Related Experiment Videos
New strategies for improving speech enhancement.
M W White1, R M Holdaway, Y Guo
1Electrical and Computer Engineering Department, North Carolina State University, Raleigh 27695.
International Journal of Bio-Medical Computing
|April 1, 1990
Summary
Two novel design strategies, utilizing adaptive signal processing and neural networks, can significantly enhance speech clarity in noisy environments. These methods focus on improving the accurate perception of speech, leading to better communication outcomes.
Area of Science:
- Signal Processing
- Acoustics
- Machine Learning
Background:
- Speech enhancement systems aim to improve intelligibility and quality of speech signals degraded by noise.
- Current systems face challenges in effectively recovering speech while preserving naturalness.
Purpose of the Study:
- To introduce and evaluate two novel design strategies for substantially improving speech enhancement system performance.
- To demonstrate the potential benefits of these strategies through a preliminary study on pulse recovery.
Main Methods:
- Implementing a non-linear, adaptive signal processing approach for speech-in-noise recovery.
- Optimizing enhancement system design based on maximizing the evocation of target speech percepts.
- Utilizing neural network learning algorithms to determine optimal enhancement parameters for both strategies.
Main Results:
- Preliminary results from pulse recovery illustrate the potential benefits of the proposed strategies.
- The second strategy, optimizing for speech percepts, is hypothesized to yield superior performance due to direct relation to communication goals.
Conclusions:
- The proposed design strategies offer a promising direction for advancing speech enhancement technology.
- Adaptive signal processing and neural networks are key components for achieving improved speech clarity and accurate speech perception.