Vowel recognition using discrete Tchebichef transform

• Main Authors: Ernawan, Ferda, Abu, Nor Azman, Herman, Nanna Suryana
• Format: Article
• Language: English
• Published: Asian Network For Scientific Information 2013
• Subjects: T Technology (General); TK Electrical engineering. Electronics Nuclear engineering
• Online Access: http://eprints.utem.edu.my/id/eprint/23051/2/465-471.pdf; http://eprints.utem.edu.my/id/eprint/23051/; http://docsdrive.com/pdfs/ansinet/jas/2013/465-471.pdf
• Institution: Universiti Teknikal Malaysia Melaka

Spectrum analysis has become an elementary operation in vowel recognition. Fast Fourier Transform (FFT) has been used as a famous technique to analyze frequency spectrum of the signal in vowel recognition. Traditionally, vowel recognition required large FFT computation on each window. This study has proposed the Discrete Tchebichef Transform (DTT) as a possible alternative to the popular FFT. DTT has had lower computational complexity and it did not require complex transform with imaginary numbers. This study has proposed an approach based on 256 DTT for efficient vowel recognition. The method used a simplify set of recurrence relation matrix to compute within each window. Unlike the FFT, DTT has provided a simpler matrix setting which involves real coefficient numbers only. The experiment on vowel recognition using 256 DTT, 1024 DTT and 1024 FFT has been conducted to recognize five vowels. The experimental results have indicated the practical advantage of 256 DTT in terms of spectral frequency and time taken for vowel recognition performance. 256 DTT has been produced frequency formants that were relatively similar output of 1024 DTT and 1024 FFT in terms of vowel recognition. The 256 DTT has become potential to be a competitive candidate for computationally efficient dynamic vowel recognition.