Packet Loss Concealment via Forced Waveform Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional packet loss concealment methods in digital communication systems often result in audible distortion due to destructive interference between extrapolated and normally-decoded waveforms after frame erasure or packet loss, as they fail to properly align the waveforms.
Innovation Solution
A method that generates an extrapolated waveform based on both the segment preceding and following the lost segment, using a first-pass periodic extrapolation and a second-pass periodic extrapolation guided by a calculated pitch contour to align the waveforms, thereby reducing or eliminating destructive interference.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional packet loss concealment methods are used, then the decoder can continue operating during packet loss, but audible distortion occurs due to destructive interference between extrapolated and normally-decoded waveforms
Solution Approach 1:
The patent performs preliminary waveform alignment by calculating a pitch contour based on the time lag between the first-pass extrapolated waveform and the decoded waveform. This pitch contour is then used in a second-pass extrapolation to pre-align the waveform phases before the actual waveform replacement, preventing destructive interference from occurring in the first place.
Solution Approach 2:
The patent changes the parameter of waveform phase alignment by using a pitch contour derived from time lag analysis. Instead of using a fixed or simple extrapolation method, the system dynamically adjusts the pitch parameters of the extrapolated waveform to match the decoded waveform's phase characteristics, thereby eliminating destructive interference.
2Device complexity
If waveform extrapolation is performed based only on the preceding waveform, then the extrapolation process is simple, but the extrapolated waveform becomes out of phase with the normally-decoded waveform after packet loss
Solution Approach 1:
The patent performs a preliminary analysis step before the main extrapolation process. It first calculates the time lag between the extrapolated and decoded waveforms, then derives a pitch contour from this time lag. This preliminary action provides the alignment information needed for the second-pass extrapolation to achieve precise phase matching.
Solution Approach 2:
The patent uses feedback from the time lag measurement to adjust the extrapolation process. The time lag between the first-pass extrapolated waveform and the decoded waveform is measured, and this feedback information is used to calculate a pitch contour that guides the second-pass extrapolation, ensuring the final waveform is properly aligned in phase.
3Stability of the object's composition
If overlap-add method is used to reduce waveform discontinuity, then waveform continuity is improved, but destructive interference between out-of-phase waveforms cannot be fixed
Solution Approach 1:
The patent performs preliminary waveform alignment before the overlap-add operation by calculating the pitch contour from time lag analysis. This ensures that when the overlap-add method is applied, the waveforms are already in phase, so the overlap-add operation only needs to handle continuity rather than fixing phase mismatches.
Solution Approach 2:
The system uses feedback from measuring the time lag between waveforms to adjust the pitch parameters of the extrapolated waveform. This feedback mechanism ensures that the waveform phase is corrected before the overlap-add operation, allowing the overlap-add method to work effectively for continuity without needing to compensate for phase mismatches.
Data Source
AI summary
A packet loss concealment method and system is described that attempts to reduce or eliminate destructive interference that can occur when an extrapolated waveform representing a lost segment of a speech or audio signal is merged with a good segment after a packet loss. This is achieved by guiding a waveform extrapolation that is performed to replace the bad segment using a waveform available in the first good segment or segments after the packet loss. In another aspect of the invention, a selection is made between a packet loss concealment method that performs the aforementioned guided waveform extrapolation and one that does not. The selection may be made responsive to determining whether the first good segment or segments after the packet loss are available and also to whether a segment preceding the lost segment and the first good segment following the lost segment are deemed voiced.


