Receiver-Side Speech Enhancement for Noisy PSTN Audio Streams
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Public switched telephone network (PSTN) callers, particularly from cellphones, generate background noise in online meetings due to the lack of processing capabilities for background noise removal (BNR) at the sender side, which is not an option for PSTN callers, and the available processing budget is insufficient for conferencing services.
Innovation Solution
Implementing receiver-side noise reduction and speech enhancement techniques using an audio enhancement model trained with audio enhancements, which applies these enhancements to audio codec streams within a look-ahead buffer to generate an enhanced audio stream, leveraging machine learning and artificial intelligence to improve audio quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If background noise removal is implemented at the sender side, then noise reduction effectiveness is improved, but it is not feasible for PSTN callers due to lack of processing capabilities
Solution Approach 1:
The patent inverts the traditional noise removal approach by implementing background noise removal at the receiver side instead of the sender side. This allows PSTN callers to join meetings without requiring processing capabilities on their end, while still achieving effective noise reduction through server-side processing of the audio stream after it reaches the conference platform.
Solution Approach 2:
The patent introduces an intermediary processing layer at the conference server that receives audio streams from PSTN callers and applies background noise removal algorithms before distributing to other participants. This intermediary approach enables noise reduction for PSTN callers without modifying their devices or the PSTN infrastructure itself.
2Object-affected harmful factors
If background noise removal is implemented on PSTN servers, then noise reduction capability is improved, but the processing budget is insufficient to run BNR at the required scale
Solution Approach 1:
The patent implements a universal audio processing pipeline at the conference server that handles multiple functions including background noise removal, speech enhancement, and audio mixing in a unified manner. This multi-functional approach consolidates processing resources and avoids redundant computations that would otherwise be required if separate systems were used for each function.
Solution Approach 2:
The patent applies preliminary speech enhancement and noise reduction processing to audio streams from PSTN callers as they enter the conference system, before the audio is mixed and distributed to participants. This preliminary action reduces the overall processing burden on the conference server by pre-treating the audio streams, making subsequent mixing and distribution more computationally efficient.
3Manufacturing precision
If receiver-side audio enhancement is applied, then speech clarity is improved, but additional processing time and resources are required
Solution Approach 1:
The patent implements periodic processing of audio streams at the receiver side, where speech enhancement and noise reduction are applied in regular intervals rather than continuously. This periodic approach maintains speech clarity by processing audio frames at optimal intervals while reducing overall processing time and computational resource requirements compared to continuous real-time processing.
Data Source
AI summary
In one embodiment, an illustrative method herein may comprise: receiving, at a receiver, an audio codec stream; determining, by the receiver, a length of time associated with a look-ahead buffer of the audio codec stream; inputting, by the receiver, the audio codec stream into an audio enhancement model trained with one or more audio enhancements to cause the audio enhancement model to apply the one or more audio enhancements to the audio codec stream to generate an enhanced audio codec stream within the length of time associated with the look-ahead buffer of the audio codec stream; and outputting, by the receiver, the enhanced audio codec stream.