Audio Noise Suppression Using Input Signal Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing noise suppression algorithms in real-time communications struggle to reliably differentiate between keyboard clicking noise and speech onset, leading to either unacceptable delays or distortion of speech components in audio signals during calls.
Innovation Solution
A method and system that utilize input signals from user-operated devices to determine noise-generating activities, enabling selective noise suppression only when such activities occur, thereby minimizing distortion of speech components by activating noise suppression algorithms only during device activity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If background noise reduction methods are used to suppress noise in audio signals, then stationary noise can be effectively removed, but transient and non-stationary noise components such as keyboard clicking noise cannot be reliably identified and attenuated
Solution Approach 1:
The system changes the parameters of noise analysis by switching between different noise reduction algorithms based on the statistical characteristics of the audio signal. When transient noise is detected (through statistics-based detection), a specific transient noise reduction algorithm is applied; otherwise, standard background noise reduction is used. This adaptive parameter change resolves the contradiction by matching the algorithm to the noise type.
Solution Approach 2:
The noise reduction system is made dynamic by continuously monitoring the audio signal statistics and adjusting the noise reduction strategy in real-time. The system transitions between different operational states (standard noise reduction vs. transient noise reduction) based on detected signal characteristics, enabling reliable identification and attenuation of transient noise while maintaining effectiveness against stationary noise.
2Object-affected harmful factors
If keyboard noise attenuation algorithms are used to detect and filter clicking noise, then keyboard noise can be suppressed, but speech components may be distorted because clicking noise and speech onset have similar initial characteristics
Solution Approach 1:
The system adjusts the detection parameters dynamically based on the current audio context. By monitoring signal statistics and detecting transient noise characteristics, the system only activates aggressive keyboard noise attenuation when transient noise is actually present. This prevents misclassification of speech onsets as keyboard noise, thereby preserving speech integrity while still suppressing keyboard clicking when detected.
Solution Approach 2:
The noise attenuation process incorporates feedback mechanisms where the detected noise characteristics inform the attenuation strength and timing. The system uses statistics-based detection to provide feedback about the presence of transient noise, which then modulates the attenuation algorithm's behavior. This feedback loop ensures that speech components are not distorted while keyboard noise is effectively suppressed.
3Object-affected harmful factors
If noise detection is performed to differentiate between speech and clicking noise, then noise can be attenuated, but detection reliability is reduced when user is engaged in a call because clicking noise and speech onset have similar characteristics
Solution Approach 1:
The detection parameters are changed based on the operational context. The system uses statistics-based detection to identify transient noise characteristics that differ from speech patterns. By monitoring parameters such as signal variance, zero-crossing rates, and spectral features, the system can differentiate between keyboard clicking and speech onsets even when they have similar initial characteristics, thereby improving detection reliability during calls.
Solution Approach 2:
The detection process is segmented into multiple analysis stages. Rather than relying on a single detection threshold, the system divides the analysis into preliminary detection, confirmation, and attenuation phases. This segmentation allows the system to use different detection strategies at different stages, improving overall differentiation accuracy between noise and speech while maintaining noise suppression effectiveness.
4Reliability
If full click detection is used to reliably detect clicking noise, then detection accuracy improves, but processing delay is introduced which is unacceptable for real-time communications
Solution Approach 1:
The system performs preliminary noise detection using statistics-based methods that analyze incoming audio characteristics in real-time. Rather than waiting for a full click to complete before detecting it, the system proactively identifies transient noise patterns as they emerge. This preliminary action enables early attenuation of keyboard noise without requiring full click detection, thereby maintaining detection accuracy while minimizing processing delay for real-time communications.
Data Source
AI summary
A method and computing system for suppressing noise in an audio signal, comprising: receiving the audio signal at signal processing means; determining that another signal is input to the signal processing means, the input signal resulting from an activity which generates noise in the audio signal; and selectively suppressing noise in the audio signal in dependence on the determination that the input signal is input to the signal processing means to thereby suppress the generated noise in the audio signal.