Speech Encoding Dual Noise Shaping Filter
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Speech coding technologies face challenges in reducing quantization noise, which affects the quality of transmitted speech signals, as quantization distorts the signal and introduces noise that is not always perceptually favorable across the frequency spectrum.
Innovation Solution
The method involves using a dual noise shaping filter approach, where a first noise shaping filter processes the input signal before quantization and a second filter processes the quantized output signal, allowing for independent manipulation of the signal and coding noise spectra to minimize distortion and noise perception.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If quantization is applied to reduce signal representation complexity, then the number of bits required for transmission is reduced, but quantization noise is introduced that degrades speech quality
Solution Approach 1:
The patent applies noise shaping to convert the harmful quantization noise into a beneficial tool by filtering it through a noise shaping filter that redirects noise energy to frequency regions where the human ear is less sensitive. This transforms the unavoidable quantization noise into a perceptually acceptable form, improving speech quality while maintaining bit rate efficiency.
Solution Approach 2:
The patent implements different noise shaping filter coefficients for different frequency bands, allowing the noise spectrum to be locally optimized. By applying frequency-dependent filtering, the system reduces noise in perceptually critical frequency regions while tolerating noise in less sensitive regions, achieving localized quality improvement across the speech spectrum.
2Object-affected harmful factors
If noise shaping filtering is applied to reduce perceptible quantization noise, then speech quality is improved, but the device complexity increases due to additional filtering operations
Solution Approach 1:
The patent divides the noise shaping process into separate filter stages with distinct coefficient sets for different frequency regions. By segmenting the filtering operation, the system achieves comprehensive noise shaping across the speech spectrum while maintaining manageable computational complexity through modular filter design and independent coefficient optimization.
3Device complexity
If a single noise shaping filter is used for quantization, then the system remains simple, but independent manipulation of signal and noise spectra is limited
Solution Approach 1:
The patent employs dynamic coefficient selection where the noise shaping filter coefficients are adapted based on the input signal characteristics and desired output spectrum. This allows the system to dynamically adjust the noise shaping behavior to match different speech conditions, achieving versatile spectrum manipulation while maintaining a relatively simple filter structure through parameter adaptation rather than structural complexity.
Data Source
AI summary
Some embodiments describe methods, programs, and systems for speech encoding. Among other things, a received input signal representing a property of speech is quantized to generate a quantized output signal. Prior to the quantization, a version of the input signal is supplied to a first noise shaping filter having a first set of filter coefficients effective to generate a first filtered signal. Following the quantization, the quantized output signal is supplied to a second noise shaping filter having a second set of filter coefficients, thus generating a second filtered signal. A noise shaping operation is performed to control a frequency spectrum of a noise effect in the quantized output signal caused by the quantization, wherein the noise shaping operation is based on both the first and second filtered signals. Finally, the quantized output signal is transmitted in an encoded signal.