Scalable Speech Coding Error-Band Search Across Frequency Layers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing scalable coding schemes face challenges in accurately specifying error bands across full frequency bands with high computational complexity, particularly when the number of layers increases, which affects speech quality and efficiency in mobile communication systems.
Innovation Solution
The proposed solution involves a configuration with a first layer encoding and decoding section, and a second layer encoding and decoding section that searches for error transform coefficients across wider and narrower bandwidths to identify and encode target frequency bands with maximum errors, using step sizes that minimize computational complexity while improving sound quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a fine search is performed across the full band to accurately specify error bands, then measurement precision is improved, but device complexity increases due to the large number of calculations required
Solution Approach 1:
The full frequency band is divided into multiple subbands. The error band specification is performed separately in each subband rather than across the entire band at once. This segmentation reduces the computational load in each step while maintaining overall accuracy, as each subband requires fewer calculations than the full band would require
Solution Approach 2:
The search process is organized into two dimensional stages: first searching across subbands to identify which subbands contain error bands, then performing a finer search within those identified subbands. This two-dimensional approach reduces the total number of calculations compared to a single exhaustive search across the full band
2Manufacturing precision
If the number of coding layers is increased to improve speech quality and fidelity, then manufacturing precision is improved, but device complexity increases
Solution Approach 1:
The speech signal is divided into multiple coding layers, with each layer handling different aspects of the signal. The first layer performs coarse encoding while subsequent layers refine the representation. This segmentation allows high-quality reconstruction through multiple passes rather than requiring a single complex encoder
Solution Approach 2:
The coding layers are structured hierarchically where each layer builds upon and refines the previous layer's output. The error bands identified in outer layers guide the focus of inner layers, creating a nested structure where computational effort is concentrated on the most significant error regions at each level
3Measurement precision
If a wide bandwidth is searched with fine step size to accurately identify error bands, then measurement precision is improved, but productivity decreases due to increased computational time
Solution Approach 1:
The frequency band is segmented into subbands, and the search is performed in stages. A coarse search with larger steps identifies candidate subbands first, then a fine search with smaller steps is performed only within those candidate regions. This segmentation strategy maintains accuracy while reducing total computational time
Solution Approach 2:
A preliminary coarse search is performed across the full band using wider steps to identify candidate subbands that contain error bands. This preliminary action filters out regions without errors, so the subsequent fine search with smaller steps needs to be performed only in relevant regions, significantly reducing total computation time
Data Source
AI summary
Disclosed is an encoding device which can accurately specify a band having a large error among all the bands by using a small calculation amount. A first position identifier uses a first layer error conversion coefficient indicating an error of a decoding signal for an input signal so as to search for a band having a large error in a relatively wide bandwidth in all the bands of the input signal and generates first position information indicating the identified band. A second position identifier searches for a target frequency band having a large error in a relatively narrow bandwidth in the band identified by the first position identifier and generates second position information indicating the identified target frequency band. An encoder encodes a first layer decoding error conversion coefficient contained in the target frequency band.