FingerType: One-Handed Thumb-to-Finger Text Input Using 3D Hand Tracking.

Area of Science:

• Human-Computer Interaction
• Virtual Reality Input Methods
• Gesture Recognition

Background:

• Traditional text input methods in virtual reality (VR) can be cumbersome and limit immersion.
• One-handed input methods are desirable for VR to free up the other hand for interaction.
• Thumb-to-finger gestures offer a potential modality for intuitive one-handed VR input.

Purpose of the Study:

• To develop and evaluate FingerType, a one-handed text input system utilizing thumb-to-finger gestures for VR.
• To investigate the learnability and performance of FingerType compared to existing VR input methods.
• To analyze user comfort and error patterns associated with the proposed input technique.

Main Methods:

• Designed FingerType using Temporal Convolutional Networks (TCN) for tap event detection from 3D hand data.
• Developed an n-gram language model for decoding tap sequences into words.
• Incorporated user feedback on thumb-to-finger interactions and finger region comfort to refine a T9-style layout.
• Conducted a six-block user study comparing FingerType with controller and touch input, measuring Words Per Minute (WPM) and display attention.

Main Results:

• FingerType achieved 94.97% accuracy in tap detection and operated at 72 frames per second.
• User typing speed increased from 5.88 WPM to 10.63 WPM over the study blocks.
• FingerType enabled significantly more eyes-free typing (84.41% display attention) compared to touch input (69.47%).

Conclusions:

• FingerType demonstrates a viable and efficient one-handed text input method for VR environments.
• The system shows potential for improving user experience through increased typing speed and reduced visual attention requirements.
• Further improvements in gesture recognition accuracy are expected to enhance performance and competitiveness with established VR input techniques.