WAVELET BASED DENOISIONG OF ACOUSTIC SIGNAL

Abstract

This paper introduces the method of noise reduction in acoustic emission signal based on wavelet transform. Acquiring real time signal & analyzing it, is a very difficult task. In data acquisition (DAQ) system, the data can be acquired, measured and analyzed from electrical, mechanical and physical phenomena. In this paper, acoustic signal is generated from pencil lead break at different locations on metal sheet. Acoustic emission (AE) sensor is used to acquire the data of physical phenomenon. It converts the physical parameters into the electrical signal, which is very small (in mV). This has been amplified by using a preamplifier of suitable gain. Amplified signal is transformed into digital signal by using USB data acquisition module. But this signal contains noise. So for processing, digitized signal is connected to computer using USB and processing is done in Matlab software. This amplified signal is filtered using conventional filtering or wavelet filtering. In conventional filtering, Butterworth filter is used but it gives more attenuation. Hence wavelet filtering is better. In wavelet filtering, there are four methods of wavelet threshold denoising that are heursure, minimaxi, rigrsure and sqtwolog. These methods are compared based on the energy, standard deviation and percentage attenuation in filtered signal. The best choice for threshold noise reduction in case of AE signal obtained by pencil lead break is rigrsure. The filtered signal is good for analysis so that different parameters can be easily measured.