A Hierarchical Classifier for Text-Independent Speaker Identification Based on a Comparison of Vowel Waveforms

摘要

In the paper a hierarchical classifier is described that can be used for text-independent speaker identification. Patterns representing voice of particular speakers are formed from the samples of pitch period waveforms of particular Czech vowels. These patterns are compared by two different distance subclassifiers that work independently on the first (lower) level of the hierarchical classifier. To identify an unknown speaker a combiner is used on the second (upper) level of the hierarchical classifier. This combiner allows to take into account distance measures provided at the outputs of the particular subclassifiers on the first level.