Auditory motivated discrete time frequency signal representation and its application to vowel classification

In this thesis the Auditory Motivated Discrete Time Frequency Signal Representation method is presented. The method is simple and independent from the window function, which affect the obtained time frequency resolution in classical methods. The numerical simulations with different SNR values show that the proposed method is applicable for time frequency signal analysis. The proposed method is applied to the speech vowel signals and similar spectral shapes are obtained from the same vowel signals independent from the speakers, which is good evidence for existing similar spectral shapes inside the same vowels. The vowel classification based on vowel patterns extracted from spectral peaks distribution is performed in order to test the existence of the similar spectral shapes, and the obtained results show that the proposed method can be used to extract additional vowel patterns in speech recognition applications to improve the speech recognition performance.