MCL), assessment of hearing aid fitting and the like. The masking noise is an amplitude-modulated, speech-shaped noise signal, which is designed to simulate a 4-person speech babble in order to assess both the frequency selectivity and the temporal resolution. The speech material is described and the long-term power spectra and modulation spectra are given.

Evoked acoustic emissions from the human ear. V. Developmental changes.

Twenty young children, in whom evoked acoustic emissions were recorded at birth, were re-examined at the age of 4 years. None of the children showed evidence of sensorineural hearing impairment and it was possible to record a reproducible emission in all ears, which displayed normal otoscopy and tympanometry (n = 9). The new recordings were compared with those obtained at birth and the latency and the amplitude of the response both appeared unchanged. However, in some ears the frequency content of the dominant part of the emission was considerably lower at the age of four. The implication of this finding is briefly discussed in view of recent data on cochlear development, obtained from animal research.

Evoked acoustic emissions from the human ear. IV. Final results in 100 neonates.

Evoked acoustic emissions were recorded from both ears in a series of 100 consecutive normal newborns. We used the same stimulus, a 2-kHz click, and recording technique as previously described. Analysis of the data showed that evoked emissions could be identified in all ears, except one at 70 dBaud (i.e. approximately 30 dB nHL). No significant differences could be demonstrated between males and females or between left and right ears with regard to the latency of the emissions, the peak-to-peak amplitude, the main frequency component, or the waveform correlation between the two 70 dBaud recordings in each ear. However, a significant correlation between left and right ears was found for the amplitude and frequency of the emissions. Practical and methodological problems related to the recording were elucidated. The tail of the stimulus artifact sometimes interfered with the first part of the emissions even though the recordings were made in a time window delayed 5 ms relative to the stimulus onset. We tried to solve this artifact problem by different off-line techniques, but found no useful solution. We therefore continued to use only a cosine tapering of the first 2 ms of the time window. Three different ways of determining latencies were evaluated and we found that the 'envelope' technique was the most simple and reliable. Recording of evoked acoustic emissions is a quick and non-invasive method and provided that the presence of the emissions is related to normal cochlear function, it can be used as a screening test in newborns.(ABSTRACT TRUNCATED AT 250 WORDS)

Reference data for ABRs in retrocochlear diagnosis.

A retrospective analysis has been carried out on the auditory brainstem responses obtained in 235 female and 249 male neurologically normal patients. The analysis is focused on the wave latencies and the I-V and I-III intervals are compared with the classification parameters sex, age and high-frequency hearing loss. Since age and hearing loss are correlated, a multiple regression analysis is used to analyse the effects from the two variables. Both intervals are found to differ with sex (i.e. females display smaller values than males), to increase with age, and to decrease with hearing loss. Corrections are made for the classification parameters relative to 25 years and 0 dB HL. For the I-V interval, we find a mean value of 4.01 ms for the females and 4.19 ms for the males and a residual variance (SD) of 0.20 ms. For the I-III interval, we find a mean value of 2.13 ms for the females and 2.27 ms for the males and a residual variance (SD) of 0.16 ms. However, for the I-III interval, only the male data show dependence of hearing loss. In 78 patients, the electrocochleographic N1 is used to ensure the wave I latency and there is in general a good agreement between the two latency measures. However, in patients with profound high-frequency hearing loss differences up to 0.5 ms can be found.

Threshold characteristics of the human auditory brain stem response.

Auditory brain stem responses (ABRs) were recorded from ten normal-hearing subjects in response to 100-microseconds clicks from a TDH 49 earphone at a rate of 48 pps and at levels randomly varied in 2-dB steps between 34 and 52 dB p.e. SPL. At each level, 10 000 epochs were averaged with use of a weighted concept and a running estimate was made of the signal-to-noise ratio (SNR). This quantity was used to detect the presence of the ABR and the median threshold was found at 38 dB p.e. SPL. The mean averaged background noise level was 11.3 nVrms, and the "true" ABRrms amplitude function crossed this value at 35.5 dB p.e. SPL, which indicates the level where the SNR = 1. By extrapolation, it was found that the ABR amplitude became zero at 32 dB p.e. SPL. The perceptual thresholds of the click were estimated by means of a modified block up-down procedure, and the median value was found at 33 dB p.e. SPL. The slope of the amplitude function and the magnitude of the averaged background noise are the two factors responsible for the ABR threshold sensitivity, which thus depends on both physiological and technical parameters. Therefore, these have to be considered together with the method of detection when the ABR is used to indicate the hearing sensitivity.

Detection functions for the human auditory brainstem response.

Previously obtained data characterizing the auditory brainstem response near the threshold for detection in 10 normal-hearing subjects are used to evaluate the detection method applied. The basic detection formula is described in terms of rates of true positive and false positive ABR detection and in combination with the normative ABR values used to calculate the ABR detection functions as well as the corresponding receiver operating characteristics (ROC curves). The observed distribution of the ABR-threshold levels is similar to that derived from the detection function, and therefore verifies the present results which are based partly on theoretical considerations.