Anshu Chittora - Academia.edu (original) (raw)
Address: Gandhinagar, Gujarat, India
less
Related Authors
Graduate Center of the City University of New York
Uploads
Papers by Anshu Chittora
2013 International Conference Oriental COCOSDA held jointly with 2013 Conference on Asian Spoken Language Research and Evaluation (O-COCOSDA/CASLRE), 2013
In this paper, a new feature extraction technique based on modulation spectrogram is proposed. Mo... more In this paper, a new feature extraction technique based on modulation spectrogram is proposed. Modulation spectrogram gives a 2-dimensional (2-D) feature set for each obstruent segment. Since the size of feature vector given by modulation spectrogram is of very large dimension, Higher Order Singular Value Decomposition (HOSVD) theorem is used to reduce the size of feature vector. The reduced feature vector is then applied to a classifier, which classify the obstruent in three broad classes, viz., stop, affricate and fricative. Four-fold cross-validation experiments have been conducted on TIMIT database to find accuracy (in %) of obstruent classification at phoneme-level and recognition of manner of articulation of obstruents. Our results show 92.22 % and 94.85 % accuracies for obstruent classification at phoneme-level and recognition of manner of articulation of obstruents, respectively, using 3-nearest neighbor classifier while with same experimental setup Mel Frequency Cepstral Co...
2014 International Conference on Asian Language Processing (IALP), 2014
In this paper, feature derived from the glottal inverse filtering of the speech signal is used fo... more In this paper, feature derived from the glottal inverse filtering of the speech signal is used for classification of pathological infant cries. Glottal inverse filtering is used to estimate the glottal volume velocity waveform (i.e., the source of voicing for infant cry). Here, GIF is used to separate the glottal source and vocal tract filter. The source and the filter features are used for classification of pathological cries. Through the experimental results, importance of both the features in cry classification is investigated. State-of-the-art feature set, viz., Mel Frequency Cepstral Coefficients (MFCC) is also used to compare performance of the proposed feature set. Experimental results show classification accuracy of 76.28 % with the proposed features as opposed to state-of-the-art, MFCC feature which shows classification accuracy of 71.13 %. Fusion of the proposed feature set with MFCC gives classification accuracy of 78.35 % indicating that proposed feature captures the complimentary information in infant cry signal. All experiments were conducted with SVM classifier with radial basis function (RBF) kernel. Keywords-Glottal inverse filtering (GIF), iterative adaptive inverse filtering (IAIF), glottal volume velocity waveform, vocal tract filter, linear prediction coefficients (LPC), support vector machine(SVM) classifier. I.
2013 International Conference Oriental COCOSDA held jointly with 2013 Conference on Asian Spoken Language Research and Evaluation (O-COCOSDA/CASLRE), 2013
In this paper, a new feature extraction technique based on modulation spectrogram is proposed. Mo... more In this paper, a new feature extraction technique based on modulation spectrogram is proposed. Modulation spectrogram gives a 2-dimensional (2-D) feature set for each obstruent segment. Since the size of feature vector given by modulation spectrogram is of very large dimension, Higher Order Singular Value Decomposition (HOSVD) theorem is used to reduce the size of feature vector. The reduced feature vector is then applied to a classifier, which classify the obstruent in three broad classes, viz., stop, affricate and fricative. Four-fold cross-validation experiments have been conducted on TIMIT database to find accuracy (in %) of obstruent classification at phoneme-level and recognition of manner of articulation of obstruents. Our results show 92.22 % and 94.85 % accuracies for obstruent classification at phoneme-level and recognition of manner of articulation of obstruents, respectively, using 3-nearest neighbor classifier while with same experimental setup Mel Frequency Cepstral Co...
2014 International Conference on Asian Language Processing (IALP), 2014
In this paper, feature derived from the glottal inverse filtering of the speech signal is used fo... more In this paper, feature derived from the glottal inverse filtering of the speech signal is used for classification of pathological infant cries. Glottal inverse filtering is used to estimate the glottal volume velocity waveform (i.e., the source of voicing for infant cry). Here, GIF is used to separate the glottal source and vocal tract filter. The source and the filter features are used for classification of pathological cries. Through the experimental results, importance of both the features in cry classification is investigated. State-of-the-art feature set, viz., Mel Frequency Cepstral Coefficients (MFCC) is also used to compare performance of the proposed feature set. Experimental results show classification accuracy of 76.28 % with the proposed features as opposed to state-of-the-art, MFCC feature which shows classification accuracy of 71.13 %. Fusion of the proposed feature set with MFCC gives classification accuracy of 78.35 % indicating that proposed feature captures the complimentary information in infant cry signal. All experiments were conducted with SVM classifier with radial basis function (RBF) kernel. Keywords-Glottal inverse filtering (GIF), iterative adaptive inverse filtering (IAIF), glottal volume velocity waveform, vocal tract filter, linear prediction coefficients (LPC), support vector machine(SVM) classifier. I.