Circuitry Options: Signal Processing Under Altered Signal ...

Chapter Four

Circuitry Options : Signal Processing Under Altered Signal

Conditions

by Paul H. Stypulkowski, PhD

Paul H. Stypulkowski, PhD, was Senior Technical Specialist with the 3M Health Hearing Laboratory in St. Paul, Minnesota at the timethis segment was written . He is currently Marketing and Technical Support Manager for SONAR Hearing Health in Eagan, MN

INTRODUCTION

Signal processing, as it applies to hearing aids, is aterm that has many connotations . In the broadest sense,any change made to an input signal by a circuit can beconsidered signal processing . Thus, even the simplesthearing aid performs some degree of processing by am-plifying and frequency-shaping an input signal . Gener-ally however, hearing aid signal processing usually refersto somewhat more complex signal modification intendedto enhance, in a defined, predictable manner, the outputsignal relative to the input signal.

Signal processing can be beneficial or it can bedetrimental, depending upon its intended function andthe resulting outcome. Peak clipping is an example of asimple form of signal processing . A peak-clipping circuitclearly modifies the input signal in a well-defined, pre-dictable manner by limiting the amplitude of signals thatexceed a certain level and thereby maintaining the outputat a specified level . This processing can be viewed asbeneficial if looked at from the design goal of preventingthe hearing aid output from reaching sound pressure lev-els that exceed a user's comfort level (UCL) . On the

Address all correspondence to: Paul H . Stypulkowski, PhD, Director ofMarketing, Sonar Hearing Health, 3130 Lexington Avenue South, Eagan, MN55121 .

other hand, peak clipping can be considered detrimentalin view of its effect upon the frequency content andsound quality of the output signal. Peak clipping resultsin severe distortion and significantly degrades soundquality . This example illustrates that signal processingcan be viewed in different ways, and that in and of itself,signal processing is neither inherently good nor bad, butmust be examined in the context of its intended applica-tion, how well it achieves the desired goal, and the conse-quences of the processing on various aspects of theoutput signal.

GOALS OF SIGNAL PROCESSING

Hearing aid signal processing can generally be cate-gorized into two very broad areas of intended applica-tion. The first includes those circuits designed to alter theinput signal in some manner to better fit the hearing im-pairment of the user. In other words, their intent is tocompensate for the hearing loss. Linear amplificationwith a prescriptive frequency response would represent avery simple example in this category. A more complexexample is a compression circuit designed to transformthe wide range of sounds in the environment into a nar-rower, restricted range of output levels to fit within a re-duced dynamic range.

49

50

RRDS Practical Hearing Aid Selection and Fitting

The second category includes circuits designed toenhance the signal in some manner, based upon the prop-erties of the input signal itself. In this case, the processingis intended to somehow improve the signal, thereby mak-ing it more intelligible or more pleasant . Almost allforms of noise reduction fall into this category, sincetheir primary design goal is to modify the input signal,based upon its frequency, intensity, or temporal charac-teristics, in such a way that the output contains less noiseand more signal . Although some processing in this cate-gory may yield an improved fit to the hearing impair-ment, this is typically a secondary effect. An example ofthis would be traditional ASP (automatic signal process-ing) hearing aids that use an adaptive filter to reduce lowfrequency output when input levels exceed a definedthreshold . The original design intent of such circuits wasto reduce the amplification of background noise domi-nated by low frequency, and thereby reduce the inci-dence of the upward spread of masking (i .e ., the loss ofaudibility of high frequency sounds caused by louderlow frequency sounds).

Within both application categories exists a range ofsignal processing complexity, from the simple exampleslisted, to very complex implementations under develop-ment, such as systems using multiple microphones, anddigital signal processors incorporating sophisticatednoise reduction algorithms.

CHARACTERIZING SIGNAL PROCESSING

Signal processing circuits are usually characterizedin terms of their effects on three primary signal domains:the frequency domain, the intensity domain, and the timedomain. Although each is identified individually, thethree represent inseparable attributes of any analog sig-nal, such as speech. For example, speech can be charac-terized in the time-intensity domain by a waveform likethat shown in Figure 1, where the top graph represents aplot of sound intensity versus time. The various individ-ual speech components in this utterance of the word"shoot" are indicated below the corresponding section ofthe waveform. In this example, the vowel portion ("oo")has a greater intensity than the consonants ("sh" and"t") . In the frequency domain, the speech signal can bedescribed by the relationship between frequency and in-tensity of the long-term average of speech, known as thespeech spectrum (Figure 1, bottom) . As in the shortspeech sample, the long-term speech spectrum has the

r60 —

100

1000

10000

L0 Octave Band Center Frequencies (Hz)

Figure 1.Speech characteristics . Top : time-intensity waveform of the word"shoot." Bottom : typical speech spectrum.

highest intensity in the low frequency vowel region andless energy in the high frequency consonant region . Toassess the effects of hearing aid signal processing onspeech or any other signal, the input and output signalscan be compared in these different domains . The differ-ences between the output and the input are the result ofthe signal processing.

CLASSIFYING SIGNAL PROCESSING

Recently, several authors have summarized andcreated a hierarchy of the different types of signal pro-cessing currently available, particularly in programma-ble hearing aids (1,2) . For the most part, thesecategorizations are in general agreement . The classifica-tion scheme presented below will ignore for the momentthe aspect of programmability and the potential for mul-tiple-response capabilities, which represents a furtherpoint of differentiation, and focus specifically on analogsignal processing.

Number of ChannelsA main point of differentiation between circuits is

the number of channels . Instruments with multiple chan-

so

to 40

30

20

51

Chapter Four : Circuitry Options-Altered Signals

nels contain filters that separate the input signal into fre-quency bands that can then be processed independently.These instruments generally offer the ability to more pre-cisely shape the response to match the hearing loss.

Number of Channels of CompressionGenerally viewed as more important than the num-

ber of channels is the number of channels of compressionwithin an instrument . Single- and multiple-channel hear-ing instruments may contain zero, one, or several com-pression circuits . For either instrument, linearamplification combined with peak-clipping circuits tolimit output represents the simplest form of signal pro-cessing.

Single-channel compression instruments can gener-ally be classified into one of three categories based uponthe frequency region that the compression circuit con-trols (Figure 2) . Compression systems that operate pri-marily in the low frequency region produce bass increaseat low levels (BILL) processing (3) . These systems re-duce low frequency gain in response to higher input lev-els in that region (note: some BILL devices do not usecompression circuits but accomplish a similar effectthrough the use of active filters that change the responseslope in the low frequencies) . High frequency compres-sion systems provide treble increase at low levels(TILL), reducing high frequency gain in response tolouder inputs and increase high frequency gain as inputlevels decrease . Single-channel compression systemsthat control overall gain are generally known as auto-matic gain control (AGC) circuits ; these vary the overallgain across the full frequency range of the instrument,much like a volume control, rather than having a fre-quency-specific effect.

Multichannel instruments may or may not containcompression circuits ; some use simple linear amplifiersand peak-clipping circuits . Those with compression mayhave: (a) one compression circuit that controls the over-all gain of the instrument; (b) a compression circuit inone channel and peak clipping in others ; (c) individualcompression circuits within each channel ; or (d) a com-bination of the above. In the signal processing hierarchy,instruments that utilize independent compression cir-cuits within each channel are generally considered themost advanced . This design, known as multichannelcompression, allows the gain of each channel to be con-trolled by inputs within each channel's bandwidth. Mul-tichannel compression instruments provide the mostflexibility, and can adapt their signal processing in re-

BILL

Frequency (Hz)

TILL

AGC

Figure 2.Single-channel compression processing . Top : BILL (bass increase atlow levels) ; middle: TILL (treble increase at low levels) ; bottom: AGC(automatic gain control). Each panel illustrates the characteristicchange in frequency response for each circuit as the input level variesfrom soft (dotted) to loud (solid).

sponse to changes in input level across the frequencyspectrum. Unlike single-channel systems, limited to onetype of processing, some multichannel compression in-

Frequency (Hz)

J

52


struments can produce BILL, TILL, or overall AGC ef-fects in response to different types of input signals (lowfrequency, high frequency, or broadband), providingvariations of the three different response patterns illus-trated in Figure 2, depending upon the intensity and fre-

quency content of the input signal.

TYPE OF COMPRESSION CIRCUIT

Compression LimitingThe compression circuit within an instrument also

represents a point of differentiation . Circuits with highcompression ratios (5 :1 or greater) and high thresholds ofcompression (65 dB SPL or greater) are known as com-pression limiters, and are designed to prevent output fromexceeding a predetermined level to avoid circuit overloadand user discomfort (Figure 3) . Compression limiters per-foiut essentially the same function as peak clippers : theylimit the hearing aid output at a specified level, but gener-ally do so with significantly lower distortion levels.

Wide Range CompressionOther circuits designed to compensate for the re-

duced dynamic range associated with hearing impairmenttypically use lower compression ratios and thresholds(Figure 3) . These circuits operate over a wider range ofinput levels and preserve relative intensity information of

Figure 3.Compression input-output characteristics . Representative illustrationsof input-output curves for different compression circuits .

speech and other inputs, compared to compression limit-ing systems that severely degrade intensity informationabove the compression threshold . Some instruments alsocombine different compression circuits within the samehearing aid . Typically, these consist of a combination ofone or more wide-range compression circuits paired witha compression limiting circuit that activates at a muchhigher level and functions to limit output.

Variable Release TimesSome circuits also employ a variable release time

feature. Associated primarily with single-channel com-pression systems, variable release time circuits adapt thecompressor release time, based upon the duration of theinput that triggers the compression circuit: longer releasetimes are used for longer duration inputs and shorter re-lease times are used for more transient inputs . This sys-tem is typically not required with multichannelcompression systems where the release time for eachchannel is set for the appropriate duration of signalswithin the channel's bandwidth.

Classifying Hearing Aid Signal Processing

Number of channelsSingle channelMultiple channel

Number of channels of compression0 — peak clipping1 — Single Channel: BILL, TILL, AGC, variable release time2 or more — Multichannel compression

`lope of compression circuitInput/OutputCompression limitingWide range compression

Input/Output CompressionFor hearing aids with a volume control, the com-

pression system can also be categorized as either inputor output compression . Very simply, with an input com-pression system, the volume control affects both the gainand the maximum output of the hearing aid, providingthe user with control over the level of both soft and loudsounds . With an output compression circuit, the maxi-mum output of the aid is fixed and independent of thevolume control, which affects the gain and the compres-sion threshold. Because the user does not have control ofmaximum output of the aid with this system, an appro-priate output setting is critical to successful use . Typi-cally, output compression systems tend to be associated

CompressionLimiting

I'

'1'1'1'1'120

30

40

50

60

70

80

90

Input (dB)

53


with compression limiting circuits . Wide range compres-sion circuits are more often configured as input com-pression systems.

CHARACTERIZING ALTERED SIGNALCONDITIONS

Which one of the various types currently available inhearing aids is the most appropriate under "altered" signalconditions? Assuming that an appropriate fitting exists for"average" conditions (conversational speech in quiet), oneapproach to answer this question is to determine how thealtered conditions differ from the average condition, andattempt to create signal processing to compensate for thedifferences . For some conditions, a simple change in fre-quency response of a linear instrument may be sufficient;for others, the use of more complex multichannel com-pression may be required to best optimize performance.

Fittings designed for altered listening conditions, orspecific listening environments, are generally associatedwith the use of multiple-response instruments (i .e ., multi-ple-memory programmables) . With single-response instru-ments, it is unlikely that the fitting will be optimized forthe characteristics of one specific listening situation, butmore likely to be fit for average conditions . In attemptingto determine the most appropriate signal processing forspecific conditions there are a number of characteristics oflistening situations that should be defined including : 1) thesignal of interest ; 2) the background noise conditions ; and3) the acoustic nature of the listening environment.

Signal Of InterestThe primary consideration in creating a fitting for

any specific listening situation is the signal of interestand how it differs from the conversational speech signalfor which the baseline fitting was optimized . The alteredsignal of interest can then be compared in the differentdomains outlined earlier to determine how it differs fromthe average speech signal . For example, are the differ-ences primarily in the frequency content or in the inten-sity domain? By defining these basic differences, theappropriate signal processing modifications may becomemore evident.

Background NoiseSimilarly, the level and composition of any back-

ground noise in the altered conditions that will compete

with the signal of interest need to be defined . Here again,the characteristics of the background noise : its frequencycontent (low frequency versus broadband) ; its intensityrange, and its temporal nature (constant versus intermit-tent) will all factor into the signal processing modifica-tions that may be appropriate.

Listening EnvironmentFinally, the acoustic nature and specific conditions

of the listening environment also require consideration.For example, situations with multiple speakers located atvarious distances from the listener, as in a large meeting,can be problematic for many users due to large differ-ences in the vocal levels of the various speakers . Is thesignal of interest presented live or via a sound reproduc-tion system, the characteristics of which may alter thesignal? Are the surroundings reverberant? These factorswill also influence the signal processing required to opti-mize performance.

Characterizing "Altered" Listening Conditions

Signal of InterestFrequency compositionIntensity levels

Background NoiseFrequency compositionIntensity levelsTemporal characteristics

Listening EnvironmentSignal source(s)Acoustics (reverberance)

ALTERED SPEECH SIGNALS

Everyday speech covers a wide range of frequenciesand intensity levels . Figure 4 illustrates a series of long-term average spectra produced at different vocal efforts(4) . In addition to changes in overall intensity, there arealso changes in the shape of the spectrum, and in the rela-tive levels of the low and high frequency components.

Whispered SpeechThe waveform and spectrum of whispered speech is

considerably different from that of typical voiced speech,as shown in Figure 5 . The top panel shows two wave-forms of the word "shoot," one spoken normally (upper)and the second whispered (lower) . In whispered speech,all components are essentially unvoiced . Therefore, the

54


Figure 4.Speech spectra . Examples of speech spectra produced at different vocalefforts (adapted from Pearsons KS, Bennett RL, Fidell S . Speech levelsin various noise environments . Washington, DC : U .S . EnvironmentalProtection Agency. Report No . EPA-60011-77-025, 1977).

Figure 5.Whispered speech . Top: time-intensity waveforms of the word"shoot" voiced (upper graph) and whispered (lower graph) ; bottom:examples of spectra of voiced and whispered speech .

dominant low frequency portion of the spectrum, madeup of harmonics of vocal cord vibrations and vocal tractresonances associated with vowels, is considerably re-duced. This is illustrated in the bottom panel, whichshows a comparison of the spectrum of voiced speechversus whispered speech.

Modifying a hearing aid to listen specifically towhispered speech would primarily involve altering theshape of the frequency response to accommodate the al-tered speech spectrum, and providing adequate gain toensure audibility of the reduced input levels . A flatter re-sponse, with more mid- and low-frequency gain mayhelp to emphasize the reduced mid-range spectrum inwhispered speech . Appropriate signal processing for thistype of listening situation would include TILL, whichprovides maximum gain for soft, high frequency sounds,or wide range, multichannel compression, which wouldboost softer inputs. Even a linear, peak-clipping circuit,with an appropriate frequency response, would likelysuffice under this listening condition, as long as inputlevels remain low, and the user has access to a volumecontrol.

Loud SpeechAt the other end of the speech intensity range is

shouting, or loud speech, which has a slightly differentspectrum from casual speech (Figure 4) . For loudspeech, the peak of the spectrum shifts upward, above1,000 Hz, and the roll-off between the peak and the lowerintensity, high frequency portion of the spectrum issteeper . During these intense vocal efforts, the vocalcords open and close more rapidly and remain open for ashorter period of time, which increases the amplitude ofthe higher harmonics. For most individuals, less gainwould be needed for this signal, since the level is signifi-cantly elevated compared to average conditions . Also,because the low and mid-frequencies contribute themajor portion of the loudness of speech, a frequency re-sponse with reduced low and mid-frequency gain may bemore appropriate . BILL processing, single-channel AGC,or multichannel compression would be appropriate pro-cessing to deal with the elevated low and mid-frequencyinput levels to maintain user comfort . Peak-clipping cir-cuits may exhibit severe distortion under these conditionsand, therefore, compression limiting would be the prefer-able form of output limiting.

For a multichannel compression instrument, wherethe crossover frequency between channels may be near1,000 Hz, the change in the loud speech spectrum may

shout

Loud

i

Raised

Norrttal

Casual

1000

10000

1/3 Octave Band Center Frequencies (Hz)

70

60 —

!i50-co40 -

30 —

20 .

too

so

0 .00

Time secs .

0 .55

sh

o0

t

50 —

10

1000

10000

1/3 Octave Band Center Frequencies (Hz)

0too

40

30ii

m 20

55


actually shift the peak energy from one channel into an-other. Because the loudest signal within a channel'sbandwidth controls the gain of a compression amplifier,this shift of spectral peak may actually result in a largereduction in gain in the high frequency channel . Onepossible modification for a multichannel compressionaid for this type of input signal would be to shift thecrossover frequency upward, to maintain the peak of thespectrum within the low frequency channel, therebyminimizing its effect on gain for softer high frequencycomponents.

COMMUNICATION SYSTEMS

Speech that has been preprocessed through anothercommunication device, such as a telephone or intercom,will have a different frequency spectrum from that shownin Figure 1, owing to the transmission characteristics ofthe device. In addition to the system's frequency re-sponse, some of these devices contain built-in AGC cir-cuits, which further alter the normal speech signal.

TelephoneThe telephone, for example, has a bandpass charac-

teristic that is reasonably flat from approximately 300 to3,000 Hz. Below and above those frequencies, however,the spectrum rolls off quite dramatically . There are twomain considerations for acoustically coupling a hearingaid to a telephone handset : (a) minimize feedback, and(b) compensate for the altered speech spectrum . The firstgoal can usually be accomplished by significantly reduc-ing the gain above 3 kHz through frequency-shapingmodifications, such as a high cut trimmer or program-mable gain . Because of the limited bandwidth of thetelephone system, reducing the gain in this frequency re-gion does not alter the speech signal, yet does signifi-cantly reduce the likelihood of feedback . More emphasiscan be provided to the mid-frequency region for thistype of fitting, to maximize audibility of the speech in-formation in that range . Typically, most telephonespeech occurs over a fairly restricted intensity range, andsignal-processing considerations are, therefore, of sec-ondary importance compared to providing an appropri-ate frequency response with adequate gain and minimalfeedback.

For magnetic telecoil systems, response modifica-tions are fairly limited in most instruments ; however,some programmables do allow the frequency response of

the telecoil system to be modified independently of that ofthe microphone (5) . There is recent evidence that shapingthe telecoil frequency response to provide real ear gainthat matches a speech-based prescriptive target, such asNAL, may improve performance (6) . Use of a telecoil hasbecome more difficult in many of today's office environ-ments, where high levels of electrical interference fromfluorescent lighting, computer terminals, or other elec-tronic equipment exist . These devices radiate energy atharmonics of the frequency of the power supply circuit(60 Hz) . By reducing the gain of the telecoil in the fre-quency region of the higher level harmonics (below 300Hz), this electrical background noise can be minimized.Since the telephone signal rolls off below 300 Hz, thereis little information lost through the modification.

IntercomsFor intercoms and other speaker systems, the fre-

quency response, as well as its overall level, will dictatethe type of fitting modification required . Most intercomsystems do not use particularly high quality components;therefore, the output signal may be significantly de-graded to begin with . In general, the frequency responseof the system is the major consideration, as opposed tointensity concerns, since here again, the output range formost of these systems is fairly restricted . Typically, theycontain relatively small speakers, which reproduce mid-and high frequencies more efficiently than low frequen-cies . The spectrum of the signal of interest then, willlikely be more mid- and high-frequency biased than livespeech. To compensate for this variation in input signal,more low frequency gain can be provided in the hearingaid response to improve the sound quality . In many cases,intelligibility may be determined more by the characteris-tics of the output device, rather than by the processing ofthe hearing aid.

SPEECH COMBINED WITH OTHERSIGNALS

Because of the importance of communication ineveryday life, speech remains the primary signal of in-terest in the majority of typical listening situations.There are times, however, when nonspeech signals arethe point of interest, such as listening to music, or othersituations, often entertainment-related, where speechmay be combined with other signals, such as music orsound effects .

56


TelevisionListening to television is often one of the first situ-

ations where hearing loss becomes a noticeable problem.Because television programming contains speech at avariety of levels, as well as music, audience soundtracks, and any number of other sounds, the viewer isexposed to dynamically changing intensity levels . Fromthe softest speech sounds of a quiet conversation onemoment, to a loud trumpet fanfare introducing thenewest variety of breakfast cereal the next, the individ-ual with hearing loss, aided or unaided, is faced with achallenging listening situation. Unaided, most raise thetelevision volume to the level needed to hear the softestspeech segments . The much louder commercials andmusic may then be too loud, due to the reduced dynamicrange associated with many types of hearing loss . For ahearing aid wearer, particularly if using a linear circuit, asimilar outcome may result (the television volume con-trol is simply a linear amplifier) . By turning up the gainof the instrument to hear the soft speech, too much gainmay be provided for louder sounds, or distortion frompeak clipping may occur, either one an uncomfortablelistening experience.

On television there may be multiple signals of inter-est, although speech most likely remains the primary one.The speakers used in most sets are often of fairly lowquality, although that aspect has improved recently withthe advent of stereo broadcasts . In general, however, tele-vision speakers are smaller in size, which limits their abil-ity to reproduce low frequency sound efficiently. From afrequency response point of view, a hearing aid fittingwith additional low frequency gain will produce a morepleasant sound quality for listening to television, but theprimary concern of a fitting specific to this situation is thewide range of intensities to which the listener is exposed.This artificially manipulated range may, in fact, be greaterover a relatively short period of time, than one might oth-erwise experience during a typical day away from the TVset . Because of this, the use of some type of compressioncircuit will generally provide the most comfortable listen-ing experience for most users . The more adaptive multi-channel compression processing, which can respondappropriately to the broad spectral and intensity changes,may provide the best performance in this situation. Sin-gle-channel devices with compression-limiting circuitswould also provide an improvement over peak-clippingdevices, in terms of sound quality . For this situation, theissue of intensity dynamics generally outweighs the con-sideration of frequency response .

Theaters and AuditoriumsMuch of the above discussion related to television

is germane to considerations for movie theaters, live the-ater, and other large listening environments . In these situ-ations, consideration of intensity dynamics remains theprimary focus over frequency content issues . In fact,most theaters have relatively high quality sound repro-duction systems (capable of producing high output lev-els) ; therefore, the frequency response of the outputsource is generally not a limiting factor. However, inthese larger venues room acoustics may become more ofa factor for consideration . Places of worship, auditori-ums, and the like, with bright, solid, reflective surfacesoften tend to create reverberant listening conditions . Re-verberation of low frequencies degrades the temporal en-velope of speech, masking many lower intensity speechcues . In such a situation, a multichannel compressionsystem, where low frequency speech components areprocessed separately from high frequency components,may prove advantageous . By separating speech informa-tion into multiple bands, appropriate gain can be pro-vided for softer high frequency components, independentof reflected, low frequency energy . Single-channel BILLprocessing would also serve to minimize low frequencygain in these types of conditions, and would likely pro-vide an improvement over linear processing.

MusicListening to live or recorded music and to many of

the situations discussed above where music may be partof the overall content, is a different listening experiencefrom most others . Certainly, the frequency content ofmusic can differ quite significantly from that of speech.Similarly, the intensity range of music is also quite likelyto exceed that of speech . In addition, the receptive atti-tude of the listener may also constitute a major factor infitting for this listening situation . Many individuals listento music at relatively loud levels (some to the extent ofbeing dangerously loud) compared to other types of ma-terials . In terms of absolute intensity, these levels may befar beyond what the same individual would tolerate forany period of time for listening to speech, or to a babycrying, or to nails on a chalkboard . These examples illus-trate that the nature of the material, rather than simply theabsolute levels alone, exerts a major influence on the ac-ceptability of intensity levels for specific stimuli . Bentlerobserved exactly this phenomena in a study that exam-ined the relationship of stimulus material to the reportedUCL (7). Her data showed a clear trend toward lower re-

57

Chapter Four: Circuitry Options-Altered Signals

ported UCLs for stimuli typically considered to be aver-sive, compared to those considered pleasant.

Thus, in addition to frequency content and dynamicconsiderations, the nature of the signal of interest, in thecase of music, must also be accounted for. In general,most listeners will tolerate, and may in fact prefer, louderlevels for music than for other materials . Hearing aid fit-tings for listening to music should address all of these is-sues . Compared to listening to speech, the preferredfrequency response for music will generally have morelow frequency gain, and a somewhat flatter response (8).This setting will provide a fuller, richer sound quality andprovide emphasis to the bass components, compared tothe typical high frequency emphasis desired for speech,where priority is placed on intelligibility. In a study ofuser preferences of different frequency responses for var-ious listening materials, Fabry and Stypulkowski foundthat most listeners tended to select the greatest amount oflow frequency gain for listening to music, compared tolistening to speech in quiet or noise (9).

The intensity range of music is also often greaterthan that of everyday speech, suggesting that compres-sion circuits may be appropriate, certainly over the use ofpeak-clipping circuits, since sound quality is a major fac-tor when listening to music . Multichannel, wide-rangecompression may improve sound quality and naturalnessin perceived intensity levels, particularly if the dynamicrange is significantly different in the low and high fre-quency regions . Because of the nature of the material,output levels may be set higher for this specific listeningcondition than for others, as described above . Input com-pression circuits, which provide the user with control ofboth gain and output levels, would be more appropriatehere than would output compression circuits, where theuser would be unable to control the loudness of higher in-tensity sounds.

In general, as the relative spectral differences be-tween the low and high frequencies of an input signal be-come greater, multichannel compression offers a greaterpotential to improve signal processing performance . Sim-ilarly, as differences in the user's dynamic range becomegreater in different frequency regions, multichannel pro-cessing may also prove more beneficial.

MULTIPLE-RESPONSE INSTRUMENTS

As outlined earlier, the use of very specific fittingsor signal processing for individual listening situations islikely to involve the use of multiple-response instru-ments . Creating different fittings for specific listening

conditions is a luxury that is really only available withsuch programmable instruments . The decision as towhether to prescribe one can be based to a great extentupon user lifestyle and exposure to distinctly differentlistening situations. Many of the situations described inthis and earlier chapters represent listening environmentswhere multiple-response instruments can be beneficialwhen appropriately fit to optimize performance for thespecific conditions . A number of recent studies haveshown that individuals will consistently and effectivelyutilize a number of different programs of a multiple-memory instrument in different listening situations(10-13) . For example, Ringdahl reported that most sub-jects used 2-5 different programs of a multiple memory,multichannel compression instrument in their daily use,and consistently selected specific programs for specificlistening conditions (14).

It should be noted that just as with single-responsehearing aids, multiple-response instruments span a com-plete range of signal-processing capabilities . Some multi-ple-response instruments contain single-channel, linear,peak-clipping circuits. This system primarily allows dif-ferences in frequency response to be created between thedifferent programs. At the other end of the spectrum aremultiple-response, multiple-channel compression instru-ments that not only allow changes in the frequency re-sponse of different programs, but also in the signalprocessing that can be created for different listening situ-ations (15) . Instruments with this capability offer thegreatest flexibility, and the greatest ability to create dra-matic differences in program responses within the samehearing instrument.

Successful use of multiple-response instruments isnot dictated by audiogram type, prior hearing aid experi-ence, or age of the user. Studies have reported successfuluse of them across a wide population of users with di-verse audiometric profiles and history of use, including alarge population of children (10,14) . In general, individ-uals with hearing losses on the extremes of the distribu-tion (i .e ., very mild or very severe) may not benefit asmuch from the use of multiple memories due to limita-tions in frequency response or signal processing differ-ence that can be provided between programs for thesetypes of losses . Most users, however, can benefit fromthe availability of one or two additional programs de-signed for specific listening circumstances in their life.Experienced hearing aid users, familiar with the situa-tions that present listening difficulties for them, are oftenthe most successful users of multiple-response instru-ments .

58


REFERENCES

1. Mueller HG. Update on programmable hearing aids . Hear J1994 ;47(5) :13.

2. Marion M. Programmable hearing aids, fact or fiction . Pre-sented at the American Academy of Audiology Meeting.Phoenix, AZ, 1993.

3. Killion MC, Staab WJ, Preves DA . Classifying automatic sig-nal processors . Hear Instrum 1990 ;41(8) :24-6.

4. Pearsons KS, Bennett RL, Fidell S . Speech levels in variousnoise environments . Washington, DC : U .S . Environmental Pro-tection Agency. Report No . EPA-60011-77-025, 1977.

5. Stypulkowski PH. 3M programmable hearing instruments. In:Sandlin R, editor. The application of digital technology to hear-ing aid devices . Boston : Allyn & Bacon, 1992.

6. Davidson SA, Noe CM . Programmable telecoil responses : po-tential advantages for assistive listening devices . Am J Audiol1994 ;3 :59-64.

7. Bentler RA . Relationship of perceived quality dimensions tothreshold of discomfort. Presented at the American Academy ofAudiology Convention, Phoenix, 1993.

8. Franks JR. Judgments of hearing aid processed music. Ear Hear1982 ;3(1) :18-23.

9. Fabry D, Stypulkowski PH . User selected hearing aid fittingsfor different listening environments . Presented at the AmericanAcademy of Audiology Convention, Nashville, TN, 1992.

10. Ringdahl A, Eriksson-Mangold M, Israelsson B, Lindkvist A,

Mangold S . Clinical trials with a programmable hearing aid setfor various listening environments . Br J Audiol1990;24 :235-42.

11. Goldstein D, Shields A, Sandlin R . A multiple-memory, digi-tally-controlled hearing instrument . Hear Instrum1991 ;42(1) :18-21.

12. Kuk FK. Evaluation of the efficacy of a multimemory hearingaid . J Am Acad Audiol 1992 ;3 :338-48.

13. Stypulkowski PH, Hodgson WA, Raskind LA. Clinical evalua-tion of a new programmable multiple memory ITE. Hear In-strum 1992 ;43(6) :25-9.

14. Ringdahl A . Listening strategies and benefits when using a pro-grammable hearing instrument with eight programs . Ear NoseThroat J 1994;73(3) :192-6.

15. Stypulkowski PH. Fitting strategies for multiple memory pro-grammable hearing instruments . Am J Audiol 1993 ;2(2) :19-28.

PAUL H. STYPULKOWSKI, PhD is currently Marketingand Technical Support Manager for SONAR Hearing Health inEagan, Minnesota. Dr. Stypulkowski received his doctorate inAuditory Physiology and Neurobiology from the University ofConnecticut. His research and publications include landmarkwork on cochlear implant devices as well as high technologyhearing aids .

Circuitry Options: Signal Processing Under Altered Signal ...

Documents

Transcript of Circuitry Options: Signal Processing Under Altered Signal ...