Performance evaluation of dwt based speech enhancement
Citations Over Time
Abstract
Tracking of stationary noise is not an issue when compared to non-stationary noise tracking as stationary noise properties remain constant throughout time. When noise properties itself is changing with time noise standard deviation estimation is found to be a better tool in analyzing noise properties. It is estimated to find out the noise power spectrum which is assumed to be same as noise variance. This computation is done by E-DATE which has its base in STFT. Estimation quality of noise power spectrum of the new algorithm is determined based on how efficient the algorithm is in removal of noise component during speech enhancement when compared to original algorithm. Since both STFT and DWT follow the weak sparseness property which is the main basis for E-DATE algorithm, STFT could be replaced effectively by DWT in the algorithm. DWT has an advantage of providing good time-frequency resolution over STFT. A comparison on proposed work based on different wavelets used is also done. Performance evaluation of proposed work is done based on output SNR values and PESQ-MOS scores when noise power spectrum estimation was integrated with noise suppression/reduction algorithm.
Related Papers
- → Switching Bilateral Filter With a Texture/Noise Detector for Universal Noise Removal(2010)156 cited
- Image Denoising with Linear and Non-Linear Filters: A REVIEW(2013)
- → Switching bilateral filter with a texture/noise detector for universal noise removal(2010)15 cited
- → Review on Linear and Non - Linear Filter(2023)1 cited
- → Noise filtering method for color images based on LDA and nonlinear diffusion(2008)1 cited