Page 462 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

approve, reject, or queue them based on predefined rules. This enables fair, organized,
and efficient turn-taking.
• REQ-04: To ensure clear and professional audio quality, even in noisy environments,
the AI-driven system must apply noise reduction, echo cancellation, and voice activity
detection before sending audio to the PA system.
• REQ-05: The system must handle many users at once. The server must be able to track
active, waiting, and muted users, and switch audio sources seamlessly between speakers
to avoid overlaps, dead air, or audio collisions.

4 Sequence Diagram

5 Reference

[1] A. V. Oppenheim and R. W. Schafer, "Discrete-Time Signal Processing," 3rd ed., Pearson,
2009.
[2] D. Jurafsky and J. H. Martin, "Speech and Language Processing," 3rd ed. (draft), 2023.
[Online]. Available: https:// web .stanford .edu/ ~jurafsky/ slp3/
[3] R. S. Sutton and A. G. Barto, "Reinforcement Learning: An Introduction," 2nd ed., MIT
Press, 2018.

426