Deep Neural Network Audio Noise Suppression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing noise reduction systems in audio signals are inadequate in efficiently suppressing a variety of noises, especially when speech and noise have similar frequency levels, and fail to provide real-time dynamic noise reduction and signal enhancement.
Innovation Solution
A system and method utilizing a deep neural network to identify and suppress multiple types of noises, including environmental and Gaussian noises, by converting audio signals into the PCM format and applying a noise suppression module that learns noise behavior in real-time, enabling effective noise reduction and signal enhancement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional static filters are used for noise reduction, then the filtering process is simple and fast, but the noise cannot be effectively reduced when speech and noise have similar frequency levels
Solution Approach 1:
The patent transforms static filters into dynamic adaptive filters that automatically adjust their parameters based on the input signal characteristics. The filter coefficients are continuously updated using algorithms like LMS (Least Mean Squares) to adapt to changing noise conditions, enabling effective noise reduction even when noise and speech frequencies overlap.
Solution Approach 2:
The patent changes the filter parameters dynamically based on the estimated noise characteristics. By monitoring signal statistics and adapting filter cutoff frequencies, bandwidths, and coefficients in real-time, the system optimizes noise reduction performance for different acoustic environments and noise types.
2Reliability
If threshold-based noise filtering is applied, then the implementation is straightforward, but noise is not reduced in overlapping noise and voice signals
Solution Approach 1:
The patent introduces noise estimation algorithms as intermediary components that analyze the signal to separate noise from speech. These algorithms use techniques like spectral subtraction, Wiener filtering, or deep learning-based noise estimation to create a noise profile that guides the filtering process, enabling accurate noise suppression without affecting speech quality.
Solution Approach 2:
The patent replaces simple threshold-based mechanical filtering with sophisticated signal processing algorithms. Instead of using fixed threshold comparisons, the system employs adaptive filtering, spectral analysis, and machine learning models that can intelligently distinguish between noise and speech components based on their statistical and spectral characteristics.
3Productivity
If hardware-based noise reduction solutions are used, then the static noise from microphones is reduced, but real-time dynamic noise reduction and signal enhancement cannot be achieved
Solution Approach 1:
The patent implements self-adjusting noise reduction systems that automatically adapt to changing acoustic environments without manual intervention. The system continuously monitors the input signal, estimates noise characteristics, and adjusts filter parameters in real-time, enabling dynamic noise reduction that responds to varying speech and noise conditions.
Solution Approach 2:
The patent incorporates feedback mechanisms where the output of the noise reduction system is monitored and fed back to adjust the filtering parameters. This closed-loop approach allows the system to learn from its performance and continuously optimize noise reduction effectiveness, adapting to new noise patterns and speech characteristics encountered during operation.
Data Source
AI summary
This disclosure relates generally to a system and method to identify a plurality of noises or their combination to suppress them and enhancing the deteriorated input signal in a dynamic manner. It identifies noises in the audio signal and categorizing them based on the trained database of noises. A combination of deep neural network (DNN) and artificial Intelligence (AI) helps the system for self-learning to understand and capture noises in the environment and retain the model to reduce noises from the next attempt. The system suppresses unwanted noise coming from the external environment with the help of AI based algorithms, by understanding, differentiating, and enhancing human voice in a live environment. The system will help in the reduction of unwanted noises and enhance the experience of business and public meetings, video conferences, musical events, speech broadcasts etc. that could cause distractions, disturbances and create barriers in the conversation.