Design of wavelet based phone pass filter (ppf) using genetic algorithm

The paper introduces new direction of handling phone classification problem. The problem is treated by designing a filter like algorithm to separate certain phone or a group of phones from the complete set. The designed filter conveys a good information regarding the phone it extracts. The speech data that is introduced to the filters are prepared by a certain speech model that is based on wavelet transform. This model is introduced into Amr M. Gody (2003) and Amr M. Gody (2002). A brief illustration of the model is included in the introduction section. The work is applied on the Arabic language phone set. A database of many hours of Arabic dialogs is used. The database is recorded from 30 male speakers. Genetic algorithm is utilized in phone pass filter (PPF) parameters design. The 6 coefficients filters are analyzed to extract all the possible phone properties. The speaker dependency is localized. An average PPF extraction efficiency of 94% is achieved over the trustable dataset