Adaptive Speech Filter for Ambient Noise Attenuation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for attenuating ambient noise in video recordings often compromise the integrity of the signal or require significant user interaction and computational resources, especially when dealing with speech content and wind noise.
Innovation Solution
An adaptive speech filter system that analyzes input signals using recursive spectral analysis to differentiate noise from speech, generating a filter curve based on energy spectral components and zero points, and applies filtering in both the frequency and time domains with pre- and post-ringing filters to minimize side effects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If spectral subtraction is used to attenuate ambient noise, then noise reduction is achieved, but the integrity of the speech signal may be compromised
Solution Approach 1:
The patent applies different processing strategies to different parts of the signal spectrum. It identifies speech segments versus non-speech segments and applies noise attenuation selectively, preserving speech components while removing ambient noise. The system analyzes local signal characteristics (spectral flux, zero-crossing rate) to determine which portions should be attenuated and which should be preserved.
Solution Approach 2:
The patent uses dynamic thresholding and adaptive parameters that change based on signal conditions. The noise attenuation factor is adjusted in real-time based on the detected speech activity and noise characteristics. This dynamic approach allows the system to adapt to varying speech and noise levels, maintaining signal integrity across different recording conditions.
2Ease of operation
If automatic noise estimation is used by searching for non-speech sequences, then user interaction is minimized, but computational resources and processing time increase
Solution Approach 1:
The patent performs noise estimation on a subset of the signal rather than analyzing the entire recording. It identifies and processes only the non-speech segments for noise characterization, applying the estimated noise parameters to the entire signal. This partial analysis approach reduces computational burden while maintaining effective noise reduction.
Solution Approach 2:
The patent performs noise characterization in advance by analyzing non-speech portions of the signal before applying noise attenuation to the speech segments. The noise profile is estimated once and then reused, avoiding repeated heavy computations during the main processing phase. This preliminary estimation reduces overall processing time.
3Object-affected harmful factors
If multiple microphones are used for beam forming or active noise suppression, then noise attenuation effectiveness improves, but device complexity increases
Solution Approach 1:
The patent extracts the noise component from the mixed signal by identifying non-speech segments and using them to characterize the ambient noise profile. This extracted noise model is then subtracted from the speech portions, achieving noise reduction without requiring multiple microphones or complex spatial processing. The harmful noise component is separated and removed through signal analysis rather than spatial filtering.
Data Source
AI summary
According to a preferred aspect of the instant invention, there is provided a system and method that allows the user to attenuate ambient noise in speech recordings in the audio part of a video recording. The user does not need to define particular sections or samples or individual parameters. The system is automatically analyzing the input signal and in a plurality of individual steps detects the ambient noise, determines an adaptive filter, implements the filter and therewith attenuates the ambient noise accordingly.