Home
Add
Get on Google Play
Home
> Edit
Add/Update Thesis
Title*
Author's Name*
Supervisor's Name
Abstract
Adaptive algorithms are mostly optimized using integer order derivatives for error minimization. The Least Mean Squares (LMS) and Recursive Least Squares (RLS) adaptive filters are among the most commonly employed schemes. The LMS algo rithm is simple to implement, has robust tracking performance in nonstationary environments and is less sensitive to floating point precision effects. However, it has the issue of slow convergence especially when the number of weights is large. The RLS achieves faster convergence but is computationally expensive, and prob lematic in nonstationary environments. This study introduces fractional calculus techniques in stochastic gradient algo rithms. In addition to first order derivative, Fractional Order (FO) derivatives are proposed in the optimization of gradient algorithms. Four configurations have been considered based on whether the fractional derivative is applied to the in stantaneous present or posterior error. For evaluation of the FO algorithms, three applications have been considered, that is, (a) adaptive equalization of multi path channels (b) Active Noise Control Systems (ANCS) and (c) tracking of time varying Rayleigh fading sequences. In equalization, both supervised and unsu pervised algorithms are considered. For the supervised case, FO variants of LMS and Normalized LMS (NLMS) are applied in both feed-forward and decision feed back configurations. In the unsupervised case, FO variants of Gordad and con stant modulus algorithms are developed. In ANCS, FO variants of the NLMS, Filtered-x (input) LMS, Modified FxLMS and Filtered-error LMS algorithms are developed. The noises are modelled as binary, Gaussian and impulsive sources characterized by fractional lower order moments. In tracking, the behavior of FO variants is evaluated for nonstationary environments. A Rayleigh channel has also been considered having Doppler frequency shifts of 0.8KHz to 3KHz. The frac tional algorithms are compared with the standard NLMS, RLS and Extended-RLS schemes. The main performance metrics include (1) mean squared error (2) mean squared deviation (3) relative modelling error (4) model accuracy using both frequency and time domain analysis and (5) symbol error rate. The former three performance metrics help compare the convergence speed and steady state performance; the latter two are application specific. Simulation results are shown for different step sizes and fractional orders. It is seen that the fractional variants show superior performance in all the three applications and hold great promise for future use.
Subject/Specialization
Language
Program
Faculty/Department's Name
Institute Name
Univeristy Type
Public
Private
Campus (if any)
Institute Affiliation Inforamtion (if any)
City where institute is located
Province
Country
Degree Starting Year
Degree Completion Year
Year of Viva Voce Exam
Thesis Completion Year
Thesis Status
Completed
Incomplete
Number of Pages
Urdu Keywords
English Keywords
Link
Select Category
Religious Studies
Social Sciences & Humanities
Science
Technology
Any other inforamtion you want to share such as Table of Contents, Conclusion.
Your email address*