Personalized Denoise Engine for Virtual Meeting Audio
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital communication systems, particularly in virtual meetings, face challenges in isolating and denoising voice content from ambient audio interference, leading to poor audio quality due to the presence of multiple speakers or suboptimal audio capture devices.
Innovation Solution
A Denoise Engine is implemented that collects and filters voice content segments from user accounts, generates personalized denoised audio by using speaker embeddings and a personalized denoise model, and bypasses the model when high-quality voice content is detected, ensuring clear audio output.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a denoise model is applied to all audio data, then audio quality is improved, but processing time and computational resources are increased
Solution Approach 1:
The system performs preliminary actions by collecting and storing voice content segments before the actual denoising operation. These pre-collected segments are used to generate speaker embeddings that characterize the user's voice pattern, enabling faster and more accurate denoising when needed without requiring reprocessing of the entire audio stream.
Solution Approach 2:
The system extracts only the necessary components for denoising by isolating voice content segments from the full audio data, generating speaker embeddings from these extracted segments, and using only these embeddings for the denoising operation. This extraction approach reduces the computational burden while maintaining denoising effectiveness.
2Measurement precision
If speaker embedding is generated from all voice segments, then denoising accuracy is improved, but data processing complexity is increased
Solution Approach 1:
The system segments the audio data into distinct voice content segments and non-voice segments. By segmenting the data, the system can process only the relevant voice portions for embedding generation, reducing the complexity of data processing while maintaining the accuracy needed for effective denoising.
Solution Approach 2:
The system applies local quality by generating speaker embeddings specifically from voice content segments rather than processing the entire audio stream uniformly. This localized approach focuses computational resources on the most critical portions of the data, improving denoising accuracy without proportionally increasing overall processing complexity.
3Quantity of substance
If ambient audio interference is not filtered, then audio content is preserved, but audio quality deteriorates due to noise
Solution Approach 1:
The system converts the harmful ambient audio interference into a beneficial feature by using the presence of multiple speakers in the audio data to identify and characterize the user's voice pattern more accurately. The interference from other speakers actually helps the system distinguish the user's voice characteristics, improving the effectiveness of the denoising operation.
Data Source
AI summary
Various embodiments of an apparatus, method(s), system(s) and computer program product(s) described herein are directed to a Denoise Engine. The Denoise Engine collects segments of voice content of a first user account from audio data associated with a virtual meeting. The audio data further includes additional types of audio content. The Denoise Engine identifies an audio embedding model. The Denoise Engine receives a speaker embedding generated by the audio embedding model. The speaker embedding based on the collected segments of voice content. The Denoise Engine generates personalized denoised voice content of the first user account for the virtual meeting by applying the speaker embedding to the audio data associated with a virtual meeting.