Pronunciation Detection for Foreign Language Learning Using MFCC and SVM

Abstract

As technology improves, people around the world are given more effective tools to communicate with each other. This has caused a sensation of secondary language learning. Many countries have now included this as an obligatory component of their education systems. However, the lack of appointing right professionals has led to misleading the practicing the pronunciation of the new language, because students often follow the pronunciation that non-native teachers have. This paper aims to provide a model that has a potential to help learners with increasing the recipient for understanding the speaker. The model records the learner’s English pronunciation of a given context, analyses it and provides feedback on the screen. The system has shown an accuracy of 98.3%. Throughout the research we have discovered that several factors such as the learner’s predefined accent from his mother-tongue language, the noise level of an environment where the learner uses the system as well as different types of English accents interfere with providing accurate feedback to the learner.