Speech Pre-Processing With Frequency-Dependent Gain for Noise Intelligibility
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio systems struggle to maintain speech intelligibility in noisy environments, as previous methods either fail to adapt automatically to varying noise levels or compromise speech intelligibility by keeping the overall speech level constant.
Innovation Solution
An audio system processes speech signals using a speech filter and gain computer to maintain a target Speech Intelligibility Index (SII) while minimizing overall speech level, applying frequency-dependent gains to the speech signal and environmental sound signals to preserve intelligibility across different noise levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the overall speech level is kept constant, then listening comfort is maintained, but speech intelligibility deteriorates in noisy environments
Solution Approach 1:
The patent applies different gain adjustments to different frequency bands of the speech signal rather than uniformly adjusting the overall speech level. The speech processing circuit identifies frequency regions critical for intelligibility (such as formant regions) and applies selective amplification to these regions while minimizing amplification in less critical regions, thereby maintaining speech intelligibility without proportionally increasing overall loudness.
Solution Approach 2:
The system dynamically changes the spectral distribution parameters of the speech signal by applying frequency-dependent gains. The speech processing circuit analyzes the speech signal's spectral characteristics and adjusts the gain in each frequency band to optimize the Speech Intelligibility Index (SII), adapting to varying noise conditions while maintaining natural speech perception.
2Reliability
If speech level is increased to overcome background noise, then speech intelligibility improves, but overall loudness and distortions increase
Solution Approach 1:
The patent selectively amplifies only the frequency regions that are critical for speech intelligibility rather than uniformly increasing the overall speech level. By identifying and enhancing formant regions and other intelligibility-critical frequency bands, the system achieves improved speech understanding without proportionally increasing overall loudness, thereby avoiding listener discomfort and potential distortions.
Solution Approach 2:
The system converts the harmful effect of background noise into a beneficial guide for selective speech enhancement. By analyzing the spectral characteristics of both speech and noise, the speech processing circuit identifies which frequency regions contain speech information that is most degraded by noise and selectively enhances only those regions, transforming the noise problem into an opportunity for targeted speech reinforcement.
3Adaptability or versatility
If automatic adaptation to varying noise levels is implemented, then speech intelligibility is maintained, but system complexity increases
Solution Approach 1:
The patent implements a feedback mechanism where the speech processing circuit continuously monitors the Speech Intelligibility Index (SII) of the processed speech signal and automatically adjusts the frequency-dependent gains to maintain a target SII value. This closed-loop control enables automatic adaptation to varying noise conditions without requiring manual intervention, while the feedback mechanism guides the complexity of processing by only making adjustments when intelligibility degradation is detected.
Solution Approach 2:
The speech processing system performs self-adjustment by automatically analyzing its own output and adapting to changing acoustic conditions. The speech processing circuit monitors the processed speech signal's intelligibility characteristics and autonomously modifies the frequency-dependent gain structure to maintain optimal speech understanding, reducing the need for external control or complex manual configuration.
Data Source
AI summary
An audio system processes a speech signal to maintain a target value of the speech intelligibility index (SII) while minimizing the overall speech level so that speech intelligibility is preserved across different environmental sound levels while possible distortions and overall loudness are mitigated. In one embodiment, a hearing aid processes a speech signal received from another device to maintain a target value of the SII while minimizing the overall speech level before mixing the speech signal with a microphone signal.