Scalable Decoder Gain Attenuation for Speech Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In scalable speech coding, when the enhancement layer signal is lost, the interpolated gain coefficients based on past information can lead to disagreeable sound and subjective quality deterioration due to uneven power changes in the core and enhancement layer decoded speech signals.
Innovation Solution
A scalable decoding apparatus that detects changes in the power of the core layer decoded signal and variably sets the attenuation rate for the enhancement layer gain coefficients, interpolating them for a specific interval based on the detected power changes to maintain consistent quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If enhancement layer gain coefficients are interpolated using monotone decrease function based on past information, then the decoded speech can be maintained during signal loss, but the subjective quality deteriorates due to disagreeable sound and fluctuation in signal power
Solution Approach 1:
The patent applies dynamics by making the attenuation rate variable rather than fixed. The attenuation rate is dynamically adjusted based on the power of the core layer decoded speech signal, allowing the system to adapt to changing signal conditions. This resolves the contradiction by enabling the system to maintain decoded speech continuity while avoiding disagreeable sound through adaptive control of the attenuation characteristics.
Solution Approach 2:
The patent changes the parameter of attenuation rate from a constant value to a variable value that depends on core layer power. By modifying this key parameter based on real-time signal conditions, the system can maintain reliability during signal loss while preventing subjective quality deterioration. The attenuation rate is set to 1/(1+α) where α is determined by core layer power, creating a dynamic parameter adjustment mechanism.
2Reliability
If the gain of enhancement layer decoded speech signal is attenuated at extremely slow pace, then the interpolated speech can be maintained, but the enhancement layer decoded speech deteriorates and becomes obvious
Solution Approach 1:
The system dynamically adjusts the attenuation rate based on core layer power to balance speech maintenance and quality. When core layer power is low, a slower attenuation rate is applied to maintain interpolated speech. When core layer power is high, a faster attenuation rate prevents enhancement layer deterioration from becoming obvious. This dynamic adjustment resolves the contradiction between maintaining speech and preventing deterioration.
Solution Approach 2:
The system uses feedback from the core layer decoded speech power to control the attenuation rate of enhancement layer gain coefficients. The power measurement of the core layer signal provides feedback that determines the appropriate attenuation rate, creating a closed-loop control system that automatically adjusts to prevent deterioration while maintaining speech continuity.
3Adaptability or versatility
If the power of core layer decoded speech signal changes dramatically, then the signal adaptation to loss conditions is achieved, but band change is perceived by listeners
Solution Approach 1:
The system dynamically controls the attenuation rate based on core layer power to manage band change perception. When core layer power changes dramatically, the attenuation rate is adjusted to compensate for the power variation, preventing perceived band changes. This dynamic control allows the system to adapt to loss conditions while maintaining perceptual consistency.
Data Source
AI summary
A scalable decoder capable of avoiding deterioration in subjective quality of a listener. The scalable decoder for decoding core layer encoding data and extension layer encoding data including an extension layer gain coefficient, wherein a voice analysis section detects variation in power of a core layer decoding voice signal being obtained from the core layer encoding data, a gain attenuation rate calculating section (140) sets the attenuation intensity variable depending on variation in power, and a gain attenuation section (143) attenuates the extension layer gain coefficient in a second period preceding a first period according to a set attenuation intensity when extension layer encoding data in the first period is missing, thus interpolating the extension layer gain coefficient in the first period.