Signals, frequency spectrum, and reconstructed phase spaces for logistic map with .
The overall structure of the proposed feature extraction method.
Mutual information diagram across different time lags for /a/ phoneme (first minimum is 6).
(a) The speech signal of a sample of /a/ phone. (b) The reconstructed phase space by and (c) its Poincaré section.
Individual -ratio values for 30 different features (sorted by their pruning effects in -ratio values).
The labels assigned to the features stated in the 99-dimensional feature vector.
COR% for isolated phoneme recognition using different feature sets (GMM with four mixtures).
The CSR results for the clean speech signal using MFCC and the proposed features distinctively.
CSR results in details for three different feature sets.
COR% for isolated phoneme recognition for different noisy conditions.
Article metrics loading...
Full text loading...