Related Experiment Videos
Coding of the speech spectrum in three time-varying sinusoids
Annals of the New York Academy of Sciences
|January 1, 1983
Summary
Phonetic information is perceived in sinewave speech, which lacks natural acoustic elements. Preserving temporal variations in the acoustic signal allows for phonetic perception, suggesting a focus on spectrum variation for speech processing.
Area of Science:
- Psychoacoustics
- Speech Perception
- Auditory Neuroscience
Background:
- Natural speech signals contain complex acoustic elements like harmonics and broadband formants.
- Sinewave replicas of speech are simplified tonal patterns preserving only formant peak frequencies and amplitudes.
- These sinewave signals are spectrally unnatural, lacking typical speech characteristics.
Purpose of the Study:
- To investigate whether phonetic information can be conveyed by sinewave speech replicas.
- To determine the critical acoustic properties for phonetic perception in altered speech signals.
- To explore the role of temporal spectrum variation in speech perception.
Main Methods:
- Adult listeners were presented with sinewave replicas of speech signals.
- The sinewave replicas mimicked the first three formant peaks of natural speech.
- Experiments focused on preserving the temporal dynamics of spectral changes.
Main Results:
- Listeners successfully perceived phonetic information from sinewave speech replicas.
- The unnatural timbre of sinewave signals did not prevent phonetic perception.
- Phonetic perception was maintained when the temporal properties of spectrum variation were preserved.
Conclusions:
- Phonetic perception can be achieved with highly simplified, spectrally unnatural speech signals.
- The temporal variation of the acoustic spectrum, a second-order property, appears crucial for speech perception.
- This suggests that coherent spectrum variation, rather than specific acoustic features, underlies phonetic perception.