Improving Sound Quality with High-Resolution Hearing Instruments

Sound quality has been revealed to be the second-most important factor related to customer satisfaction improvement, as well as the second-most desired improvement sought by hearing instrument wearers.¹ Clearly, advances in this area would contribute greatly to wearer satisfaction with hearing instruments.

But what exactly is meant by sound quality? One aspect of sound quality is that which allows the ear to distinguish between sounds having the same pitch and loudness, such as the difference between the same musical note played on a piano versus a guitar. Also known as timbre, this is mainly determined by the harmonic content and dynamic features of the sound, like vibrato and attack and decay times.

When speaking of sound reproduction, sound quality refers to the fidelity with which the original sound is duplicated by the sound-generating device. Although it is an indication of how accurately the system gives the impression of the original sound, it is also subjective. For example, listeners who were asked to rate various dimensions of sound quality of music played through different speaker systems appeared to weigh these dimensions differently when asked to judge the similarity of the speaker systems.² Some listeners gave greater weight to “brightness” while others weighed “loudness” more when comparing reproductions for similarity.

As pointed out in a previous article in The Hearing Review,³ sound quality is more difficult to define for the hearing instrument wearer, because hearing instrument processing strategies significantly change the incoming sounds as they aim to compensate for lost auditory function. For example, the prescriptive fitting formulae we typically use to adjust the frequency response of hearing instruments will always result in a weighing of gain in certain frequency regions, thereby negating the idea of an acoustic signal which is a true reproduction of the original.

For hearing instrument wearers, it may be more appropriate to think of sound quality as describing how amplified sounds fit within their range of hearing, whether sounds are distorted, and the degree to which undesired sounds (eg, background noises, acoustic feedback, or signal processing artifacts) are heard. This is consistent with the way in which various dimensions of hearing instrument satisfaction and desired improvements in the MarkeTrak surveys are grouped as relating to sound quality, including not only “better sound quality,” but also “less whistling and buzzing,” “more soft sounds audible,” and “loud sounds less painful.” In addition, “clearness of tone and sound,” “natural sounding” and “sound of voice” were suggested by MarkeTrak¹ to be dimensions of sound quality for hearing instrument wearers.

No one feature or characteristic of a hearing instrument can ensure satisfactory sound quality. Mechanical and electroacoustic designs are critical components of how the final product performs and sounds. Sampling rate, A/D conversion, and processing speed are some factors which can affect the sound output of a digital hearing instrument.

It is intuitively appealing to think of resolution in the sound processing as analogous to the realm of digital photography, where a higher number of pixels enhances the quality of the photograph. The following discusses how sound processing algorithms and some acoustic aspects of the fitting impact sound quality for wearers of the ReSound Pixel hearing instrument. In developing this device, the concept of resolution was key. The goal was to provide processing that could resolve frequencies similar to the human auditory system, and to offer sufficient resolution in the temporal and intensity domains that maximize sound quality.

“Better Sound Quality”
The foundation of the sound processing in the Pixel hearing instrument is the 17-band Warp compression system. The Warp compressor utilizes a technique called frequency warping to achieve frequency representation similar to the human ear. The peripheral auditory system can be thought of as a bank of bandpass filters—now referred to as auditory filters—which perform frequency analysis on the incoming sound.

Beginning with the work of Fletcher,⁴ many investigators have developed models of the auditory filters based on psychoacoustic experiments^5,6 and this concept has become widely accepted. It has been found that the auditory filters vary in bandwidth depending on the frequency, with frequency resolution greatest at lower frequencies (narrower bandwidth) and decreasing with increasing frequency (broader bandwidth).

Most digital signal processing techniques for frequency analysis yield constant bandwidth with uniform spacing of the bands, which is unlike the non-uniform spacing of the auditory filters. This difference in spacing of the bands is illustrated in the top two panels of Figure 1. Because these graphs are plotted on a linear scale, the finer low frequency resolution of the auditory system is also readily apparent.

FIGURE 1. In the human ear, the bandwidth of the auditory filters increases with increasing frequency (top panel). Most digital techniques for frequency analysis, such as the discrete Fourier transform (DFT), result in uniform spacing of the bands (middle panel). In contrast, frequency warping can provide non-uniform frequency analysis similar to the auditory filters (bottom panel).

Frequency warping is a frequency analysis technique which modifies the uniform band spacing inherent to other digital techniques to a non-uniform scale. The parameters utilized for frequency warping in the Warp compression system result in band spacing that is very similar to that of the auditory filters, as illustrated in the lower panel of Figure 1. The mapping of frequencies by the Warp compressor is, in fact, so good a representation of the auditory system that the center frequencies of its bands are nearly identical to the first 17 bands of the auditory model. In terms of performance metrics, advantages of employing frequency warping include lower processing delay than for conventionally implemented systems with comparable frequency resolution and very low distortion.

“More Soft Sounds Audible” and “Loud Sounds Less Painful”
The Cochlea Dynamics compression strategy implemented in most GN ReSound hearing aids encompasses a set of compression characteristics that is intended to compensate for the loss of the ear’s natural compressive non-linearity caused by damage to outer hair cells in the cochlea. The compression characteristics include fast syllabic attack and release times, low-compression thresholds, and compression ratios ranging from 1:1 to 3:1. This type of compression scheme squeezes the normal range of sound levels into the reduced dynamic range of the hearing-impaired individual. In this way, audibility of soft sounds is attained, while overamplification of loud sounds is avoided.

Addressing Background Noise
Background noise was one of the top reasons why people who own but do not use their hearing instruments (eg, “in-the-drawer” aids) reported giving up on them.⁷ In the words of one respondent to this survey, “I will not wear my hearing aids because they increase background noise. After awhile I get a headache and get somewhat nervous.” This statement reflects the annoyance and fatigue many hearing instrument wearers experience when bombarded with amplification of undesired sound. Noise reduction strategies have the goal of reducing this annoyance, thereby enhancing the quality of the listening experience. User preferences when listening with noise reduction processing in laboratory tests suggest that the algorithms are helpful in this regard.^8,9

The noise reduction system employed by the ReSound Pixel uses a spectral subtraction approach, which entails reducing gain for noise components of the incoming sound compared to gain for speech components. Crucial elements in the success of this processing include the frequency resolution of the system, its accuracy in identifying speech and noise, its accuracy in estimating the spectrum of the noise, and its reaction times.

Frequency warping technology plays a vital role in the ReSound Pixel noise reduction system. The Warp compressor with its auditory system resolution forms the basis for the noise reduction’s precise capacity to attenuate noise while leaving the speech signal untouched. The noise reduction operates in the same 17 warped frequency bands as the Warp compressor. Identification of speech and noise components of the input sound is based on extraction of temporal and spectral features, and noise estimation is limited to time periods in which speech components are not identified. This makes possible reduction of noise elements without disrupting or degrading speech (Figure 2).

FIGURE 2. Noise reduction based on spectral subtraction estimates the spectrum of the noise during pauses in speech, and subtracts the noise estimate from the total signal.

“Less Whistling and Buzzing”
Acoustic feedback compromises hearing aid performance in several ways. First, feedback leading to oscillation causes “whistling.” The hearing instrument amplifier saturates, which leads to an increase in distortion. Ultimately, feedback limits the maximum amount of high frequency gain. Acoustic feedback has additional negative consequences for wearers. Depending on the hearing loss, the whistling may be audible to them. Even if it is not, feedback can be a source of embarrassment for the wearer because it is bothersome or disruptive to others.

The ReSound Pixel utilizes Dual Stabilizer DFS feedback cancellation to increase the amount of available gain for a particular fitting. DFS technology analyzes the properties of the amplified sound returning from the ear canal to the hearing instrument microphones and creates a signal equal and opposite in phase to cancel this feedback. DFS employs two cancellation filters: 1) a static filter, which accounts for relatively stable aspects of the feedback path, and 2) an adaptive filter, which changes its characteristics during use of the hearing instrument to account for dynamic aspects of the feedback path. The coefficients of the static filter are determined by a calibration carried out at the fitting of the hearing instrument. This calibration procedure is unique to DFS technology.

There are a number of digital hearing instruments available that can provide additional headroom through adaptive feedback cancellation. However, an essential element for sound quality of the DFS processing is a set of constraints placed on the adaptive cancellation filter. These constraints drastically reduce the chances of the system attempting to cancel sounds that are not feedback. The sound of doorbells, phones ringing, or even the tapping of fingers on a computer keyboard have the potential to cause ringing or buzzing artifacts when adaptive feedback cancellation is not constrained. Figure 3 compares the loss in fidelity for five digital hearing instruments which all provide extra headroom via feedback cancellation. This measurement was generated by subtracting the amount of distortion for a signal in the ear canal with feedback cancellation activated from the distortion without the feedback cancellation. Hearing Instrument ¹, which is the ReSound Pixel, showed no loss of fidelity with the Dual Stabilizer DFS active in this measurement. This indicates that the wearer benefits from the extra headroom provided by the system without having to endure a decline in sound quality.

FIGURE 3. Loss of fidelity with feedback cancellation turned on was determined for five digital hearing instruments. Feedback cancellation in the ReSound Pixel (Hearing Instrument 1) did not add distortion.

Another characteristic of the DFS processing in the ReSound Pixel which is necessary for good sound quality is that it is dual channel. This hearing instrument features dual microphone Enhanced Adaptive Directionality, which can achieve optimum polar attenuation patterns for varying frequencies simultaneously. Because of the variability in changes over frequency and phase caused by such an advanced adaptive directional system, the feedback path to each microphone cannot be adequately modeled by a single feedback cancellation system. A single-channel feedback cancellation system would result in lack of headroom and degraded sound quality when in directional mode. To solve this issue, Dual Stabilizer DFS assigns a dedicated feedback cancellation filter to each microphone.

“Natural Sounding” and “Sound of Voice”
Sound quality complaints in terms of unnaturalness of one’s own voice are often related to blocking of the ear canal by an earmold or hearing aid shell, creating the so-called “occlusion effect.” The occlusion effect is the build-up of low frequency sound pressure in the residual ear canal that occurs when body-conducted sound is trapped.^10,11 Hearing aid users often describe the subjective effect of occlusion as sounding like they are talking in a barrel. In addition, sounds like chewing, swallowing, throat-clearing, and even breathing can be perceived as being so loud that they interfere with the normal perception of external sounds. Alleviation of the occlusion effect and improvement in sound quality for own voice can be achieved through open venting strategies.¹²

Sound quality for ReSound Pixel wearers with good low frequency hearing can be further enhanced with open fittings. An open fitting can be achieved with either large diameter vents, or with very short vents such as the FlexVent earmold.¹³ The BTE model can also be fitted with an inconspicuous thin tube and non-occluding eartips. Because this thin tube has less than half the inner diameter of standard BTE tubing, there is an acoustic loss of high frequency amplification (Figure 4). Compensation for this loss is done automatically when the fitter clicks a button in the fitting software.

FIGURE 4. Open fittings (yellow) using thin cosmetic tubing require compensation for the reduction in high frequency gain compared to standard BTE tubing (blue).

Open fittings place unique demands on the hearing instrument, and the sound quality benefits won by the open fitting can easily be jeopardized if these demands are not met. One of these demands is a low propagation delay. All digital hearing aids introduce some delay to the passage of sound through the instrument due to A/D conversion, digital signal processing speed, and number and complexity of signal processing algorithms. Long propagation delays can lead to disturbing changes in the timbre of the sound, as sound amplified through the hearing instrument mixes with direct unamplified sound to the ear canal. The perceptual effects of this delay are particularly apparent for the wearer’s own voice. The propagation delay associated with the Resound Pixel is less than 5 ms across frequencies, which has been demonstrated to result in high ratings for naturalness of own voice and other sounds.^14,15

An additional requirement to an open-fitted hearing instrument is that it be capable of providing enough stable gain to allow fitting of a wide range of hearing losses. This entails an effective strategy for dealing with acoustic feedback. As reviewed above, the Dual Stabilizer DFS feedback cancellation system allows more useable gain for each individual fitting without loss of fidelity. This means that the sound quality benefits attained via the open fitting are not counteracted by disturbing ringing artifacts from the feedback cancellation.

This article was submitted to HR by Jennifer Groth, MA, the senior research audiologist at GN ReSound Group North America, Chicago, Ill. Correspondence can be addressed to HR or Jennifer Groth, GN ReSound Group North America, 2601 Patriot Blvd, Glenview, IL 60026; e-mail: [email protected].

References
1. Kochkin S. Consumers rate improvements sought in hearing instruments. The Hearing Review. 2002;9(11):18-22.
2. Gabrielsson A, Rosenberg U, Sjögren H. Judgments and dimension analyses of perceived sound quality of sound-reproducing systems. J Acoust Soc Am. 1974;55,854-861.
3. Groth J. A new high performance, fully transparent hearing instrument. The Hearing Review. 2005;5(12):56,58-59.
4. Fletcher H. Auditory patterns. Rev Mod Phys. 1940;12:47-65.
5. Moore BCJ, Glasberg BR. Suggested formulae for calculating auditory-filter bandwidths and excitation patterns. J Acoust Soc Am. 1983;74:750-753.
6. Zwicker E, Terhardt E. Analytical expressions for critical-band rate and critical bandwidth as a function of frequency. J Acoust Soc Am. 1980;68:1523-1525.
7. Kochkin S. MarkeTrak V: “Why my hearing aids are in the drawer”: The consumers’ perspective. Hear Jour. 2000;53(2):34-41.
8. Ricketts T, Hornsby BWY. Sound quality measures for speech in noise through a commercial hearing aid implementing “Digital Noise Reduction”. J Am Acad Audiol. 2005;16:270-277.
9. Boymans M, Dreschler WA, Shoneveld P, Vershuure H. Field trials using a digital hearing aid with active noise reduction and dual-microphone directionality. Audiology. 1999;38:99-108.
10. Studebaker GA, Zachman TA. Investigation of the acoustics of earmold vents. J Acoust Soc Am.1979;47:1107-1115.
11. Grover BC, Martin MC. Physical and subjective correlates of earmould occlusion. Audiology. 1979;18:335-50.
12. Kiessling J, Brenner B, Jespersen CT, Groth J, Jensen OD. Occlusion effect of earmolds with different venting systems. J Am Acad Audiol. 2005;16:237-249.
13. Jespersen C, Groth J. Vent is designed to reduce occlusion effect. Hearing Jour. 2004;57(10):44-46.
14. Groth J, Søndergaard M. Disturbance of varying propagation delay in non-occluding hearing aid fittings. Int J of Audiol. 2004;43:594-599.
15. Kates J, Arehart K. Multi-channel dynamic range compression using digital frequency warping. EURASIP Journal of Applied Signal Processing. 2005;18:3015-3025.