Application of Linear Predictive Code (LPC) Feature Extraction for Reading Arabic Numbers with Support Vector Machine (SVM) Heriyanto, Budi Suyanto, Tunjung Wahyu Widayati
UNIVERSITAS PEMBANGUNAN NASIONAL VETERAN YOGYAKARTA
Abstract
The pronunciation of the Arabic language is related to checking the suitability of the reading of the Arabic numerals numbers one to ten which are often read by Muslims. This research consists of three stages. The first stage, extraction of reading sound characteristics using Linear Predictive Code (LPC). The second stage is selecting the features that will be used as a feature table. The third stage is testing by checking the suitability of the reading of Arabic numbers applied in the recitation at the Nurul Huda Mosque. Testing checks the suitability of selecting the right features to the number of cepstral coefficients and frames. The features from LPC extraction are in the form of frame features and the average cepstral coefficient in each feature frame is directly checked with the reader with a reference. Research on the sound of reading Arabic numbers as well as feature extraction, feature selection, and checking the suitability of readings in application using LPC feature extraction at the Nurul Huda Mosque reached 85% with Support Vector Machine (SVM).
