Speech Signal Leveling With Voice-Aware Gain Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech signal processing systems face challenges in maintaining an optimal signal-to-noise ratio (SNR) due to dynamic speaking distances and voice levels, leading to inadequate recognition rates in speech recognition systems and intelligibility issues in hands-free communication, as automatic gain control (AGC) methods often amplify noise and fail to maintain full-scale speech signals.
Innovation Solution
A speech signal leveling system that includes a controllable-gain block, a speech detecting block, and a gain control block, which apply frequency-dependent or frequency-independent gains to maintain a predetermined mean or maximum signal level based on voice activity detection and pause detection, ensuring consistent signal quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If a peak limiter is used to amplify soft speech to full scale, then the speech signal level is improved, but the signal-to-noise ratio deteriorates because noise is also amplified
Solution Approach 1:
The patent applies different gain values to different signal components based on their classification. Speech components receive one gain value while noise components receive a different gain value, allowing selective amplification that maintains SNR while achieving full-scale speech output.
Solution Approach 2:
The system dynamically adjusts gain values based on real-time signal classification. The classification module continuously identifies speech versus noise components, and the gain control module adapts gain parameters accordingly, enabling the system to respond to changing signal conditions while maintaining optimal SNR and speech level.
2Manufacturing precision
If a peak limiter attenuates loud speech to full scale, then the speech signal level is improved, but the signal-to-noise ratio deteriorates because the limiter is more often active
Solution Approach 1:
The system applies selective gain control where speech components are treated differently from noise components. By classifying signal components and assigning appropriate gain values, the system can attenuate loud speech to full scale while preventing excessive noise amplification.
Solution Approach 2:
The gain parameters are dynamically adjusted based on the classified signal characteristics. When loud speech is detected, the system applies appropriate attenuation while monitoring noise levels, ensuring the limiter operates effectively for speech control without excessively amplifying background noise.
3Device complexity
If a simple automatic gain control is used, then the device complexity is reduced, but the speech recognition rate deteriorates due to inadequate signal leveling
Solution Approach 1:
The system segments the input signal into distinct components (speech and noise) through classification. This segmentation allows separate gain control for each component type, achieving precise speech leveling that improves recognition rates while maintaining manageable system complexity through modular processing stages.
Solution Approach 2:
The system uses feedback from the signal classification results to control gain parameters. The classification module provides information about speech versus noise components, which feeds back to the gain control module to adjust amplification/attenuation accordingly, creating a closed-loop system that adapts to maintain optimal speech levels for recognition.
Data Source
AI summary
A speech signal leveling system and method include generating an output signal by applying a frequency-dependent or frequency-independent controllable gain to an input signal, the gain being dependent on a gain control signal, and generating at least one speech detection signal indicative of voice components contained in the input signal. The system and method further include generating the gain control signal based on the input signal and the at least one speech detection signal, controlling the controllable-gain block to amplify or attenuate the input signal to have a predetermined mean or maximum or absolute peak signal level as long as voice components are detected in the input signal.