Abstract:The traditional denoising algorithm can reduce the noise from environment, yet an evident downside of this is the sharp decrease of efficacy under non-stationary noise with low signal to noise ratio (SNR). In this case, this paper presents a denoising algorithm for environmental sound based on non-negative matrix factorization (NMF) and optimally-modified log-spectral amplitude (OMLSA). First of all, prior information is obtained from the noise and environmental sound dataset to separate the noise from noisy sound using supervised NMF. Then, decision-direction approach (DD) is used to calculate a priori SNR estimation according to the power spectral density (PSD) of the separated noise. Finally, under the log minimum mean square error criterion, the OMLSA gain function is obtained to filter out the noise in the environment. In this paper, the performance of the proposed algorithm is tested with white noise, F16 and babble under -5dB, 0dB and 5dB SNR conditions respectively. The experimental results show that this algorithm has strong ability of filtering out the noise superimposed in the environmental sound and improving the SNR gains. Moreover, comparing with other noise suppression algorithms, the proposed algorithm preforms better in preserving weak environmental sound, significantly reducing environmental sound distortion and avoiding the introduction of musical noise.

Improvement of audio noise reduction system based on RLS algorithm

Simulation and Research on Adaptive Acoustic Feedback Reduction System

An Improved Nlms Algorithm Based on Speech Enhancement

The Audio Signal De-Noising Based on the Arima Model and Kalman Filter

A Kind Of Integrated Serial Algorithms For Noise Reduction And Characteristics Expanding In Respiratory Sound

Audio Ranging Signal De-Noising and Hardware-in-the-Loop Simulation

Application of RLS (recursive Least-Squares) Adaptive Filter and Wavelet Transform Method in Canceling Signal Noise

Comparison of Noise Filtering Simulations Based on Adaptive Algorithms

Low-SNR Speech Enhancement and Separation in Driving Environment

A New Speech Denoising Algorithm Based on Improved Spectral Subtraction

Impulsive noise suppression based on time-frequency spectrogram

Denoising Algorithm of Environmental Sound Fused NMF and OMLSA in Non-Stationary Noise Environment

Comparison of Adaptive Filtering Algorithms in Noise Cancellation Application:A Simulation Study

Research on voice noise reduction algorithm based on LMS algorithm and wavelet threshold

Improved spectral subtraction algorithm to restrain strong background noise in tank cockpit

Modified Model of RLS Adaptive Filter for Noise Cancellation

A New Real-Time Noise Suppression Algorithm for Far-Field Speech Communication Based on Recurrent Neural Network

A New Proportionate Filtered-x RLS Algorithm for Active Noise Control System.

Sub-band Adaptive Noise Reduction Algorithm to Improve Speech Intelligibility

Research on Speech Signal De-Noising Algorithm Based on Variable Step Size LMS Adaptive Filter with Strong Background Noises

A Robust Acoustic Echo Canceller for Noisy Environment