Audio Noise Reduction from Audio Signals and Speech Signals

Speech signal analysis is one of the important areas of research in multimedia applications. Discrete Wavelet technique is effectively reduces the unwanted higher or lower order frequency components in a speech signal. Wavelet-based algorithm for audio de-noising is worked out. We focused on audio signals corrupted with white Gaussian noise which is especially hard to remove because it is located in all frequencies. We use Discrete Wavelet transform (DWT) to transform noisy audio signal in wavelet domain. It is assumed that high amplitude DWT coefficients represent signal, and low amplitude coefficients represent noise. Using thresholding of coefficients and transforming them back to time domain it is possible to get audio signal with less noise. Our work has been modified by changing universal thresholding of coefficients which results with better audio signal. In this various parameters such as SNR, Elapsed Time, and Threshold value is analyzed on various types of wavelet techniques alike Coiflet, Daubechies, Symlet etc. In all these, best Daubechies as compared to SNR is more for Denoising and Elapsed Time is less than others for Soft thresholding. In using hard thresholding Symlet wavelet also works better than coiflet and Daubechies is best for all. Efficiency is 98.3 for de-noising audio signals which also gives us better results than various filters.