russotti duffy - three methods for unscrambling helium speech

U0 NE

Submarine Base, Groton, Conn.

REPORT NUMBER 602

AN EVALUATION OF THREE METHODS FOR UNSCRAMBLINGHELIUM SPEECH PRODUCED AT DEPTHS OF 800 AND 1000 FEET

Joseph S. Russottiand

Joseph R. Duffy

Bureau of Medicine and Surgery, Navy DepartmentResearch Work Unit MF4306.03-2020D.01

Approved and Released by:James E. Stark, CAPT MC USNCOMMANDING OFFICERNaval Submarine Medical Center30 October 1969

This document has been approved for public release and sale; its distribution is unlimited.

http://archive.rubicon-foundation.org

AN EVALUATION OF THREE METHODS FOR UNSCRAMBLINGHELIUM SPEECH PRODUCED AT DEPTHS OF 800 AND 1000 FEET

byJoseph S. Russotti

andJoseph R. Duffy

SUBMARINE MEDICAL RESEARCH LABORATORYNAVAL SUBMARINE MEDICAL CENTER REPORT NO. 602

Bureau of Medicine and Surgery, Navy DepartmentResearch Work Unit MF4306.03-2020D.01

Submitted by:

J. Donald Harris, Ph.D.Head, Auditory Research Branch

Reviewed and Approved by: Reviewed and Approved by:

Charles F. Gell, M.D., D.Sc. (Med)Scientific DirectorSubMedResLab

Joseph D. Bloom, CDR, MC, USNDirectorSubMedResLab

Approved and Released by:

E/ Stark, CAPT MC USNCOMMANDING OFFICER

Naval Submarine Medical Center

This document has been approved for public release and sale; its distribution is unlimited.

4 7 o -ý ifILr f-Tý


SUMMARY PAGE- REPORT NO. 602

THE PROBLEM

To evaluate three systems for "unscrambling" helium speech producedat the more severe depths of 800 and 1000 feet and compare these to un-altered helium speech and helium speech played back at one-half normalspeed.

FINDINGS

The increase in speech intelligibility imparted by the three commercialsystems was limited at these depths although two of the units showed asignificant improvement over the unaltered helium speech. Half-speedplayback of unaltered helium speech was found superior to all other meth-ods of presentation. Methods for improving the operation of these unitsare suggested.

APPLICATIONInformation contained in this report is useful to the design of systems

for improved reception of helium speech by divers during deep-submergence.

ADMINISTRATION INFORMATION

This investigation was conducted as a part of Bureau of Medicine and SurgeryResearch Work Unit MF4306.03-2020D-Sensory Aids for Verbal Communication UnderWater by Divers and Swimmers. The present Report is No. 1 on this Work Unit. Itwas approved for publication on 30 October 1969 and designated as Submarine MedicalResearch Laboratory Report No. 602.

PUBLISHED BY THE NAVAL SUBMARINE MEDICAL CENTER

ii


ABSTRACT

Three systems for "unscrambling" helium speech produced at simu-lated depths of 800 and 1000 feet were evaluated. Recordings of the Modi-fied Rhyme test read by an experienced male diver in a pressurized heliumatmosphere were processed through three commerical frequency-shiftingdevices. The intelligibility of words passed through these systems wascompared to scores obtained when the same helium recording was presenteddirectly without alteration and when presented at one-half playback speedto groups of listeners. A within-subjects' design was employed with suit-able counterbalancing so that no word was heard twice by the samelistener. Half speed playback (58.57c intelligibility) was significantly su-perior to all other conditions. The Kay Electric Company "Varivox" (43.2%intelligibility) was significantly inferior to the other two commercial sys-tems. The Industrial Electronics Corp. and the Gotham Audio Corp. unitswere equal in performance (50.77( and 51.0% intelligibility respectively)and superior in intelligibility to unaltered helium speech (45.7%). The in-crease in the degree of intelligibility imparted by the three commercialsystems is limited at the depths under study. Methods for adapting theseunits for improved operation at greater depths are suggested.

i..


AN EVALUATION OF THREE METHODS FOR UNSCRAMBLING HELIUMSPEECH PRODUCED AT DEPTHS OF 800 AND 1000 FEET

INTRODUCTIONEfforts to "unscramble" helium-speech

have led to the development of commerciallyavailable systems which lower the frequen-cies of helium speech to a more naturalrange. At present, these mechanical, elec-tronic or electromechanical systems have alimited range through which frequencies canbe translated. Since with increased depththe introduction of helium breathing mix-tures causes a very considerable elevation inthe frequency components of speech, theworking capability of these units may wellbe exceeded. The present study evaluatedthree such systems for unscrambling heliumspeech, on the basis of the intelligibility ofthe speech at depths of 800 and 1000 ft. Forcomparison, a half-speed tape playback andan unaltered version of helium speech atthese depths were included. The units stud-ied were the Kay Electric Company "Vari-vox," the Eltro Company Rate-Changer (soldin the U.S.A. by Gotham Audio Corp. as the"Tempophon"), and the Integrated Electron-ics Corp. Model 701 Helium Voice Unscram-bler.

Both the Varivox and Eltro units use asegmented rotating head which electrome-chanically splices tape so as to lower the fre-quency positions of recorded signals by re-ducing the relative speed of short sections oftape which contact energized segments ofthe playback head. Except for the shortdelay in speech reception caused by the dis-tance between record and playback heads,real time is preserved. The all-electronic IECSystem, on the other hand, achieves real timeby using heterodyning techniques which low-er frequency. The signal is divided into twoseparate frequency bands, heterodyned andfinally reconstituted for real-time reception.

*Normal speech produced by loudspeaker and pickedup by this microphone at pressures equivalent to1000 ft submergence was easily understood and in-stantly recognized as a reproduction of normalspeech without distortions due to depth or helium.

PROCEDURE

The voice of an experienced male diver wasrecorded from inside a pressurized test cham-ber containing the gas mixtures shown inTable I, at pressures equivalent to thosefound at 800 and 1000 ft of water. Record-ings were made of the diver's voice when heread the same 50-word list from the ModifiedRhyme Test (MRT) twice at both depths,once with the microphone connected directlyto a General Radio Data Recorder locatedoutside the chamber, and then again with theIEC unit between the microphone output andthe recorder. The microphone was especiallyconstructed by IEC for use with their un-scrambler under conditions of high ambientpressure.*

For each of the two depths a working mas-ter tape was produced, which consisted ofthree identical re-recordings of the originallyrecorded unaltered helium-speech transmis-sion and one re-recording of the originallyrecorded IEC alteration. An additional re-recording of the unaltered helium transmis-sion, played back at half original speed, wasalso included on each tape.

Ten stimulus tapes for presentation togroups of listeners were finally prepared,each consisting of the same word list present-ed at either of two depths, under one of fivetest conditions: (1) unaltered helium speech,(2) half-speed playback, (3) Eltro alteration,(4) Varivox alteration, and (5) IEC altera-tion.

Field operation of the IEC system in thisstudy was by IEC technicians. The Varivoxand Eltro units were adjusted for optimumperformance and operated in this laboratoryby the authors.

The intelligibility scores reported here areby no means representative of the perform-ance of any of the three systems at the muchshallower depths for which they were de-signed.

Since all ten conditions (five at each depth)

1


TABLE IContents of Breathing Mixtures.

800 ft. 1000 ft.

96.50% Helium 97.46% Helium1.22% Nitrogen 1.46% Nitrogen

.17% Carbon Dioxide .10% Carbon Dioxide2.02% Oxygen 1.04% Oxygen

Temp. Approx. 800Rel. Humidity 40-50%

were comprised of the same 50 words, sep-arate listening groups were necessary. Toreduce the number of such groups to lessthan 10, the high split-half reliability of theMRT word list allowed constructing the tapesusing only 25-word lists, with items 1-25 al-ways taken from the 800-ft and items 26-50always from the 1000-ft tape. In an effort tocancel group differences and minimize ordereffects, the resultant five 50-word tapes,which now contained samples from bothdepths, were systematically spliced in groupsof five words so that items 1-25 contained allfive conditions at 800 ft, and items 26-50 con-tained all five conditions at 1000 ft. The orderof presentation of these two sets of five con-ditions was arranged according to two 5 x 5Latin Squares derived from a Greco-Latinsquare. Thus, the design for presentation ofthe data made use of a sample of listenersdetermined from the responses of five groupsof actual listeners.

An advantage to this unusually complexdesign for presentation of words was that aspecific set of five words for a particulartreatment at either depth was never heardmore than once within or between groups,but each of the 25 words was heard 28 timesfor a particular treatment. In other words,a sampling of 28 listeners theoretically atrandom heard 25 different words for each offive treatment conditions, at two differentdepths, although the equivalency inherent inusing the same 25 words for all treatmentsat a particular depth was retained.

Each of five groups of 28 listeners, all withnormal hearing, heard a different one of thefive 50-word lists monaurally in a roomequipped with 49 matched Telephonics TDH-

39 earphones mounted in Otocups. The meanlevel of the speech signals on a stimulus tapewere set at 75 db Sound Pressure Level(SPL). Responses were made on a machine-scoring answer sheet especially constructedfor the MRT test (See Appendix I).

RESULTS

Data from the five groups were combinedto control order effects, and each experimen-tal condition at both depths was analyzed bymeans of a treatments x treatments analysisof variance for repeated measures. Table IIshows a significant difference (level of confi-

TABLE IIAnalysis of Variance of Test Condition X Depth.

df MS F

Between Ss 139 1.084 .969Within Ss 1260 1.189Depth 1 .686 .613Test Condition 4 23.990 21.457*Depth X Test Cond. 4 .983 .879Total Error 1251 1.118Total 1399 1.179

*Significant at 99% level.

dence better than 991%) for the conditions ofaltering the helium speech heard by each lis-tener. No significant difference was found,however, between intelligibility for the twodepths. Further, the interaction betweendepth and condition was not significant (95%level).

Based upon these results, data for the twodepths were combined, and a revised sum-mary table for a one-way analysis of varianceusing repeated measures was performed.Table III, a summary of this analysis, showsthat the difference for condition is decreased,but still significant (99% level).

The mean correct responses to the variousconditions are in Table IV. Comparison ofconditions by a Cochran statistic showed thatthe assumption of homogeneity of varianceis valid.

2


TABLE IIISummary of Revised Analysis of Variance.

df MS F

Between Ss 139 2.168Within Ss 560 2.654Test Condition 4 47.980 16.717*Residual Error 556 2.870Total 699 2.557

*Significant at 99% level.

According to a Newman-Keuls comparison(see Table V) the half-speed condition wassignificantly better (99 % level) than all otherconditions. The Varivox was significantlyinferior (99% level) to the Eltro, IEC, andhalf-speed conditions, and slightly worse thanunaltered speech, though this trend was notformally significant (<95%). Finally, theEltro and IEC units yielded similar data, bothsignificantly better (95% level) than unal-tered speech.

TABLE IVPercent Correct Responses Based Upon 50 Words

Under Each Condition for 28 Theoretical Listeners.*

UnalteredVarivox He0 2 IEC Eltro Half Speed

43.2% 45.7% 50.7% 51.0% 58.5%*28 Ss each group x 5 groups = 140 actual listeners.

TABLE VSummary Table of Newman-Keuls Comparison of

Test Conditions Showing Level of SignificantDifference Between Pairs of Means.

HalfSpeed Eltro IEC Unalt Varivox

Half Speed 99% 99% 99% 99%Eltro * 95% 99%IEC 95% 99%Unaltered **Not Significant.

DISCUSSIONSince speech intelligibility was best under

half-speed playback and was, in fact, signifi-cantly better than any other condition, onemay conclude that the splicing/heterodyningprocesses employed in maintaining real timewhile shifting frequency do so at the expenseof intelligibility.

Considering together all the ways usedhere for treating helium speech, the intelligi-bility appears poorer than might have beenexpected. While this overall decrement isprimarily due to the excessive distortionsproduced at these depths, the particular de-sign employed may have contributed to someextent. In order that each listening grouphear the five experimental conditions at bothdepths, five words from each condition werepresented. Since after each group of fivewords a novel situation confronted the listen-er, learning and adaptation were kept to aminimum. Had listeners been able to "getused to" a particular type of speech at eachcondition, they might well have done better.

The operating ranges of the Eltro, Varivox,and IEC units had apparently been exceededso that they were not quite able to lower thefrequencies to the degree that would theoret-ically be needed. In effect, then, a certainamount of distortion to the speech was stillpresent, and the resultant signal was not per-fectly intelligible. Further, once the Eltroand Varivox units were operated in the upperranges of frequency shift, far too many smallsections of the speech were eliminated. Bothof these systems sample small portions of therecorded signal and reduce frequency by re-ducing the relative playback speed of eachsmall section sampled. The extent of theshift, then, determines the amount each sec-tion is expanded in playback time, and there-fore since real time is maintained, fewer sec-tions are sampled when the frequency shiftis greater.

The Varivox condition was less intelligibleto these listeners than the direct unalteredtransmission; this is most probably attribu-table to the noise which this unit adds to sig-nals during the splicing operation, combinedwith the appreciable loss in speech signalcaused by the sampling techniques.

3


It should be noted that the altered speechproduced by the Eltro unit at these depths isby no means representative of this unit inimproving intelligibility at lesser depths. Ad-vanced design of the switching mechanismemployed in the rotating multiple-playbackhead assembly has lowered appreciably thesplicing noise inherent in the system. In fact,we have found that helium speech producedat one ATA could be altered by the Eltro soas to recreate in part the particular speaker'snormal voice and speech characteristics. Themachine was also successful in rendering in-telligible a sample of helium speech previous-ly recorded at 400-ft depth, although voicequality could not be studied, since no tape ofthe speaker's voice under normal conditionswas available. At the 800-ft depth, however,the frequencies could not be brought farenough down to reach a normal-soundingspeech. At maximum settings the unit sam-pled too few sections of speech and was poorin intelligibility. A compromise was made byhaving three speech-researchers shift theplayback frequencies to what they felt wouldproduce maximum intelligibility; this metwith some success.

We would emphasize that while a decreasein the number of sections sampled does ac-company an increase in frequency shift, thesampling system should not be discarded asincapable of successful operation at greaterdepths. Note that present sampling systemsincorporate four pick-up segments in the ro-tating head array. With, for example, eight-part heads, smaller sections could be sampledand expanded with perhaps smaller loss inintelligibility.

Further refinements seem in order in boththe Eltro and the IEC units, equal here inperformance, to adapt them to greater depths.Factors in favor of the IEC all-electronic unitare its capability of microminiaturization,and elimination of moving parts more proneto breakdown. On the other hand, an advan-tage of the rotating-head system is that re-sulting speech does sound more nearly nor-mal, miniaturization of the rotating-headprinciple is within present-day capabilities,should a demand exist.

SUMMARY

An experienced male diver read a 50-wordlist of the Modified Rhyme Test at depths of800 and 1000 ft simulated submergence, whilebreathing helium. A faithful rendition ofthis speech was played to a group of listenersfor intelligibility testing, together with thesame tape played at half-speed, and as al-tered by three frequency shifting devicesmanufactured by the Industrial ElectronicsCorporation, the Kay Electric Co., and oneof German make sold by the Gotham AudioCorp. Type of speech treatment and groupsof listeners were suitably counterbalanced sothat no word was heard twice by the samelistener. Significantly superior to all otherconditions was one of half-speed playback.The Kay "Varivox" was significantly inferiorto the unaltered condition, though this trendwas not significant. The IEC and the Gothamunits were not different, and both were sig-nificantly superior to the unaltered condition.It was suggested that had these latter unitsbeen able to extend even further the frequen-cy shifts of which they are now capable, in-telligibility of helium speech at these depthsmight well have been further improved.

4


NAME I FORM 3XI I GROUPI(LAST) (FI FST)

RANK/RATE[ I DATEI I TEST NO.'

gold hold sold pop shop hop bill fill till safe save sake came cape cane

21 31 41told fold cold cop top mop will hill kill sale sane same case cave cake

lame lane lace tang tab tack sop sag sad mop mat math wig rig fig

2 ---- ----- 12 22 32 42late lake lay tam tap tan sass sack sat mad mass man pig big dig

bust just rust keel feel peel gale male tale gang hang fang ban back bat-13 -23 -43dust gust must reel heel eel pale sale bale bang rang sang bad bass bath

did din dip heal heap heath peas peal peach sip rip tip test nest best

4 ----- 14 ----- --- 24-----------: ------- -34 --- -------- 44 -------------dim dig dill heave hear heat peat peak peace lip hip dip west rest vest

sin win fin paw jaw saw rent went tent beach beam beak seen seed seek5 ... ... ... 15 . .. 25 35 45

din tin pin thaw low raw bent dent sent bead beat bean seem seethe seep

sun sud sup pub pus puck sun nun gun hen ten then dun dug dub~--- 26-- ------------ 4

6 16 ..... 26 36 46sub sung sum pun puff pup run bun fun den men pen duck dud dung

lot not hot meat feat heat bus buff bug cuff cuss cub led shed red7 17 27 ----- 372---37 47got pat tot neat beat seat buck but bun cup cut cud wed fed bed

pill pick pip kit kick kin tick wick pick park mark hark tease teak tear

8 ----- 18 28 38 48pit pin pig kid kill king kick lick sick dark lark bark teal teach team

may gay pay cook book hook sin sill sit fizz fill fib bit sit hit----- ' ----- --- ,1 :.: ;;.;-11- "-; ".2. ;." :;. .:.

9 --- 19 --- -- 29 39 4day soy way shook look took sip sing sick fin fit fig wit fit kit

paAv pale pay race ray rake name fame tame soil toil oil pad pass path

10 ----- 20 30 40 50page pane pace rate rave raze came game same foil coil boil pack pan pat

APPENDIX I

Machine-score version of Modified Rhyme Test answer sheet developed at Naval SubmarineMedical Center.

5


I I i.I

Security Classification

DOCUMENT CONTROL DATA. R & D(Security classification of title, body of abstract anid indexing annotation must be entered when the overall report is classified)

1 ORIGINATING ACTIVITY (Corporate author)•.a. REPORT SECURITY CLASSIF4CATtON

U- S. NAVAl SUBMAR INE MEDICAL CENTER, Submar ineI UNCLASSI F IEDMedical Research Laboratory 2b. GROUP

3. REPORT TITLE

AN EVALUATION OF THREE METHODS FOR UNSCRAMBLING HELIUM SPEECH PRODUCED ATDEPTHS OF 800 AND 1000 FEET

4. DESCRIPTIVE NOTES (Type of report and inclusive dates)

Interim Report5. AUTHOR(S) (First name, middle initial, last name)

Joseph S. Russotti and Joseph R. Duffy

6. REPORT DATE'7a. TOTAL NO. OF PAGES 7b. NO. OF REFS

30 October 1969 4 --Sa. CONTRACT OR GRANT NO. 9a. ORIGINATOR'S REPORT NUMBER(S)

Submarine .Medical Research Laboratoryb. PROJECT NO. Report No.602

MF 4306.03-2020D.C. 9b. OTHER REPORT NO(S) (Any other numbers that may be assignedthis report)

d.

10. DISTRIBUTION STATEMENT

This document has been approved for public release and sale; Its distributionis unlimited.

Ii. SUPPLEMENTARY NOTES 12 SPONSORING MILITARY ACTIVITY

Naval Submarine Medical CenterBox 600 Naval Submarine 8aseGroton, Connecticut 06340

ABSTRACT

Three systems for."unscrambling" helium speech produced at simulated depths

of 800 and 1000 feet were evaluated. Recordings of the Modified Rhyme test readby an experienced male diver in a pressurized helium atmosphere were processed

through three commercial frequency-shifting devices. The Intelligibility of words

passed through these systems was compared to scores obtained when the same heliumrecording was presented directly without alteration and when presented at one-half

playback speed to groups of listeners. A within-subjects' design was employed withsuitable counterbalancing so that no word was heard twice by the same listener. Half

speed playback (58.5% Intelligibility) was significantly superior to all other con-

ditions. The Kay electric Company "Varlvox" (43.2% Intelligibility) was significantl]Inferior to the other two commercial systems. The Industral Electronics Corp. ald

the Gotham Audio Corp. units were equal in performance (50.7% and 51.0% Intelligibi-

lity respectively) and superior In Intelligibility to unaltered helium speech (45.7%)

The increase In the degree of Intelligibility Imparted by the three commercial system!Is limited at the depths under study. Methods for adapting these units for Improvedoperation at greater depths are suggested.DD FORM 1 73 (PAGE I1DD ,O,651473

S/N OtOl-I.-807- 680 1

UNCLASSI FIEDSecurity Classification

13.


UNLL'bbII" I I')Security Classification

KEY WORDS

Helium Speech Unscrambler3

Underwater Verbal.Communication

Frequency Shi'ting Devices for Helium Speech

D FORMS 17 (BACK)DD I(POAG 21473)(PAGE 2)

UNCLASSI.FIEDSecurity Classification


russotti duffy - three methods for unscrambling helium speech

Documents

unscramblinghelium speech

speech intelligibility

unaltered helium speech

helium recording

report number

present report

severe depths

orgsummary page report