Front Row: Automatically Generating Immersive Audio Representations of Tennis Broadcasts for Blind Viewers

Visual Impairment Technologies (Screen Readers, Tactile Graphics, Braille)Accessible GamingMental Health Apps & Online Support CommunitiesPhysicians, Nurses & CliniciansEsports Players & Live Streamers

Document Title

Front Row: Automatically Generating Immersive Audio Representations of Tennis Broadcasts for Blind Viewers

Document Information

  • Domain: Human-Computer Interaction, Computer Vision, Accessibility Design
  • Keywords: Visual impairment, sports, accessibility technology, computer vision, immersive audio, blind user experience, sports broadcasting, accessible video

Research Background and Problem

  • Identified Challenges:

    • Individuals with visual impairments struggle to fully comprehend sports content through conventional television or radio media; existing text or voice descriptions fail to provide spatial information.
    • Visually impaired individuals rely on explanations from spectators or commentators, which limits their ability to independently understand the match.
    • Current technologies (e.g., high-precision cameras and sensors) rely on expensive hardware and are applicable only to select sports events.
  • Significance: Sports broadcasting is an important way for visually impaired individuals to integrate into society and culture, yet its inaccessibility leaves many feeling isolated or excluded.

  • Research Motivation and Related Work:

    • Previous studies have explored tactile graphics and sound effect technologies, but these methods are unsuitable for rapidly changing scenes like sports videos.
    • Systems like Action Audio depend on hardware installations at venues, whereas Front Row aims to generate audio representations directly from video, enhancing the technology's universal applicability.

Solution

  • Proposed Approach:

    • Front Row System: Extracts player positions and actions from tennis match videos using computer vision and presents match details in spatialized (3D) audio format.
    • The system features three primary audio cues: player positions (continuous low-frequency sound), player actions (bell tones distinguishing forehand and backhand), and ambient game sounds (e.g., audience cheers).
  • Innovations:

    • Generates audio representations directly from video, eliminating hardware dependency and enabling accessibility for all sports broadcasts.
    • Unprecedented integration of spatialized audio technology to provide visually impaired users with real-time dynamics and positional understanding.
    • Empowers visually impaired users to independently perceive match content and analyze player strategies.
  • Implementation Steps and Key Technologies:

    1. Computer Vision Processing:
      • Detects the court, ball, and player positions using deep learning models to track the ball and players.
      • Identifies hitting moments through trajectory changes and detects hitting types (forehand/backhand) from frames around the player.
      • Uses Support Vector Machines to identify "active match moments" in the video (e.g., ignoring rendering during side changes).
    2. Audio Design:
      • Utilizes Steam Audio tools to create 3D audio scenes, rendering the court as a two-dimensional plane in front of the user.
      • Integrates ambient audio (e.g., audience sounds) with generated sound effects to enhance realism for visually impaired users.
    3. User-Customized Experience:
      • Provides intuitive spatial position cues and adjustable audio modes tailored to users with varying levels of technical expertise.

Research Outcomes

  • Specific Results:

    • Front Row significantly improved users' understanding of match content, reducing error rates: for instance, identifying hitting types saw a 90% reduction in error compared to television.
    • Enhanced immersion: users reported that Front Row enabled them to independently analyze matches and form insights into player mindset and strategies.
    • Overall user experience surpassed radio and television, particularly in spatial understanding and accuracy.
  • Advantages Comparison:

    • Compared to television: Front Row offers real-time content interpretation capabilities.
    • Compared to radio: Front Row avoids commentator bias, providing more neutral and accurate match information.
  • Experiment and Evaluation Results:

    • Technical tests showed an average accuracy of 97.2% in player position tracking.
    • User studies revealed that 8 out of 12 visually impaired participants ranked Front Row as their preferred match-viewing experience.
  • Limitations and Future Directions:

    • Limitations: The continuous "low-frequency sound" in Front Row's audio design was perceived by some users as lacking realism; the system does not fully cover complex events (e.g., multiplayer sports).
    • Future Directions:
      • Develop multimodal feedback (e.g., tactile + audio) to convey details like hitting force.
      • Support representation of different player numbers and actions in more complex scenarios (e.g., soccer or basketball).
      • Expand accessibility broadcasting to more sports (e.g., badminton, table tennis).

Quick Actions

Share

Share this page

ios_share

https://hci.top/en/papers/uist/126662/2023

AdRecommended

Learn AI Coding at CodeNow

open_in_newOpen DOI Link
DOI: https://doi.org/10.1145/3586183.3606830
At a Glance

Paper Snapshot

fact_check
dataset
Source
UIST
calendar_month
Year
2023
emoji_events
Award
No award tagged
group
Authors
7 authors
sell
Subtopics
Visual Impairment Technologies (Screen Readers, Tactile Graphics, Braille), Accessible Gaming, Mental Health Apps & Online Support Communities
work
Professions
Physicians, Nurses & Clinicians, Esports Players & Live Streamers
article
Content Status
Full text indexed
hub
Related Papers
0 related papers