Acceptability of Speech and Silent Speech Input Methods in Private and Public

Title of the Paper

Acceptability of Speech and Silent Speech Input Methods in Private and Public

Bibliographic Information

• Field of Study: Human-Computer Interaction, Speech Input, Privacy and Security
• Keywords: Silent Speech Input, Speech Input, Social Acceptability, Human-Computer Interaction, Voice Assistants, Contactless Interaction

Research Background and Issues

• Identified Problems or Challenges:
  • Although speech input has become a significant mode of human-computer interaction for mobile devices, it faces notable challenges regarding privacy and security in public settings.
  • Silent speech input (converting lip and tongue movements into text) is considered a potential alternative to speech input, but research on user acceptability is still limited.
• Importance:
  • Enhancing the social acceptability of input methods in public settings can improve user experience and technology adoption while safeguarding privacy and security.
• Research Motivation and Related Work:
  • Silent speech input technology has been preliminarily demonstrated to have application potential, especially for individuals with speech impairments or in noisy environments. However, issues related to social adaptability, error tolerance, and feedback design remain underexplored.

Solution

• Methods or Solutions:
  • Conduct three user studies: social acceptability evaluation, error tolerance analysis, and feedback mechanism comparison.
• Innovations:
  • Evaluate the social acceptability of the two input methods from both the user's and observers' perspectives.
  • Introduce a "Wizard of Oz" experiment to study user tolerance for errors.
  • Compare user acceptability of video feedback versus abstract feedback, with a comprehensive analysis of privacy and security.
• Implementation Steps and Key Techniques:
  1. Social Acceptability Test:
     • Online survey showcasing videos of silent speech and speech input methods.
     • Assess users' comfort levels in different scenarios (e.g., at home, in a restaurant, in an office).
  2. Error Tolerance Experiment:
     • Use the "Wizard of Oz" method to simulate different error recognition rates (0%, 5%, 10%, 15%, 20%) and evaluate user tolerance levels.
  3. Feedback Mechanism Experiment:
     • Compare video feedback (real-time display of camera-captured videos) and abstract feedback (display of gray dots or flashing red dots) in terms of user experience, including privacy, security, and effectiveness.

Research Findings

• Specific Findings:
  • Users generally find silent speech input more acceptable, perceiving it as offering higher privacy and security, especially in public settings.
  • When the error recognition rate is between 10%-15%, users show higher tolerance for silent speech input and remain willing to use it; however, tolerance for speech input is lower.
  • Users believe video feedback provides more detailed information, but abstract feedback emphasizes privacy and security while not interfering with the user experience.
• Comparison with Existing Solutions:
  • Silent speech input is more suitable than speech input for both public and private scenarios, particularly for users with high privacy and security requirements.
  • Abstract feedback demonstrates a low-intrusion advantage, making users feel more comfortable compared to video feedback.
• Experimental or Evaluation Results:
  • Users tend to prefer silent speech input, but its broad appeal is contingent on maintaining a reasonable error rate (5%-10%).
  • Abstract feedback significantly enhances users' positive perceptions of privacy and security without compromising effectiveness.
• Limitations and Future Directions:
  • The findings are limited by the geographical and cultural diversity of the sample, and may not be applicable to all cultural and contextual backgrounds.
  • The "Wizard of Oz" method lacks exploration of technical challenges such as latency in real-world development environments.
  • Future research should focus on developing silent speech input applications for real mobile devices and conducting long-term user studies to validate its broad applicability and the impact on user habits.