Highband Speech Regeneration Using Tonality-Dependent Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current artificial bandwidth extension methods for narrowband telephone speech often result in a metallic sound due to overly strong harmonics when regenerating wideband speech, as they fail to effectively address the limitations of narrowband audio transmission.
Innovation Solution
A method involving a system that includes up-sampling, whitening, highband excitation regeneration, and a filtering process to generate wideband speech, where tonality measures are used to filter out metallic artefacts by adjusting the energy and pitch of the signal blocks, employing a tonality-dependent gain factor to reduce signal energy and subtract pitch-delayed samples.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If spectral translation-based excitation regeneration is used to regenerate wideband speech from narrowband speech, then the high frequency band is restored, but the recovered signal becomes metallic due to overly strong harmonics
Solution Approach 1:
The patent modifies the excitation signal parameters by applying different filtering operations based on tonality measures. For high-tonality frames, a low-pass filter is applied to reduce high-frequency harmonics that cause metallic sound. For low-tonality frames, the original excitation is preserved. This dynamic parameter adjustment resolves the contradiction by selectively suppressing harmful harmonics while maintaining speech quality.
Solution Approach 2:
The patent introduces dynamic filtering where the filter characteristics change based on the tonality measure of each speech frame. The system adapts the filtering operation in real-time, switching between low-pass filtering and no filtering based on the detected tonality. This dynamic approach allows the system to eliminate metallic artifacts in appropriate contexts while preserving speech naturalness.
2Length of moving object
If narrowband speech is up-sampled and filtered to regenerate highband speech, then bandwidth is extended, but naturalness of speech is degraded due to strong harmonics
Solution Approach 1:
The patent dynamically changes the filtering parameters based on the tonality measure of each speech frame. When high tonality is detected, low-pass filtering is applied to attenuate the overly strong harmonics that cause unnatural metallic sound. When low tonality is detected, the original regenerated signal is kept. This parameter adaptation resolves the contradiction between bandwidth extension and naturalness preservation.
3Object-generated harmful factors
If tonality-dependent filtering is applied to reduce metallic artefacts, then naturalness is improved, but signal energy is reduced
Solution Approach 1:
The patent applies low-pass filtering only partially - specifically, only to high-tonality frames where metallic artifacts are present. For low-tonality frames, the original signal is preserved without filtering. This selective partial action removes harmful metallic artifacts while minimizing energy loss by avoiding unnecessary filtering of natural speech segments.
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
Provided is a method and system for processing a narrowband speech signal comprising speech samples in a first range of frequencies. The method comprises generating from the narrowband speech signal a highband speech signal in a second range of frequencies above the first range of frequencies; determining a pitch of the highband speech signal; using the pitch to generate a pitch-dependent tonality measure from samples of the highband speech signal; and filtering the speech samples using a gain factor derived from the tonality measure and selected to reduce the amplitude of harmonics in the highband speech signal.