Postfilter Control for Speech Quality via Stationarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing postfilter algorithms in speech and audio coding do not adequately account for the varying audibility of quantization noise based on speech signal dynamics, failing to optimize noise suppression in high information content and steady-state modes.
Innovation Solution
A modified postfilter control mechanism that adjusts its parameters based on signal stationarity, measured by spectral dynamics, to adapt noise suppression accordingly, reducing noise suppression during high dynamics and increasing it during low dynamics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing postfilter algorithms adapt to formant and pitch structures on a frame-basis, then the perceived quality of reconstructed speech is improved, but the algorithms fail to account for varying audibility of quantization noise based on speech signal dynamics
Solution Approach 1:
The postfilter control parameter is made dynamic by introducing a stability factor that adapts the filter behavior based on the spectral dynamics of the decoded speech signal. The stability factor varies over time to reflect changes in signal stationarity, allowing the postfilter to adjust its noise suppression aggressiveness dynamically rather than using fixed frame-based adaptation.
Solution Approach 2:
The invention changes the control parameter of the postfilter from a static or frame-based value to a time-varying parameter modulated by the stability factor. This parameter change enables the postfilter to respond to rapid spectral changes in the speech signal, improving its ability to mask quantization noise during both stationary and transient speech segments.
2Object-affected harmful factors
If noise suppression is increased during steady-state modes, then quantization noise becomes less audible, but noise suppression should be reduced during high information content modes to preserve speech quality
Solution Approach 1:
The stability factor is computed based on the spectral dynamics of the decoded speech signal, creating a feedback mechanism that continuously monitors signal characteristics and adjusts the postfilter control parameter accordingly. This feedback loop ensures that noise suppression is increased during steady-state modes where quantization noise is more audible, while being reduced during transient modes where the speech signal itself provides natural masking.
3Measurement precision
If postfilter parameters are adapted to in-frame features like pitch period and autoregressive coefficients, then signal masking is exploited effectively, but the algorithms assume speech is stationary for the current frame which limits responsiveness to rapid spectral changes
Solution Approach 1:
The stability factor is computed using a simplified measure of spectral dynamics that does not require full frame-based stationary assumptions. By calculating the stability factor from recent signal characteristics and applying it to the current frame, the system prepares the postfilter in advance for the actual speech content, enabling faster response to spectral changes without the computational burden of complete re-adaptation for each frame.
Data Source
AI summary
The present invention relates to a postfilter and a postfilter control to be associated with a postfilter for improving perceived quality of speech reconstructed at a speech decoder. The postfilter control comprises means for measuring stationarity of a speech signal reconstructed at a decoder, means for determining a coefficient to a postfilter control parameter based on the measured stationarity, and means for transmitting the determined coefficient to a postfilter, such that the postfilter can process the reconstructed speech signal by applying the determined coefficient to the postfilter control parameter to obtain an enhanced speech signal.