Auditory Processng of Dynamic 39;CenterofGravity39; Signals動(dòng)態(tài) 39聽(tīng)覺(jué)處理；重力 39中心；信號(hào)

1、psychoacoustics of dynamic center-of-gravity signals larry fethashok krishnamurthyohio state universityspectral center-of-gravitynchistovitch and lublinskaja (1976,1979)nperceptual formant at center-of-gravityntwo-formant synthetic vowelnmatched by adjustable single-formant signalncenter frequency o

2、f match depends on relative amplitudes of the two formantsexperimental paradigmchistovitch and lublinskaja resultsvoelcker two-tone signalsvoelcker two-tone signalsninitially, led to the ewaif modelnenvelope-weighted average of instantaneous frequency (time domain)npoint by point multiply e x f valu

3、esnsum over n periodsndivide by sum of weightsnindicates pitch change in periodic signalsnhelmholtz (1954, 2nd english edition)njeffress (1964)ewaif modelsignal ( ) with envelope ( )instantaneous frequency ( ).ewaif (envelope weighted average of instantaneous frequency)( ) ( )ewaif ( )( )x te ti te

4、t i t dtx te t dtiwaif model predictionstwo-tone resolution tasknfeth and omalley (1977)ntwo-tone resolutionndi = 1 db; df independent variablenvoelcker-tone pair pitch discriminationninverted “u-shaped” psychometric functionsncomponents resolved beyond 75% pointn3.5 bark separation = jnndvoelcker s

5、ignal: discrimination taskdiscrimination resultsnjnnd just not noticeable differencenfilled circlesnbreakpoint estimatesnopen circlesncr critical ratio cbwncb empirical cbwnsolid line tw envelopeiwaif modeli intensity w weighted a average of i instantaneous f frequency = centroid of signals positive

6、 power spectrum (anantharaman, et al., 1993)222020( ) ( )iwaif ( )( )| ( )| ( )|e t i t dtx te t dtf x fdfx fdfdynamic center-of-gravity effectnlublinskaja (1996)nthree-formant synthetic russian vowelsnlisteners identified vowels with: nconventional formant transitionsnco-modulated formant pairs tha

7、t exhibit the same dynamic spectral center-of-gravitynid functions were very similar with formant pairs separated by 4.3 bark or lesspsychophysicsnanantharaman (1998)ntwo-tone signals with dynamic c-o-g effectnwe called them virtual frequency glidesnlisteners matched transition rates in vf glides to

8、 those in fm glidesniwaif model predicts results for transitions from 2 to 5 erbdynamic center-of-gravity signalswaveformlong-term spectrumspectrogramrate-matching resultsmodel resultsshort-term running iwaif modeliwaif model resultsapplication of st-iwaif modelmore psychophysicsnresearch question(s

9、)nwhat is being integrated in spectral integration?orornwhere in the auditory system is the processing located?psychophysicsniyer, et al., (2001)ntemporal acuity for fm and vf glidesnstep vs. linear ramp discriminationnsimilar dt values may mean common processnmasking patterns for fm and vf glidesnp

10、eripheral process i.e., energy maskingndifferent results vf not peripheral processtemporal acuity paradigm step (red) versus glide (blue) transitions for fm tone (left panel) and virtual frequency (right panel)temporal acuity resultsjust discriminable step duration for fm (solid lines; filled symbol

11、s) and vf (dashed lines; unfilled symbols) signals. frequency separations are 2, 5 and 8 erbu. the results for 1000 hz are represented by circles and those for 4000 hz by triangles. average for 4 listeners. frequency separation (er bu)2 er bu5 er bu8 er bustep duration (msec)02468101214161820 dynami

12、c center-of-gravity maskers masking of brief probe by fm glide (left panel) and by vf glide (right panel). probe is in the spectro-temporal center of each masker. five auditory filter bands are illustrated.timeflfcfhtimeflfcfhmasking resultsmasking of a 20 ms probe by fm (light blue) and vf (darker

13、blue) maskers. the probe is placed at the beginning, middle, and end of the masker. significant differences are seen at 5 and 8 erb for the middle position and the initial position at 8 erb. average for 4 listeners.probe in initial position0.005.0010.0015.0020.0025.0030.0035.002 erbu5 erbu8 erbufreq

14、uency separation amount of masking (db spl)fm maskervf maskerprobe in medial position0.005.0010.0015.0020.0025.0030.0035.002 erbu5 erbu8 erbufrequency separation amount of masking (db spl)fm maskervf maskerprobe in final position0.005.0010.0015.0020.0025.0030.0035.002 erbu5 erbu8 erbufrequency separ

15、ation amount of masking (db spl)fm maskervf maskerglide direction asymmetryngordon and poeppeln3 frequency ranges: (for f1,f2 & f3)n 30 unpracticed listeners 20 trials / signalnone interval direction identification: up vs. dnnbest results at high frequency (f3) rangen10- through 160 ms up is eas

16、ier to id than dnnless clear-cut results at low or mid-freq. rangesglide direction asymmetrygordon and poeppel arlo (2002)identification of fm sweep direction is easier for rising than for falling tones.glide direction asymmetryndawson, (2002)ntested only high frequency range (f3)npracticed listener

17、s; 100% all conditions!nmodified procedurenrove each frequency sweep over 1 octavenpractice to asymptoteglide id resultsnaverage for 4 listenersnone-interval id taskn250 trials / datum pointnwell-practiced subjsnstarting frequency roved over 1-octave rangensummarynfm easier than vfnup easier than do

18、wnduration (ms)5102030405080160percent correct identification50556065707580859095100fm upfm downvf upvf downcv identification experimentnda ga continuum: varying f3 transitionnduration: 50 ms transition into 200 ms basenf3 onset: 2018 to 2658 hz in 80 hz stepsnf3 base: 2527 hz (constant)nformant tra

19、nsition type:nklatt synthesizernfrequency modulated tone glidenvirtual frequency glidecv identification: stimulispectrogram 1. spectrogram 1. step 1 of klatt monaural continuum/ga/ endpoint cv identification: stimulispectrogram 2. spectrogram 2. step 1 of fm monaural continuum/ga/ endpoint cv identi

20、fication: stimulispectrogram 3. spectrogram 3. step 1 of vf monaural continuum/ga/ endpoint cv identification: stimulispectrogram 4. spectrogram 4. step 1 of dichotic fm continuum/ga/ endpoint cv identification: stimulispectrogram 5. spectrogram 5. step 1 of dichotic vf continuum/ga/ endpoint cv ide

21、ntification experimentnlisteners: 8 adults with normal hearingnprocedure: one interval, 2-afcn3 transition types: klatt, fm or vfn 6 of 8 tokens testedn 20 repetitions / tokennresults are averaged for the 8 listenerscv identification: resultsfig. 4. mean responses for fm tone and virtual glide condi

22、tionsformant 3 onset frequency (in hz)190020002100220023002400250026002700% /da/ responses020406080100fm virtual glide cv identification: resultsfig. 7. mean responses for dichotic conditionf3 onset frequency (hz)190020002100220023002400250026002700% /da/ responses102030405060708090k l att k l att f

23、mvirtual glidepsychoacoustics of dynamic center-of-gravity signalsconclusionsnexcitation is integrated not signal energynthe processing is central not peripheralnmasking patterns are very different ntemporal acuity results are similar for fm & vf glidesndirection id asymmetry is similar for fm &

24、amp; vf glidespsychoacoustics of dynamic center-of-gravity signalsconclusionsn cv identification functions are similar for:nklatt synthesized soundsnfm formant soundsnvf formant soundsnthus, it doesnt matter how excitation is moved from a to b, the brain will interpret it as the same sound.nthe effe

25、ct is evident under dichotic listening; further support for central processing.collaboratorsrob foxnandini iyerjayanth anantharamanewa jacewiczrobin dawsonpsychoacoustics of dynamic center-of-gravity signalsthank youquestions?up vs. down fm glideup vs. down fm glideup vs. down vf glideup vs. down vf

26、 glideeffect of masker directionmasking produced by vf (above) and fm (below) maskers with d f = 5 erb. purple bars are “up” glides; yellow bars are “down” glides. centered probe.010203040506025012560masker durationamt. of masking (db spl)010203040506025012560masker durationamt. of masking (db spl)effect of masker