Evaluating speech-in-speech perception via a humanoid robot.

Area of Science:

• Auditory Neuroscience
• Human-Computer Interaction
• Robotics

Background:

• Investigating speech perception in noisy environments often requires lengthy psychophysical tests.
• Interactive social agents, like the NAO robot, could improve engagement during these tests.
• However, robots may have limitations in sound quality and processing speed.

Purpose of the Study:

• To evaluate the NAO robot's efficacy in presenting speech stimuli and collecting data for speech-in-speech perception tests.
• To compare the NAO robot's performance against a standard computer interface.
• To assess human-robot interaction and engagement during psychophysical testing.

Main Methods:

• Normal-hearing adults completed speech-in-speech perception tests using either a computer or a NAO robot.
• Stimuli included target sentences with keywords amidst competing speech at varying signal-to-noise ratios.
• Speech intelligibility, data collection time, and human-robot interaction (using NARS and backchannels) were measured.

Main Results:

• Speech intelligibility was comparable between the computer and NAO setups.
• Data collection took longer with the NAO robot.
• Participants exhibited higher engagement with the NAO robot, indicated by more backchannels, despite neutral attitudes towards robot social influence.

Conclusions:

• The NAO robot shows potential as a tool for presenting speech materials and gathering psychophysical data for speech-in-speech perception.
• While data collection duration is longer, increased engagement suggests benefits for certain testing scenarios.
• Further research can explore optimizing robot parameters for efficient and effective psychophysical testing.