Noise Reduction in Audio

Abstract

Noise reduction in audio is a process of minimizing the presence of noise in the audio signal. Noise reduction approach helps user understand how to remove noise from the signal. Audio is very susceptible with noise and it presence will give impact on two major ways; first it will degrade the quality of information being transfer and second limit the capacity of data being transmit. There are lot of noise reduction techniques been proposed as to remove noise from audio signal such as wiener filter and average filter. In this project, wavelet theory implementation is proposed for reducing noise from audio signal. The main objective is to identify and remove the presence of noise in the signal. Furthermore, measurements of noise signal pattern are recorded before and after the noise reduction process. To be specific, this project will focus on speech type of audio. Speech acquisition of the participant were recorded in two types of environments; controlled and uncontrolled environments. Types of speech recorded in each environment are normal, whispering and angry speech. Speech signal patterns are analysed using cross-correlation between three wavelets families consists of Coiflet 5, Daubechies 10 and Daubechies 9 in order to determine best fit the original signal.