2009/5/16 HST.723J Theme 7 Presentationweb.mit.edu/hst.723/www/ThemePapers/Pitch/... · 2009/5/16 HST.723J Theme 7 Presentation Introduction • Accurate neural coding of the pitch

2009/5/16 HST.723J Theme 7 Presentation


Introduction• Accurate neural coding of the pitch of complex

sounds is an essential part of auditory scene analysis

• Differences in pitch help segregate concurrent sounds, while similarities in pitch can help group sounds from a common source.

• In quiet, nonreverberant backgrounds, pitch can be derived from timing information in broadband high-frequency auditory channels and/or from frequency and timing information carried in narrowband low-frequency auditory channels.


f0

Repetition rate=f0

Courtesy of Dr. Bendor

How does the auditory system compute pitch?


Cochlear Filtering, Temporal Fine Structure, & Temporal-Envelope Modulation

Resolved: low number of harmonics will each pass through a different filter

Unresolved: in high frequency area　 each filter pass several harmonics

Temporal mechanism: rely on “phase-locking”

Spectral mechanism:rely on resolved harmonics in orderto determine the frequency composition of the sound.

Harmonic-related


Natural pitch-evoking sounds vs artificial ones

1. Dynamic pitch

2. Usually heard in reverberant environments: (eg. when listening to an actor on stage in a theater)


Materials and methods

Recording the responses of VCN neurons of anesthetized guinea pigs

To harmonic complex tones With time-varying F0s (F0-swept harmonic

complexes) Under a range of real-room reverberation

conditions With varying reverberation strength


Complex Stimuli

200–400 Hz F0-swept harmoniccomplex tones


Outline

1. Neuronal Responses to Dynamic PitchStimuli with Added Reverberation

2. The Effect of Harmonic Resolution3. The Effect of Phase-Locking Ability4. Pitch-Related Temporal-Envelope Cues

Are More Degraded than Fine-Structure Cues in Reverberant Spaces

5. Influence of F0 Sweep Rate


1. Neuronal Responses to Dynamic Pitch Stimuli with Added Reverberation





Responses to F0-Swept Harmonic Complex from Pri-like Unit & Low-Freq. Unit

autocorrelogram


Short Summary

• Both units can reliably represent the dynamic pitch in the absence of reverberation.

• Adding strong reverberation results in a dramatic reduction in the pitch-related response of the high-BF primary-like unit

• The effects of reverberation on single unit pitch-related responses seems to depend mainly on unit’s BF







Temporal Representations of Responses: Unresolved vs Resolved

Short Summary:Harmonic resolution is at least one important factor in determining the effect of reverb on the ability to represent the pitch of harmonic complex sounds in the temporal pattern of AP discharge.


Fine-Structure vsEnvelope-Modulation Spectra


Spectral Representations of Responses: Unresolved vs Resolved

the only cue to pitch is the noisy fine structurearound BF

No modulation band

Pitch of the unresolved stimulus relies more on AP locked to the fine structure & less on AP locked to the temporal-envelope modulation

the effect of increasing reverberation strength isto smear the frequency-domain representation over time


Effects of F0• peak reduced in amplitude

(due to the reduced envelope-modulation) & shifted to a longer period reflecting the smeared temporal fine structure arising from the addition of indirect sound components.

• the effect of increasing reverb strength is to smear the frequency-domain representation over time

F0 significant effects of harmonic resolvability and reverberation strength on the position of peak, which reflects the smearing of the fine structure through time







Population Data in Response to200–400 Hz F0-Swept Harmonic Complexes

Summary: under reverb listening condition the temporal representation of pitch in the central auditory system relies on the responses of neuronstuned within the range phase-locking

Phase lock at 3.5k Hz Phase lock 1.5 K Hz

Correlation index decreases with increasing reverberation, not significantly greater than 1 at 3.5k & 1.5k respectively










Phase lock at 3.5K Hz

Phase lock at 1.5K Hz

Temporal band

Fine structure band (preserved at early stage)

N= 56

N=72


Short Summary

• Temporal cues degraded more • Fine structure preserved in the responses

of neurons at this early stage of central auditory processing likely contributes to the increase in errors

when attempting to track a sound’s pitch over time in a reverberant environment against a complex background of other pitch-evoking sounds







The effect of F0 Sweep rate

Summary: rapidly varying F0 and reverbsignificantly diminish the magnitude and increase the smearing of the temporal representation of pitch

BF 1.35K Hz

Lose of env-mod rx

Fine structure blurred


Summary• Recording from VCN of anesthetized guinea pigs showed

that neural representation of pitch based on temporal information is severely degraded in the presence of reverb, and degradation increases with increasing reverb strength.

• In a parallel human psychophysical study, when restricted to using envelope-modulation cues in high-frequency channels under reverb. conditions, human listeners demonstrated a dramatically reduced ability to discriminate pitch-evoking sounds from noise.(supplement)

• Together, these findings explain the origin of perceptual difficulties experienced by both normal-hearing and hearing-impaired listeners in reverberant spaces.

Documents

2009/5/16 HST.723J Theme 7 Presentationweb.mit.edu/hst.723/www/ThemePapers/Pitch/... · 2009/5/16 HST.723J Theme 7 Presentation Introduction • Accurate neural coding of the pitch