Figure 7.

Scheme explaining back-scattered signal processing and post-processing steps. (1) After the acquisition of 60-s of back-scattered signal for each particle (PMMA and yeast cells), the acquired signals were downsampled from 100 kHz to 5 kHz; (2) Then, each whole acquisition was filtered using a 500 Hz high-pass Butterworth filter; (3) The 60-s segments of signal were divided into short-term portions of 2 s; (4) Noisy 2-s short-term signal portions were therefore removed taking into account the z-score value calculated before epoching. Portions whose value(s) exceeded |z-score| > 10 were removed at this stage; (5) After signal processing steps, our dataset was composed by 2-s short-term portions of back-scattered signal; (6) A set of more than 40 time- and frequency-domain features characterizing each 2-s portion was therefore created and assigned to each particle class; (7) Then, the Linear Discriminant Analysis (LDA) was applied to gather all the relevant information provided from the original set of features in a single one, in order to facilitate interpretation by an interrogation sensing system. At the end, the proposed system would be able to provide the values corresponding to that single-feature, resulting from a multivariate combination of the parameters calculated in (7).