Speech Coder LPC Excitation Signal Smoothing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech coders face challenges in handling stationary background noise during periods of voice inactivity, leading to quality degradations such as swirling due to inadequate estimation and quantization of synthesis filter coefficients and excitation signals, particularly at low bit rates.
Innovation Solution
The method involves modifying the LPC synthesis filter excitation signal by reducing power and spectral fluctuations, combined with LPC parameter smoothing, using techniques like tilt compensation and noise mixing, to improve the quality of the speech decoder output during speech inactivity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If speech coders use standard LPC synthesis filter models optimized for speech signals, then speech coding efficiency is improved, but background noise quality deteriorates due to swirling and temporal fluctuations
Solution Approach 1:
The patent applies different processing strategies to different signal types: standard LPC synthesis for speech segments and modified synthesis with reduced spectral tilt for background noise segments. The voice activity detector enables local adaptation of synthesis parameters based on signal type, improving background noise quality without compromising speech coding efficiency
Solution Approach 2:
The system dynamically adjusts synthesis filter parameters and spectral tilt compensation based on voice activity detection. During speech inactivity periods, the encoder switches to a mode that reduces temporal fluctuations and spectral variations, while maintaining standard speech coding during active speech periods
2Reliability
If codec bit rate is increased, then swirling in background noise is reduced, but transmission efficiency deteriorates
Solution Approach 1:
Instead of uniformly increasing bit rate for all segments, the patent applies enhanced synthesis parameters (reduced spectral tilt, modified excitation) only to background noise segments detected by voice activity detection. This partial application of enhanced processing reduces swirling without the full cost of increased bit rate, maintaining transmission efficiency while improving background noise quality
3Measurement precision
If synthesis filter parameters are precisely estimated and quantized, then speech quality is improved, but computational complexity increases
Solution Approach 1:
The patent applies enhanced parameter estimation and quantization only during background noise segments rather than continuously. The voice activity detector enables selective application of computationally intensive processing only when needed, reducing overall computational complexity while maintaining precision where it matters most for background noise quality
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In a method of smoothing background noise in a telecommunication speech session receiving and decoding a coded speech signal and subsequently, determining LPC parameters and an excitation signal for the received signal. Thereafter, synthesizing and outputting an output signal based on the determined LPC parameters and excitation signal. In addition, smoothing the determined set of LPC parameters by providing a low pass filtered set of LPC parameters and modifying the determined excitation signal by reducing power and spectral fluctuations of the excitation signal during periods of speech inactivity.