LPC Decoder Missing Packet Reconstruction Using Previous Frame Extrapolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Missing packets in digitized speech transmission over packetized networks cause discontinuities and degraded speech quality due to popping or clicking noises, especially when voice frames are dropped, leading to sharp edges in the waveform.
Innovation Solution
A low complexity, no delay reconstruction method for Linear Predictive Coding (LPC) decoders that uses mathematical extrapolation based on previous good or reconstructed frames, categorizing speech status to adaptively generate missing frames, relying solely on previous frames without introducing additional delay.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If missing packets are not reconstructed, then device complexity is low, but speech quality deteriorates with popping and clicking noises
Solution Approach 1:
The patent performs preliminary classification of the current frame into voiced, unvoiced, or silence categories before reconstruction is needed. This pre-categorization enables the use of simple extrapolation methods during packet loss without requiring complex real-time analysis, thus improving speech quality while maintaining low computational complexity during the reconstruction process
Solution Approach 2:
The patent reconstructs missing frames by copying and extrapolating features from previously received good frames. Instead of performing complex reconstruction algorithms, it uses the historical data from previous frames to generate replacement frames, significantly reducing computational complexity while maintaining acceptable speech quality
2Manufacturing precision
If future frames are used for reconstruction, then speech quality improves, but system delay increases
Solution Approach 1:
The patent performs frame classification in advance, before packet loss occurs. This preliminary action stores the category information (voiced, unvoiced, silence) of previous frames, enabling the reconstruction process to use only historical data without needing to wait for or analyze future frames, thus achieving accurate reconstruction without introducing delay
Solution Approach 2:
Instead of using future frames to reconstruct current lost frames (which would cause delay), the patent inverts the approach by using only past and present information. It extrapolates backwards from previously classified frames to reconstruct current missing packets, eliminating the need for future frame data and thus avoiding system delay
3Manufacturing precision
If complex reconstruction algorithms are used, then speech quality improves, but computational complexity increases
Solution Approach 1:
The patent performs computationally intensive frame classification (distinguishing voiced, unvoiced, and silence frames) in advance when network conditions are good. This preliminary classification stores essential speech characteristics that can be directly used during packet loss without requiring complex real-time computation, thus improving synthesis quality while keeping runtime complexity low
Solution Approach 2:
The patent uses simple extrapolation methods that rely on previously stored frame categories rather than performing complex reconstruction algorithms during packet loss. This approach uses cheap, computationally light operations (copying and basic signal processing) instead of expensive complex algorithms, achieving acceptable synthesis quality with minimal computational effort during the reconstruction phase
Data Source
AI summary
Lost frame reconstruction is described. A previous good or reconstructed frame may be analyzed to determine a category for the lost frame. A percentage Pi may be associated with the determined category of the lost frame. A top Pi percent magnitude samples may be zeroed out in an excitation of the previous good or reconstructed frame to produce a reconstruction excitation. The reconstruction excitation may be applied to one or more linear prediction coefficients for the previous good or reconstructed frame to generate a reconstructed frame.