The perception of speech signals.

Area of Science:

• Acoustic Phonetics
• Auditory Neuroscience
• Speech Technology

Background:

• Speech production involves articulator movements encoded in speech waveform spectra.
• Speech perception is viewed as decoding acoustic features of speech waves.
• Auditory system physiology and psychophysics are relevant to speech perception.

Purpose of the Study:

• To outline the mechanism of speech production and its acoustic encoding.
• To present a speech synthesis method based on spectral parameters.
• To describe acoustic features critical for speech sound perception and relate auditory psychophysics to hearing mechanisms.

Main Methods:

• Analysis of speech waveform spectra to identify articulator motion encoding.
• Development of a speech synthesis method using derived spectral parameters.
• Experimental investigation of acoustic features relevant to speech sound perception, including synthesized speech.
• Presentation of dichotic listening experiments to differentiate speech and non-speech sound perception.

Main Results:

• Articulator motion is demonstrably encoded in speech spectra.
• Spectral parameters provide a basis for effective speech synthesis.
• Specific acoustic features are identified as crucial for perceiving different speech sound classes.
• Dichotic listening experiments highlight distinct perceptual processing for speech versus other sounds.

Conclusions:

• Speech production and perception are fundamentally linked through acoustic spectral analysis.
• Spectral analysis offers a robust foundation for both understanding speech and developing synthesis technologies.
• Auditory psychophysics and physiology provide insights into the decoding of speech acoustics.
• Dichotic listening confirms specialized neural pathways for speech sound processing.