Who is singing? Voice recognition from spoken versus sung speech

Area of Science:

• Psychology
• Acoustic Phonetics
• Speech and Hearing Sciences

Background:

• Singing is a socially significant behavior with unique acoustic properties.
• The acoustic constraints of singing may obscure vocal identity cues present in speech.
• Limited research exists on extracting talker identity from sung speech and cross-modal recognition.

Purpose of the Study:

• To investigate the extent to which listeners can extract talker-specific information from sung speech.
• To examine talker identification accuracy across sung and spoken voice modalities.

Main Methods:

• 149 listeners were trained to recognize vocal identities in either sung or spoken voices.
• Participants were then tested on their ability to identify these same voices in both sung and spoken modalities.
• Voice recognition accuracy was compared between within-modality (sung-sung, spoken-spoken) and cross-modality (sung-spoken, spoken-sung) conditions.

Main Results:

• Initial learning of vocal identities was more efficient when trained with spoken voices compared to sung voices.
• Cross-modality voice recognition (identifying a speaker in song after hearing them in speech, or vice versa) was significantly above chance.
• Within-modality voice recognition accuracy was superior to cross-modality recognition in both sung and spoken conditions.

Conclusions:

• Talker-specific information is indeed accessible within the acoustic constraints of sung speech.
• While cross-modal voice recognition between singing and speech is feasible, it presents a greater challenge than recognizing voices within the same modality.
• These findings highlight the robustness of vocal identity cues despite significant acoustic alterations introduced by singing.