Word segmentation from noise-band vocoded speech

Area of Science:

• Auditory perception
• Speech processing
• Psychoacoustics

Background:

• Spectral degradation, often caused by noise or hearing loss, impairs the acoustic details of speech.
• Understanding spoken language structure, like word segmentation, is crucial for language acquisition and comprehension.
• Listeners may rely on various acoustic and contextual cues to overcome speech degradation.

Purpose of the Study:

• To investigate how spectral degradation affects word segmentation in artificial speech.
• To determine if listeners can utilize other speech cues, such as isolated words or altered speech rate, to improve segmentation under spectral degradation.
• To clarify the relationship between syllable recognition and word segmentation in degraded auditory environments.

Main Methods:

• Normal-hearing adults were exposed to artificial speech vocoded into 16 or 8 spectral channels (spectral degradation).
• Speech streams were presented either continuously, interspersed with isolated words, or slowed by 33%.
• Participants' ability to segment familiar versus novel syllable sequences and recognize individual syllables was tested.

Main Results:

• Spectral degradation significantly impaired both word segmentation and syllable recognition.
• Segmentation performance improved when isolated words were added to the speech stream.
• Slowing speech did not enhance word segmentation under spectral degradation.

Conclusions:

• Syllable recognition is a necessary but insufficient component for successful word segmentation.
• Isolated words serve as a valuable cue, facilitating listeners' ability to access and segment acoustically degraded speech.
• These findings have implications for understanding speech perception in noisy environments and for developing assistive listening technologies.