Dynamic Noise Suppression for Non-Stationary Speech Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional speech processing systems face challenges in noise reduction for non-stationary noisy speech signals, as they assume stationary noise, leading to errors in speech recognition and processing due to misinterpretation of noise structures as words.
Innovation Solution
A noise suppression system that transforms time-domain input signals into spectra, smooths magnitudes, determines desired noise shapes, calculates suppression factors, and generates noise suppression filter coefficients to effectively filter out noise in a dynamic and adaptive manner, suitable for both stationary and non-stationary environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If conventional speech processing systems use stationary noise assumption, then the system complexity is low, but the noise reduction accuracy deteriorates in non-stationary environments
Solution Approach 1:
The patent implements dynamic noise estimation by continuously updating noise parameters during speech pauses rather than assuming stationary noise. The system adapts noise estimates in real-time based on observed signal characteristics, allowing the noise model to dynamically adjust to changing environmental conditions while maintaining manageable computational complexity through efficient update mechanisms.
Solution Approach 2:
The system changes noise estimation parameters adaptively by detecting speech pauses and updating noise spectral characteristics only during these periods. This parameter change strategy allows the system to track non-stationary noise while avoiding unnecessary computations during active speech, thus balancing accuracy improvement with computational efficiency.
2Power
If noise estimate is updated only during speech gaps, then the computational load is reduced, but the noise estimation accuracy deteriorates for non-stationary noise
Solution Approach 1:
The system employs periodic noise estimation updates synchronized with speech pause detection. Instead of continuous updating, the noise model is refreshed periodically during identified speech gaps, which reduces computational load while maintaining accurate noise tracking. This periodic action is triggered by voice activity detection mechanisms that identify appropriate update moments.
Solution Approach 2:
The system uses feedback from speech activity detection to control when noise estimation updates occur. By monitoring speech presence and triggering updates only during detected pauses, the system creates a feedback-driven update mechanism that reduces unnecessary computations while ensuring noise estimates are refreshed at appropriate moments for non-stationary environments.
3Productivity
If the system interprets noise structures as speech words, then speech recognition coverage is improved, but the error rate increases due to false positives
Solution Approach 1:
The system extracts and removes noise components from the speech signal before recognition processing. By separating noise structures from speech content through spectral subtraction and filtering techniques, the system prevents noise-induced false positives while preserving genuine speech information, thus maintaining recognition coverage without increasing error rates.
Solution Approach 2:
The system converts harmful noise structures into beneficial information by using noise estimation during speech pauses to characterize and subsequently suppress noise during active speech. The previously harmful noise patterns are transformed into a basis for creating accurate noise models that improve recognition reliability by preventing false positive identifications.
Data Source
AI summary
Systems and methods for noise reduction are provided including operations for noisy speech signals, such as speech signals that are subject to speech processing, speech recognition and speech transmission for voice communication purposes. In one embodiment, a system for noise suppression includes an input smoothing filter to smooth magnitudes of the input spectrum, a desired noise shape determination block configured to determine a desired noise shape of the noise spectrum dependent on the smoothed-magnitude input spectrum, and a suppression factors determination block configured to determine a set of suppression factors based on the desired noise shape and the smoothed-magnitude input spectrum. In one embodiment, a filter coefficient determination block is configured to determine noise suppression filter coefficients from the desired noise shape of the noise spectrum. Embodiments are also directed to systems and methods for noise reduction. System configurations and processes are provided for formant detection.