Page 460 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

(continued)

Items Details
Model Training and Past speaker behaviour patterns to predict mic abuse or speech
Fine-Tuning quality.
Neural networks + DSP ( Audio classification e.g., noise vs. speech).
Natural Language Processing(NLP)for understanding content of
speech (for moderation).

Queue optimization using reinforcement learning or priority policies.

Testbeds or Pilot In academic and public service environments to validate its effective-
Deployments ness and usability. Also local town hall meetings, where the system was
integrated with public address (PA) systems to facilitate transparent and
structured civic engagement.

2 Use Case Description

2�1 Description

Introduction: In many professional and academic settings, such as conferences, meetings,
and large gatherings, managing real-time communication and ensuring clear audio for all
participants is a significant challenge. Traditional microphone systems are often limited,
immobile, or insufficient to meet the needs of large groups, resulting in inefficiencies, delays, and
potential disruptions. Participants may struggle to access microphones, leading to frustration
and reduced engagement. Additionally, the responsibility for managing these systems often
falls to IT departments, who may lack the necessary expertise in audio management. With the
rise of smartphones and wireless communication, there is an opportunity to utilize existing
technology to provide a more seamless and efficient communication system.

Solution Overview: The solution integrates AI with a smartphone app and a PC-based server,
creating a fully automated speaker management system. When a user requests to speak, the
server communicates with the AI module, which decides whether to grant access or add the
user to a queue based on factors like request order, user roles, and speaking history. The
system also continuously processes the audio stream, applying noise reduction and ensuring
compatibility with the PA system. This innovative approach leverages real-time multimedia
applications (RTMA), Wi-Fi connectivity, and AI decision-making, automating what was once a
manual process and making communication more efficient and accessible.

The expected impact includes increased efficiency by automating speaker management,
enhanced user experience with smartphones as microphones, and reduced costs by eliminating
the need for fixed audio equipment. This ensures smoother communication, improves
productivity, and maintains audience engagement.

Partner

Mbeya University of Science and Technology (MUST)

424