Signal Feature Extraction Using Granular Computing. Comparative Analysis with Frequency and Time Descriptors Applied to Dynamic Laser Speckle Patterns

Abstract

The laser dynamic speckle is a phenomenon caused by the fluctuant interference of the laser light reflected from an illuminated surface where some kind of activity is taking place. Signals generated by the intensity changes in each pixel through the sequence are processed with the finality of identifying underlying activity in each point. In this work we compare the performance of a Rough Fuzzy Granular Descriptor (previously published) against a set of dynamic speckle descriptors based in time and frequency processing. To perform this evaluation a numerical simulation is proposed to explore their linearity, robustness, sensitivity related to the samples quantity, as well as also by their computing time. Also the robustness to inhomogeneous spatial intensity was evaluated in an experiment performed with the illuminated surface of an actual biological object.

Visualization of natural processes applying rough-fuzzy temporal granularity to laser speckle images

“…The RFGD algorithm facilitates the processing of dynamic speckle patterns, allowing results in quasi real time thanks to the simplicity of the algorithm, its robustness due to less noise sensitivity and its need of only a few images. The method is faster than those based in frequency [24], in speed it is possible to compare favourably with time domain methods, though these have not given satisfactory results in all the tests as RFGD did.…”

“…The RFGD evaluates activity of different pixels independently of the knowledge of the studied phenomenon. Some of the descriptors normally used, described in the literature [1], are adequate for particular cases, but RFGD so far has provided good results in all cases that it was applied [24]. In general, there is not a standard or pattern of comparison for the results; they depend on what expects to be observed or it is evaluated by an expert in the problem, considering the sample and the proposed aim.…”

“…Tissue breakage or failure occurs at a microscopic level, causing defects in cells below the skin without damaging it. Breaking the skin or bruising the flesh releases enzymes that in turn cause browning of the fruit [24,30]. The bruise becomes perceptible a few days later so it is crucial to detect it early to provide better fruits for the consumer and reduce potential economic losses in the fruit industry.…”

“…The speckle pattern of bruise in apples has already been analysed with different methods [1,24,27], with the aim to distinguish damaged regions. As biological samples bear inherent variability we have used the same data obtained for the bruising of the apple that were previously employed for testing other methods to permit the easy comparison with the results shown in the bibliography.…”

“…Fig. 5 shows results obtained for the immediately after bruise stage, with RFGD and other traditional methodologies [1,24] as: autocorrelation and standard deviation (statistical), average differences (temporal) and energy of spectral band, Shannon entropy, cutoff frequency, mean frequency and wavelet entropy (frequency). In the processing of the first 20 and all together 500 speckle patterns, some of these cannot to detect the bruise and others do not make a clear difference between the bruised and not bruised regions.…”