On the use of the QOBE algorithm central and noncentral estimators in speech classification

The quasioptimal bounding ellipsoid (QOBE) algorithms have emerged as potential candidates in many digital signal processing applications for their efficiency in estimating the parameter set of an unknown system. In this work we investigate the use of the QOBE noncentral estimator for identifying an unknown vowel sound from a set of known vowels. We select estimates from different quadrants and distances within the QOBE final ellipsoid to find the best estimates. We define the best estimates as those estimates that result in the highest amount of accuracy and precision according to predefined criterion. We conduct simulations to evaluate the performance of the best QOBE estimates in vowel identification. We conclude on the different selections by comparing the accuracy of the speech identification. We also conclude that the use of QOBE noncentral estimates results in accurate vowel identification.