Wavelet DB44 and MBB Algorithm for Sasak Vowels Recognition

Ancient manuscripts are original handwriting that is at least 50 years old and has significance for civilization, history, culture, and science. In the preservation of ancient manuscripts, an attempt was made to digitize ancient documents. The hope is that by carrying out this process, the old documents can be used in historical learning and increase the interest of the younger generation in studying history. So far, the attempt to digitize ancient manuscripts is still limited to the storage of manuscripts in the form of digital documents, generally in the way of image files. So, the existing digital manuscripts are still static, while each manuscript has specificity in how to read them. Therefore, this research is a preliminary study for digitizing ancient documents, especially voice-based Sasak language. The vowel sound of Sasaknese speakers used in this research. Wavelet db44 and WPCC used as the feature extractor, and FCM implemented to build the feature reference/model. The accuracy of the recognition is evaluated using the DTW algorithm. From the evaluation obtained that the average recognition accuracy of the system using training dataset is 50% while using testing dataset is 27.14%.