A burst is a signal where oscillations have a rapid increase in amplitude from an initial reference level (generally that of the background noise), followed by a decrease, generally more gradual, to a value close to the initial level.

In order to process the discrete acoustic wave bursts separately, it is necessary to proceed with a segmentation technique. However, this paper presents a Rectangular Segmentation Algorithm (RSA) of acoustic emission signals, based on MultiResolution Analysis (MRA) and wavelets analysis.

The idea of the MRA is very similar to subband decomposition and coding, where for coding efficiency, a signal is divided into a set of frequency bands. However the MRA decomposes the envelope of the signal into the detail components and the coarsest approximation via stages of identical lowpass filters and high pass filters and subsampling the outputs. The subsampling by two produces a time compression by two and identical lowpass filters and high pass filters will have bandwidths of the lower (upper) half of the previous lowpass filters. In other way, the filters in MRA serve to divide a singal's spectrum successively by two in the subband decomposition. The task of the MRA is to characaterize non-stationary signals in the best frequency band of the sensor.

Concerning the rectangular segmentation algorithm, its advantage is to define the border of every burst. This operation, is necessary for the acoustic emission analysis system, to calculate the different parameters of data processing of acoustic emission signal which are: counts, energy, RMS, duration, peak time etc.

An other advantage, is to reduce the number of burst samples, without losing the rich information of the acoustic emission signature. Indeed, to improve computational efficiency, the rectangular segmentation algorithm reduces the data size. It does this by eliminating the feature samples that convey the least information about the source signal, using a MRA. This approach generates an envelope of the source signal which is essential for the segmentation method.

Seeing that transient elastic waves are generated by the rapid release of energy from localized sources within a material, this algorithm is validated by comparing the signal energy of the complete and segmented bursts. So, this is a way to correlate the acoustic emission signal energy and the physical phenomenon with very low number of samples.

At the end, this algorithm can be described on term of the linear filtering which can be implemented easily. This paper is attended to show this idea.