Speech Enhancement Using Neural Network

Ravi Kumar Kandagatla, P.V. Subbaiah, Ramya Sai Daddanala, Kandula Lakshmi Prasanna, KarnatiVenkata Narasamma, CheguSeshu Kumar

The main focus of Speech enhancement is to upgrade intelligibility / quality of corrupted speech signal. Traditional speech enhancement algorithms like spectral subtraction, MMSE estimators provide improved results but introduces distortion. To overcome the distortion effects, neural network perspective of speech enhancement is to be implemented in this work. The Neural network is trained by the speech and noise data base and then the trained network is used for further processing of corrupted speech signal. The performance of implemented method is analyzed using objective performance measurePerceptual Evaluation of Speech Quality (PESQ) and Signal to Noise Ratio (SNR).

Volume 12 | Issue 2

Pages: 2328-2334

DOI: 10.5373/JARDCS/V12I2/S20201279