Adaptive Codebook Filtering for Speech Coding Distortion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech coding techniques, such as CELP, face limitations in improving adaptive codebook performance and decoded speech quality due to distortion from unnecessary components and mismatched frequency bands, despite advancements in fixed excitations and spectrum envelope coding.
Innovation Solution
A speech coding apparatus and method that incorporates an excitation search section for adaptive and fixed excitations, with filtering and clipping processes to enhance adaptive codebook performance by compensating for lag-related deterioration, using a filtering section to process adaptive excitations and improve their similarity to ideal excitations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If adaptive codebook search is performed using raw adaptive excitations from the codebook, then the search process is simple and fast, but the synthesized speech signal contains distortion from unnecessary frequency components
Solution Approach 1:
The patent applies preliminary filtering to the adaptive excitations before they are used in the codebook search process. A band-limiting filter is applied to remove unnecessary frequency components from the adaptive excitations prior to the search, preventing distortion from being introduced into the synthesized speech signal while maintaining coding efficiency
2Object-affected harmful factors
If the frequency band of adaptive excitations is limited using a filter, then distortion from unnecessary components is suppressed, but the adaptive excitations may lose important speech components
Solution Approach 1:
The patent dynamically adjusts the filtering characteristics based on the actual speech signal properties. The band-limiting filter is applied adaptively, with its characteristics determined by analyzing the input speech signal's frequency content, ensuring that important speech components are preserved while removing only the unnecessary components that cause distortion
Solution Approach 2:
The patent changes the frequency domain parameters of the adaptive excitations by applying a band-limiting filter with specifically designed cutoff frequencies. This parameter transformation removes harmful high-frequency components while preserving the essential speech information in the lower frequency bands, resolving the contradiction between distortion suppression and speech quality maintenance
3Productivity
If conventional CELP coding methods are used, then coding efficiency is achieved, but adaptive codebook performance remains limited and decoded speech quality cannot be further improved
Solution Approach 1:
The patent improves the local quality of adaptive excitations by applying targeted band-limiting filtering specifically to the adaptive codebook excitations. This localized processing enhances the quality of excitations used for voiced speech segments without affecting other parts of the coding system, thereby improving overall decoded speech quality while maintaining coding efficiency
Data Source
AI summary
Provided is an audio encoding device capable of improving performance of an adaptive codebook and improving quality of a decoded audio. In this audio encoding device, an adaptive codebook cuts out a vector specified by a comparator from adaptive code vectors stored in an internal buffer and outputs it to a filter and a switch. The filter performs a predetermined filtering process on the adaptive sound source signal and outputs the obtained adaptive code vector to the switch. According to an instruction from the comparator, the switch outputs the adaptive code vector directly output from the adaptive codebook to a adjuster when the adaptive codebook is searched and outputs the adaptive code vector output from the filter after being subjected to the filtering process to the gain adjuster when a fixed sound source is searched after the adaptive sound source search.