Jawi Character Speech-to-Text Engine Using Linear Predictive and Neural Network for Effective Reading

摘要

Jawi is an old version of Malay language writing that need to be preserved. Therefore, it is important to develop tools for teaching kids about Jawi characters and speech-to-text (STT) application can serve this purpose well. Unlike English, Jawi uses special characters similar to Arabic characters. However, its pronunciations are in Malay language. This uniqueness makes STT development a challenging task. In this paper, we investigate the applicability of linear predictive coding to extract important features from voice signal and neural network with backpropagation to classify and recognize spoken words into Jawi characters. A total of 225 samples of words in Jawi characters are recorded from speakers with over 95% accuracy. Jawi characters speech-to-text engine aims to help students to read Jawi document accurately and independently without the need for close monitoring from parents or teachers.