On the application of quantum clustering on speech data

Abstract

In this work, Quantum clustering (QC) algorithm is applied to a labeled dataset of Arabic vowels. The accuracy and processing time are, then, compared with nonhierarchical kernel approaches for unsupervised clustering; namely, k-means, self-organizing map and fuzzy c-means. The choice of speech data is according to large database statistics which reveal that vowels class represents about 60–70% of Arabic speech whereas the remaining percentage is distributed among other sounds. The analysis features, in this work, are the mel-frequency cepstarl coefficients. The results show that all algorithms are competitive from accuracy point of view while QC still guarantees the solution stability.