A peak deconvolution procedure used for the analysis of data corresponding to simultaneous overlapping processes begins with separation of individual processes using functions such as Gaussian, Lorentzian, Weibull, and Fraser-Suzuki (FS) followed by application of kinetic analysis methods to the separated peaks. We propose a coupled peak deconvolution procedure to link the parameters of the FS functions of similar peaks in two DTG curves obtained at different linear heating rates, so that the coordinates of each peak can be obtained in a constrained manner. The proposed technique is a kinetic deconvolution method rather than a pure mathematical deconvolution technique. To analyze individual peaks in our study, the non-parametric kinetic and Freidman's isoconversional methods have been applied to determine kinetic triplet of each process. This technique has been tested with both simulated and experimental data. Using this technique, the effects of molecular weight and degree of hydrolysis of polyvinyl alcohol (PVA) samples on reaction mechanism and activation energy of thermal degradation were studied. The presence of acetate group in the PVA samples causes thermal stability, decreases the rate of main reactions, and increases the activation energy. The results of this study may help tailor heat-resistant materials with proper choice of polymer characteristics.