Using voice recognition to measure trust during interactions with automated vehicles

Area of Science:

• Human-computer interaction
• Automotive engineering
• Psychology

Background:

• Driver trust is crucial for automated vehicle (AV) adoption and safety.
• Understanding factors influencing trust, such as system feedback, is essential.
• Speech patterns offer a potential, non-intrusive method for assessing driver trust.

Purpose of the Study:

• To investigate the impact of varying automated vehicle system (AVS) performance and feedback on driver trust.
• To evaluate the efficacy of using speech features to measure driver trust in AVs.
• To develop a voice-based method for accurately classifying driver trust states.

Main Methods:

• Seventy-five participants were assigned to high-trust (100% correct, 0 crashes, visual-auditory feedback) or low-trust (60% correct, 40% crash rate, visual-only feedback) AV conditions.
• Speech data was collected during driving tasks using voice interaction.
• Extracted speech features were used to train a back-propagation neural network for trust classification.

Main Results:

• Experimental conditions successfully induced distinct high-trust and low-trust states in participants.
• Speech feature analysis enabled accurate prediction of driver trust levels.
• The highest classification accuracy for predicting trust reached 90.80%.

Conclusions:

• Speech analysis is a viable and accurate method for measuring driver trust in automated vehicles.
• Voice recognition technology can be leveraged to monitor and potentially manage driver trust in AVs.
• This research contributes to enhancing the safety and user experience of automated driving systems.