Stereo Audio Artifact Suppression Using Mid-Channel Blending
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Stereo sound separation models introduce phase inconsistencies and spatial artifacts, causing audio sources to shift erratically between left and right channels, leading to an undesirable and jarring listening experience.
Innovation Solution
A method that leverages temporal dynamics of amplitude differences between stereo signals to identify and suppress spatial artifacts by calculating a confidence map and blending weights, combining left and right channels to produce stable output.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If mask-based stereo separation models are used to separate audio sources, then source separation efficiency is improved, but phase inconsistencies and spatial artifacts are introduced causing erratic source location shifts
Solution Approach 1:
The patent introduces a mid-channel spectrogram as an intermediary representation that captures the common spatial information between left and right channels. By processing this mid-channel representation and using it to guide the separation of individual audio sources, the system maintains spatial consistency while achieving effective source separation. The mid-channel acts as a mediator that prevents phase inconsistencies from causing erratic spatial shifts.
2Measurement precision
If independent mask estimation is performed for multiple stereo signals, then source separation accuracy is improved, but spatial artifacts and phase discrepancies increase
Solution Approach 1:
The patent merges the processing of left and right channel signals by first computing their average to obtain a mid-channel spectrogram. This merged representation is then used to guide the separation process for individual sources. By combining information from both channels before separation, the method maintains high source separation accuracy while reducing spatial artifacts and phase discrepancies that would result from completely independent processing.
3Productivity
If conventional mask-based separation is applied, then separation speed is maintained, but perceptual quality deteriorates due to jumping effects
Solution Approach 1:
The patent performs preliminary processing by computing the mid-channel spectrogram from the left and right channels before applying mask-based separation. This preliminary action establishes a stable spatial reference that guides the subsequent separation process. By preparing this spatial guidance information in advance, the method maintains separation speed while preventing the jumping effects and perceptual quality issues that arise from conventional approaches.
Data Source
AI summary
An audio application determines a left magnitude of the audio source (LSk) and a right magnitude of the audio source (RSk). The audio application determines an amplitude difference (Dk). The audio application calculates a temporal derivative d(Dk) of the Dk. The audio application determines an average of LSk and RSk to obtain a mid-channel spectrogram (MCSk). The audio application normalizes the MCSk to obtain a normalized value (Rk). The audio application divides d(DK) by Rk to obtain a confidence map. The audio application computes a blending weight by scaling and clipping the confidence map. The audio application combines the MCSk, the blending weight, and the Lt to obtain a left modified channel, and combining the MCSk, the blending weight, and the Rt to obtain a right modified channel.