A processing tool having to extract useful spectral information from a ultrasonic signal will be in front of, at least, two difficult problems: (i) the analysis of a complex nature signal, with parameters of nondeterminist type and continuously changing with medium conditions and (ii) the discernment among factors providing a useful information and those that just act as undesirable perturbations, masking the information to be extracted. Thus, the computer evaluation in laboratory, of the reliability and robustness of any spectral processing tool for this aim, is a crucial aspect for future uses of such tool in medical diagnosis, before applying it to detect changes in real tissues.
digital signal processing proakis manolakis pdf download
Download File: https://gohhs.com/2vD9Dx
In order to overcome this limitative aspect for precise diagnosis by noninvasive thermal estimation [18], a new signal processing step is proposed by authors to be added to the analytical procedure, looking for the consecution of a higher frequency resolution, that with the basic spectral approach detailed in the previous subsection. Initial results from our improved procedure for processing the multiple power spectra involved in multipoint measures provide a better resolution than using previous ultrasonic estimation options. In addition, the Burg method is applied here as an alternative option to the classical approach for power spectrum estimation, giving a better final frequency resolution. In our estimation of the AR parameters by Burg method, a minimization of the direct and inverse errors of the linear predictors is made, with the restriction that AR parameters should satisfy the Levinson-Durbin recursion.
The high resolution (HR) thermal detection is completed by properly applying and adapting, to our estimation problem involving rather short-time windows, a procedure in certain way parallel to some techniques applied in other digital signal processing areas, by properly decomposing the echotraces in many sufficiently small fractional time-windows and artificially extending their digital lengths in all of them, before to be parametrically analyzed in the power spectrum domain with an improved frequency resolution. Each resulting time fringe is previously extended, before and after of the occurrence of each original acquired short echo-segment, with many null-value new samples in number enough to attain a total of Nf digital samples. So, the needed high resolution in the frequency domain can be attained for the subsequent power spectra calculated with Burg method from the extended digital registers associated to the successive time subtraces, instead the original registers with Ni samples, Nf = xNi.
An accurate evaluation of performance was performed, for raw bioultrasonic signals patterns (deterministic echo-pulses from regular structure phantoms, and also more complex echo-traces from randomly distributed scattered media). The potential effectiveness of our approach, as a possible reliable diagnosis tool, was demonstrated. It is based on an implementation of the Burg algorithm and adding a new processing step to obtain a higher resolution in the frequency peaks measure of the spectrum overtones. Other advantage of this approach, in respect to possible alternatives based on complex imaging systems, is that only one transducer and signal acquisition channel are needed.
2ff7e9595c
Comments