Command Recognition Using Binarized Convolutional Neural Network with Voice and Radar Sensors for Human-Vehicle...

Area of Science:

• Human-Computer Interaction
• Automotive Technology
• Sensor Fusion

Background:

• Increasing in-vehicle infotainment complexity diverts driver attention, posing safety risks.
• Effective human-vehicle interaction is crucial for mitigating driver distraction.
• Resource-constrained vehicle environments necessitate efficient sensing and processing.

Purpose of the Study:

• To propose a novel human-vehicle interaction system to minimize driver distraction.
• To enhance the reliability of driver monitoring through sensor fusion.
• To reduce computational load for real-time command classification.

Main Methods:

• Utilized low-complexity voice and continuous-wave radar sensors.
• Implemented sensor fusion techniques to combine data from multiple sensors.
• Employed a binarized convolutional neural network for efficient command classification.

Main Results:

• Achieved a recognition accuracy of 96.4% in noisy environments.
• Demonstrated a 7.6% accuracy improvement over voice-only systems.
• Showed a 9.0% accuracy improvement over radar-only systems.

Conclusions:

• The proposed system effectively reduces driver distraction by integrating voice and radar sensing.
• Sensor fusion and a binarized neural network offer a robust and efficient solution for in-vehicle interaction.
• This approach enhances driving safety in complex automotive environments.