Classification of Fricatives Using Novel Modulation Spectrogram Based Features

In this paper, we propose the use of a novel feature set, i.e., modulation spectrogram for fricative classification. Modulation spectrogram gives 2-dimensional (i.e., 2-D) feature vector for each phoneme. Higher Order Singular Value Decomposition (HOSVD) is used to reduce the size of large dimensional feature vector obtained by modulation spectrogram. These features are then used to classify the fricatives in five broad classes on the basis of place of articulation (viz., labiodental, dental, alveolar, post-alveolar and glottal). Four-fold cross-validation experiments have been conducted on TIMIT database. Our experimental results show 89.09 % and 87.51 % accuracies for recognition of place of articulation of fricatives and phoneme-level fricative classification, respectively, using 3-nearest neighbor classifier.