Decoder State Resynchronization After Audio Packet Loss Concealment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing packet loss concealment methods are inadequate for sub-band predictive coders like the ITU-T G.722 wideband speech coder, particularly due to structural issues and challenges posed by Adaptive Differential Pulse Code Modulation (ADPCM) predictive coders, which are not addressed by prior-art techniques.
Innovation Solution
A method and system for updating the decoder state in a predictive coding system by re-encoding the synthesized audio signal associated with a lost frame using a simplified encoder, and updating the decoder state based on the generated encoder state to address sub-band-specific architectural issues and backward-adaptive ADPCM challenges.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If prior-art packet loss concealment methods are applied to sub-band predictive coders, then packet loss effects can be concealed, but sub-band-specific structural issues and backward-adaptive ADPCM challenges are not addressed, resulting in poor audio quality
Solution Approach 1:
The patent applies local quality by treating each sub-band independently with sub-band-specific packet loss concealment processing. The decoder state update mechanism is specifically adapted for each sub-band's ADPCM structure, applying localized corrections to predictor coefficients and quantizer step sizes rather than uniform processing across all bands.
Solution Approach 2:
The patent changes parameters by dynamically updating decoder internal states (predictor coefficients, quantizer step sizes) to match encoder states after packet loss. The method adjusts these parameters based on re-encoded synthesized audio, ensuring the decoder operates with correct state values that adapt to the actual signal characteristics.
2Measurement precision
If a full encoder is used to re-encode synthesized audio for decoder state update, then accurate state alignment is achieved, but computational complexity and processing time increase significantly
Solution Approach 1:
The patent extracts only the necessary encoder components needed for state update, rather than implementing a complete encoder. Specifically, only the sub-band analysis filter bank and ADPCM encoding logic are used to generate state information, while other encoder functions are omitted. This extraction reduces complexity while maintaining the essential state alignment capability.
Solution Approach 2:
The patent performs preliminary action by pre-calculating and storing encoder state information during normal operation. When packet loss occurs, the system uses these pre-computed states rather than performing full re-encoding, significantly reducing the computational burden during loss concealment while maintaining accuracy.
3Productivity
If decoder state is not updated after packet loss concealment, then processing is simpler, but internal state differences between encoding and decoding accumulate, causing audio quality degradation
Solution Approach 1:
The patent implements feedback by using the synthesized concealed audio as input to the simplified encoder, then using the resulting encoder state to update the decoder state. This closed-loop feedback mechanism ensures the decoder state continuously aligns with what the encoder would produce, maintaining long-term synchronization and audio quality despite periodic packet losses.
Solution Approach 2:
The patent applies self-service by having the decoder update its own state using information generated from its own synthesized output. The decoder synthesizes concealed audio, feeds it through a simplified encoder model, and uses the resulting state to correct its internal registers, making the system self-correcting without external intervention.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A technique is described herein for updating a state of a decoder in a predictive coding system after synthesizing an audio output signal corresponding to a lost frame in a series of frames representing an encoded audio signal. In accordance with the technique, an audio signal associated with the synthesized output audio signal is re-encoded in an encoder to generate an encoder state, wherein the encoder is simplified with respect to an encoder used to generate the encoded audio signal. The state of the decoder is then updated based on the generated encoder state.