Learning to recognize talkers from natural, sinewave, and reversed speech samples

Area of Science:

• Auditory Perception
• Speech Processing
• Psycholinguistics

Background:

• Talker recognition is crucial for effective communication.
• Understanding how listeners learn to identify unfamiliar voices is key to speech perception research.
• Previous studies have explored acoustic cues in voice identification.

Purpose of the Study:

• To investigate how listeners learn to recognize unfamiliar talkers.
• To examine the generalization of talker recognition from trained utterances to novel instances.
• To determine the role of phonetic and suprasegmental information in talker identification.

Main Methods:

• Five experiments were conducted involving listener training over several days.
• Listeners identified 10 talkers using natural, sinewave, or reversed speech sentences.
• Sinewave speech preserved phonetic/suprasegmental properties but lacked vocal quality; reversed speech preserved vocal quality but distorted phonetic properties.

Main Results:

• Listeners successfully learned to identify talkers from sinewave speech, indicating recognition without natural vocal quality.
• Generalization performance varied across speech types and was influenced by the salience of phonetic information.
• The findings suggest that phonetic attributes are important for both talker recognition and phonetic perception.

Conclusions:

• Talker recognition can be achieved using acoustic signals that lack natural vocal quality.
• Phonetic information plays a significant role in the generalization of talker recognition.
• Similar phonetic attributes may underlie both talker recognition and phonetic perception processes.