|
High-frequency hearing (2010)
Noise reduction in hearing aids - Predictive ability of computational measures in terms of speech intelligibility, noise loudness and sound quality
Hearing aid Amplification at low input levels (2010)
Hearing aid noise reduction – Measured long-term average gain reduction for 12 modern hearing aids (2008)
Gain for Conductive Losses – How do different hearing aid manufacturers compensate for conductive losses in their prescribed gain? (2009)
Gain and Acclimatization – Measured long-term average gain for 12 modern hearing aids (2008)
Children and Hearing Aid Gain – A study at Alviksskolan, Stockholm (2008)
Here summaries of finished projects are found.
Aim
The primary aim of this project was to learn more about hearing thresholds at high frequencies.
Procedure
10 listeners without hearing impairment and 34 persons with hearing impairment at the traditionally measured audiometric frequencies from 125 Hz to 8 kHz participated in the study. Hearing thresholds were determined in the frequency range 125 Hz to 20 kHz using a clinical audiometer and high-frequency headphones.
Main findings
Noise problems related to the narrow range of presentation levels for the highest frequencies were found. It was difficult to predict the hearing thresholds at the highest frequencies based on the hearing thresholds at the traditionally measured frequencies.
Presentations/Publications
The results have been reported at Audionomdagarna (National Swedish conference for Audiologists) in Västerås, Sweden, May 2009.
Back to top
Aim
The aim of this project was to study the predictive ability of computational measures for evaluating the performance of noise reduction algorithms with regard to speech intelligibility, noise loudness and sound quality.
Procedure
20 listeners with hearing impairment participated in a laboratory test where they performed an adaptive speech test in noise and rated three quality aspects (Preference, Noise Loudness, and Speech Clarity) with a modified paired comparison test, the paired comparison rating (PCR). Three NR algorithms were included in the project.
Predictions from a number of computational measures were compared to the results of the two listening tests. The results from the speech test were compared to the calculated Speech Intelligibility Index (SII) and an extended version of the SII (ESII), which takes into account fluctuations in the masking noise and the effects of forward masking. The results from the PCR test of Noise Loudness were compared to the physical improvement in SNR the various noise reduction algorithms could provide, and with the calculated noise loudness. The PCR results from the Preference and Speech Clarity measures were compared to a distortion measure, the perceptual evaluation of speech quality (PESQ), and a composite measure.
Main Findings
It was shown that the calculated noise loudness predicted the results of the PCR Noise Loudness well. Both the predictive measures for speech recognition overestimated the performance for noise reduction algorithms that managed to increase the physical SNR but at the cost of introducing high amounts of distortion (described as musical or residual noise). The tested distortion measure could not capture this distortion, and the results from the sound quality related measures were difficult to interpret, but seemed to rank order the three noise reduction algorithms correctly.
Presentations/Publications
The work is described in two theses:
Anders Nilsson’s MSc thesis.
Florian Wolter’s BSc thesis.
The work has also been presented at the following conferences:
Smeds K, Wolters F, Nilsson A, Båsjö S, Hertzman S, and Leijon, A. 2010. Objective measures to quantify the perceptual effects of noise reduction in hearing aids The AES 38th International Conference: Sound
Quality Evaluation. Piteå, Sweden: Audio Engineering Society, Inc., pp. 101-107.
Audionomdagarna (national Swedish conference for audiologists) in Västerås, Sweden, May 2009.
STAF (Svensk Teknisk Audiologisk Förening, national Swedish conference for engineers and technicians) meeting in Eskilstuna, Sweden in March 2010.
Back to top
Aim
There is little consensus regarding how much hearing-aid gain is appropriate, in particular for sounds at low input levels. An important hearing aid parameter for determining gain for soft sounds is the compression threshold (CT). Previous research using fast-acting single-channel HAs indicates that most hearing-aid users prefer a moderate-level CT rather than a low-level CT. It is not clear if these results can be applied to HAs with other time constants. In addition, emerging research indicates that new and experienced HA users prefer different amounts of gain due to the HA acclimatization effect. The purpose of the current study is to investigate if the subjectively preferred CT is influenced by either compression speed and/or the participants’ previous HA experience.
Procedure
The participants were 10 new and 10 experienced HA users with mild-moderate sensorineural hearing losses. In a double-blind field trial, the participants compared moderate and low CTs implemented in a 2-program, 15-channel HA with a 2:1 compression ratio. The trial was divided into two 2-week trial periods. In each trial period, the two CT programs were combined with either fast- or slow-acting compression in a crossover, counter-balanced design.
Main findings
The results showed the inexperienced HA users were less likely to prefer low CT compared to the experienced HA users. The overall CT preference was not influenced by compression speed. In specific listening situations, the participants most often had no preference for either a moderate or low CT. If they had a preference in specific listening situations, they most often preferred the moderate CT, except for situations with quiet speech when combined with slow-acting compression. Overall, the results were not strongly in favour of the use of either a low or moderate CT. The findings also provide preliminary evidence that CT preference is related to the HA gain acclimatization effect.
Presentations/Publications.
The work is summarized in a PhD thesis:
Connor Sørensen H. 2010. Hearing aid amplification at soft input levels Centre for Applied Hearing Research (CAHR). Copenhagen: Technical University of Denmark, p. 145.
The work has been presented at the following conferences:
Connor Sørensen H., Båsjö S. & Smeds K. 2010. Field trial investigation of the preferred hearing-aid compression threshold International Conference on Adult Hearing Screening. Cernobbio (Como Lake), Italy.
Back to top
Aim
The aim of this project was to explore ways to visualize the effect of noise reduction in hearing aids and to get a rough picture of the differences between various noise reduction algorithms implemented in a number of hearing aids on the market 2008.
Procedure
Long-term average gain measurements were performed in a dedicated hearing aid test box. Gain-curves from the hearing aids were first measured by setting most of the features to OFF and then another set of corresponding gain-curves with the noise reduction set as default suggested by the different fitting software (ON). The gain reduction accounted for by the noise reduction algorithm was then calculated from the two corresponding gain-curves (OFF – ON). The hearing aids were programmed with three different audiometric hearing losses. To study the dependence on the Signal to Noise Ratio (SNR) of the input signal, the hearing aids were presented with various signal mixes with different SNRs. The input signal levels were set to have a fixed speech level at 62 and 75 dB SPL. As speech signal the new ISTS signal produced by the ISMADHA-group was used. The ISTS is a real speech signal mix of different female speakers reading the same text in six different languages. The speech-signal used here is a draft submitted to both IEC and ANSI for standardization. The noise signal was IRCA1, an un-modulated speech-weighted noise signal.
Main findings
The main conclusions from the project were that contour plots, where the gain reduction as a function of frequency and SNR is plotted, is a feasible way to illustrate the effect of noise reduction, and that there are large variations in how various hearing aid manufacturers have implemented their noise reduction systems.
Presentations/Publications
The results have been presented at the following conferences/meetings:
EU09 ”Etterutdanningskurs för audiopedagoger, audiografer och ingeniører” (National conference for Audiologists and Engineers), Oslo, Norway, November 2009.
Power Point presentation.
Widex Nordisk Seminar, Stockholm, Sweden, October 2009.
Presentation for download.
ISAAR (International Symposium on Auditory and Audiological Research), Helsingør, Denmark, August 2009. Poster presentation. A paper will appear in the proceedings from the conference (not available in June 2010).
EFAS (European Federation of Audiological Societies) conference in Terife, Spain, June 2009.
ICRA (International Collegium of Rehabilitative Audiology) meeting in Germany, May 2009.
TEMA Hörsel (National conference for Audiologists, Engineers and ENT specialists), Jönköping, Sweden, May 2009.
STAF (Svensk Teknisk Audiologisk Förening) meeting in Hudiksvall, Sweden in April 2008.
Abstract in Swedish Power Point presentation in Swedish
Back to top
Aims
For this study the aims were twofold:
- to investigate how the manufacturers’ implemented NAL-NL1 gain compared to NAL-NL1’s own compensation for conductive losses.
- to study the gain compensation for conductive losses in hearing aids from ten different hearing aid manufactures and
Procedure
The hearing aids were chosen in order to represent a variety of high-end devices from ten different hearing aid companies. The devices were programmed with four different audiograms representing a flat and a sloping hearing loss of both sensorineural and conductive origin. The hearing aids’ noise reduction algorithms were switched on. 711-coupler measurements of hearing aid gain and OSPL90 were performed in an acoustic test chamber. The signals used were the ISTS speech signal in ICRA1 background noise for the gain measurements and pure tones for the OSPL90 measurements. The data were compared to NAL-NL1 prescribed gain for the four audiograms.
Main findings
The study provides two main conclusions.
1. Different hearing aid companies provide very different gain for the same audiogram configuration, even when the prescription is called “NAL-NL1”. Despite the use of NAL-NL1 as prescription method, the variation between the device providing most gain and the device providing least gain was about 20 dB for all measured input levels. Hearing aid manufactures also provide a more linear gain setting for the sensorineural hearing losses than expected.
2. The strategies for compensating for conductive hearing losses are not very clear. One observation, though, is that in those cases where a choice between an own prescription method and the NAL-NAL method is offered, the NAL-NL1 compensation is always somewhat bigger than for the company-specific compensation. The “conductive compensation curves” for the NAL-NL1 showed to be very dissimilar to the “conductive compensation” that was advocated for the NAL-RP. The Maximum Power Output measurements (OSPL 90) showed that there is generally a difference of about 10 dB between OSPL 90 for sensorineural and conductive prescriptions. This is about 20 dB less than what is prescribed by the NAL-NL1 formula.
Presentations/Publications
TEMA Hörsel (National conference for Audiologists, Engineers and ENT specialists), Jönköping, Sweden, May 2009.
Back to top
Aims
The aims of the study were
1. to document differences in prescribed gain for a selection of hearing aids implementing the NAL-NL1 prescription.
2. to document how acclimatization stages are used to adjust the prescribed gain.
Procedure
The hearing aids were set to two different audiogram configurations and measured in an acoustic test chamber using a 2cc-coupler. Two different ICRA noises were presented at three different input levels: ICRA4 at 50 and 65 dB SPL and ICRA8 at 85 dB SPL. All hearing aids were tested in two software conditions, the lowest and highest degree of user experience.
Main findings
The study provides two main conclusions.
1. The measured gain for hearing aids programmed with the same generic prescription, NAL-NL1, differed with over 20 dB for certain input levels. This means that some implementations of the NAL-NL1 are unacceptable, and it is impossible for an audiologist to know what a “NAL-NL1 fitting” represents.
2. Hearing aid manufacturers have chosen to implement gain acclimatization stages in very different ways. Some manufacturers do not employ gain acclimatization at all, whereas others reduce the gain for a first-time hearing aid wearer up to 10 dB. Most manufacturers who have implemented acclimatization stages reduce the gain in a uniform way across input level, but three manufacturers reduces the gain more for low-level sounds than for high-level sounds.
Presentations/Publications
TEMA Hörsel (National conference for Audiologists, Engineers and ENT specialists), Jönköping, Sweden, May 2009.
Back to top
Aim
The project was undertaken to document the hearing aid gain used by school-aged hearing-impaired children.
Procedure
In total 35 children with sensorineural hearing loss (19 boys and 16 girls, median age 9.9 years, ranging from 7.3 to 14.6 years) participated in the study. The basic goal was to measure the gain the children normally use. During a short excerpt from a TV program, containing a dialogue between a female and a male talker at a presentation level of 65 dB SPL, the participants selected hearing aid program and changed the volume control so that the speech was as clear as possible and with the appropriate loudness. Using an Aurical test equipment, real ear insertion gain measurements were performed using modulated speech-weighted noise at 65 dB SPL. These measured insertion gain results were compared to individually calculated NAL-RP gain targets. Coupler gain measurements were performed using modulated speech-weighted noise at 50, 65, and 80 dB SPL and output measurements were performed for pure tones at 90 dB SPL. All gain measurements started after a 15 second pre-conditioning period.
Main findings
In the frequency range 250-2000 Hz the median gain used resembled that of the NAL-RP prescription (apart from the fact that the “NAL-boost” at 1 kHz was not mirrored). At higher frequencies, the median gain used was less than that of the NAL-RP prescription. The coupler gain measurements were used to calculate two static “compression ratios” at each frequency. The calculated compression ratios were fairly low, ranging from 1.1 to 1.7. Test-retest data on the program choice and VC setting showed that most participants produced hearing aid settings that were highly reproducible.
Back to top
|