Fractal dimension and iterated function system (IFS) for speech recognition

A Study of Speech Fractal Dimensions

“…Robinson (22), estimate that D needs only be constant for a suitable wide range of iterations, 10 or more, for example, which is not achieved in our study, if we consider 3pD which are on the smallest scales. (14) does not give the precise value of this parameter. In our experience it is difficult to overcome differences in amplitudes of recordings; only McDowell mentions how he tried to avoid this risk.…”

Using vector of fractal dimensions for feature reduction and phoneme recognition and classification

“…dimension of the signal is a measure of this irregularity [6].The main argument is that if fractals can effectively model chaotic nature of phenomena and if speech is a natural chaotic phenomenon, then fractals should be a promising model for speech [7]. Phonemes, the basic parts of human speech, show a high degree of repetition and from a fractal point of view, this high degree of repetition could be view as self-similarity and frequency content is the roughness, which is measured by the Fractal Dimension [8].…”

“…Erik and Bohez presented a speech recognition method based on two-level Clustering using these features [8]. Bohez and Van Winden used fractal dimension of speech waveform and the IFS parameters that represent the waveform as the parameters for recognition [6]. Personal authentication for speaker recognition based on Mel-scale spectral dimension and Mel Frequency Cepstral Coefficients (MFCC) or Multi-Scale Fractal Dimension (MFD) were presented in [11], [12].…”