Selective Audio Stream Inverse MDCT Decoding for Server Load Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional conference servers face a computational burden in fully decoding multiple audio streams during large multimedia conferences, as they often perform inverse MDCT on all streams, even though only a fraction is needed for mixing, and many systems lack support for RTP header extensions that facilitate selective decoding.
Innovation Solution
A collaboration server estimates audio energy of MDCT frequency domain streams, selectively performing inverse MDCT only on streams with sufficient energy, and transmitting others in their compressed form, using transform determination logic to minimize processing and be compatible with non-RTP streams and equipment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If the server fully decodes all audio streams using inverse MDCT, then the audio mixing quality is improved, but the computational burden increases significantly
Solution Approach 1:
The server performs inverse MDCT decoding only on audio streams that contain actual speech content, rather than decoding all received streams. By estimating audio energy and selectively decoding only necessary streams, the system achieves adequate mixing quality while significantly reducing computational burden.
Solution Approach 2:
Different processing approaches are applied to different audio streams based on their individual characteristics. Streams with sufficient audio energy undergo full inverse MDCT decoding, while low-energy streams are handled differently, optimizing both quality and computational efficiency for each stream locally.
2Reliability
If the server performs inverse MDCT on all audio streams, then complete audio reconstruction is achieved, but the processing time increases
Solution Approach 1:
The server performs audio energy estimation on incoming streams before deciding whether to perform the computationally intensive inverse MDCT decoding. This preliminary assessment allows the system to avoid unnecessary decoding operations, reducing processing time while maintaining reconstruction completeness only when needed.
Solution Approach 2:
Instead of universally applying full inverse MDCT decoding to all streams, the system applies partial decoding only to streams that require it based on energy thresholds, thereby reducing overall processing time while maintaining reliability for relevant audio content.
3Loss of information
If the server processes all audio streams with full decoding, then no audio information is lost, but the energy consumption increases
Solution Approach 1:
The server performs full inverse MDCT decoding only on a subset of audio streams that contain meaningful speech content, as determined by energy estimation. This selective approach retains all necessary audio information while significantly reducing the energy consumption associated with processing every incoming stream.
Solution Approach 2:
The system extracts and processes only the relevant audio streams that contain speech content, separating them from low-energy background noise streams. This extraction approach ensures no useful audio information is lost while minimizing energy consumption by excluding unnecessary processing.
Data Source
AI summary
A server receives, from each of a plurality of participant devices in a communication session, a respective one of a plurality of audio streams. The server estimates an audio energy of each of the plurality of audio streams and determines whether to perform a transform on at least one of the plurality of audio streams. If so, the server performs the transform on the at least one of the plurality of audio streams and transmits the at least one of the plurality of audio streams to at least one of the plurality of participant devices.