Active Speaker Detection via Loudness Ratio in Conference Sessions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In audio and video conferencing, accurately determining active speakers is challenging due to background noises and hardware/software differences, making it difficult to mute or hide non-active users effectively.
Innovation Solution
A system that uses a time-decaying average and loudness ratio to sort users based on their audio activity, accounting for background noise and hardware/software variations, allowing for automatic determination of active users and subsequent muting or hiding of inactive users.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all users' audio and video are transmitted in a conference session, then complete information is provided to all users, but users experience distraction and impracticality from non-active users
Solution Approach 1:
The system extracts and identifies active speakers from the group of all users using audio level analysis, then selectively transmits only active users' audio and video streams to participants. This extraction principle removes non-active users from the transmission stream, reducing distraction while preserving information from active participants.
Solution Approach 2:
The system applies different quality levels of service to different users based on their activity status. Active users receive full-quality audio and video transmission, while inactive users have their streams muted or hidden. This local differentiation optimizes user experience without sacrificing necessary information.
2Measurement precision
If background noise and hardware differences are considered in determining active users, then measurement accuracy improves, but system complexity increases
Solution Approach 1:
The system changes the parameter of audio level measurement from absolute decibel values to relative loudness ratios. By comparing each user's audio level against a dynamic reference (initially the maximum level in the conference), the system adapts to different hardware configurations and background noise conditions without requiring complex calibration procedures.
Solution Approach 2:
The reference audio level is made dynamic rather than static. The system continuously updates the reference level based on the maximum audio level detected among all users, allowing the active user determination to adapt in real-time to changing conference conditions, including background noise variations and different microphone sensitivities.
Data Source
AI summary
The present disclosure relates to systems and methods for automatically determining the active users in an audio or video conference session including a plurality of users. In one implementation, the system may include a memory storing instructions and a processor configured to execute the instructions to receive packets from a device associated with each of a plurality of users, update a loudest level associated with each user if the received packet from the user reflects a communication louder than all of the previously received packets from the user, calculate a loudness ratio associated with each user based on the loudest level associated with each user and a corresponding time-decaying average, sort the plurality of users by a loudness score associated with each user, and mute audio or hide video received from a subset of the plurality of users in the conference session. The loudness score may be based on the loudness ratio associated with each user, and the muted or hidden subset may include one or more users for which the associated loudness score is lowest according to the sorting.