INDONESIAN SPEECH RECOGNITION SYSTEM USING DISCRIMINANT FEATURE EXTRACTION – NEURAL PREDICTIVE CODING (DFE-NPC) AND PROBABILISTIC NEURAL NETWORK

FEBRUARY 2012 INDONESIAN SPEECH RECOGNITION SYSTEM USING DISCRIMINANT FEATURE EXTRACTION – NEURAL PREDICTIVE CODING (DFE-NPC) AND PROBABILISTIC NEURAL NETWORK
A THESIS SUBMITTED TO THE GRADUATE SCHOOL OF TELKOM INSTITUTE OF TECHNOLOGY
BY UNTARI NOVIA WISESTY 213100004
Prof. Dr. The Houw Liong Supervisor
Adiwijaya, SSi., MSi Co-Supervisor
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF INFORMATICS IN THE INFORMATICS STUDY PROGRAM
ABSTRACT
Along with advances in information technology, it has been developed the technology to facilitate human life, one of which is speech recognition. Speech recognition is widely applied to speech to text, speech to emotion, in order to make gadget and computer easier to use, or to help people with hearing disability. However, the development of speech recognition to produce the text from the input voice has not well developed because of time processing. This is certainly going to make the animators and engineers need more time using speech recognition. Therefore, it needs a method to solve the time processing problem and with a good accuracy. In this study proposes a speech recognition system using Discriminant Feature Extraction – Neural Predictive Coding (DFE-NPC) as feature extraction and Probabilistic Neural Network as recognition method. This system can accelerate time processing because it only uses one iteration in training process. Time processing of proposed method is decrease significantly until 1:95 compared to Fuzzy Hidden Markov Model. The best accuracy of the system is 100% when number of class is 2 and 3, and the worst one is 56% when number of class is 10. Keywords: Speech Recognition System, DFE-NPC, PNN, time processing.