Wideband Speech Coding via Highband Spectral Extrapolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wideband speech coding techniques face challenges in efficiently extending narrowband speech coding to support higher frequency ranges without significant computational overhead or bandwidth increase, particularly in mobile and embedded applications, and require transcoding for compatibility with narrowband systems.
Innovation Solution
A method and apparatus that extend narrowband speech coding to support wideband speech signals by filtering the signal into narrowband and highband components, encoding each separately, and combining them for transmission, allowing for efficient allocation of bits between channels without requiring transcoding, using a filter bank to split the signal into subbands with reduced sampling rates and overlapping frequency ranges to reduce computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If narrowband speech coding technique is scaled to cover wideband spectrum with higher sampling rate and more filter coefficients, then wideband speech quality and intelligibility are improved, but computational complexity increases significantly making it impractical for mobile and embedded applications
Solution Approach 1:
The patent divides the speech signal into narrowband and highband components using a filter bank. The narrowband signal is processed using traditional CELP coding, while the highband signal is generated through spectral extrapolation. This segmentation allows wideband processing to be achieved without applying computationally intensive algorithms to the entire spectrum, thereby reducing overall computational complexity while maintaining speech quality.
2Measurement precision
If the entire spectrum of wideband signal is encoded to desired quality using narrowband coding technique, then speech quality is improved, but bandwidth increases unacceptably large
Solution Approach 1:
The patent extracts only the essential narrowband components for transmission using traditional CELP coding, which is bandwidth-efficient. The highband components are generated locally at the receiver through spectral extrapolation algorithms rather than being transmitted. This extraction approach allows wideband speech quality to be achieved without transmitting the entire wideband spectrum, thereby avoiding unacceptably large bandwidth increases.
3Adaptability or versatility
If narrowband speech coding is extended to wideband without transcoding, then compatibility with narrowband systems is maintained, but the coarse spectral envelope of highband portion cannot be predicted accurately from narrowband spectral envelope
Solution Approach 1:
The patent performs preliminary spectral extrapolation to generate highband components from narrowband components using algorithms that predict highband spectral characteristics. By preparing the highband signal in advance through spectral extrapolation rather than relying solely on real-time prediction during decoding, the system achieves both compatibility with narrowband systems and improved accuracy of the highband spectral envelope.
Data Source
Figure 1a~1b
Figure 2a~2b
Figure 3a~3b
AI summary
A wideband speech encoder according to one embodiment includes a narrowband encoder and a highband encoder. The narrowband encoder is configured to encode a narrowband portion of a wideband speech signal into a set of filter parameters and a corresponding encoded excitation signal. The highband encoder is configured to encode, according to a highband excitation signal, a highband portion of the wideband speech signal into a set of filter parameters. The highband encoder is configured to generate the highband excitation signal by applying a nonlinear function to a signal based on the encoded narrowband excitation signal to generate a spectrally extended signal.