Indonesian Word Recognition System using LabVIEW for Controlling Equipment in the Lecture Hall
This paper describes the implementation of the Indonesian word recognition system using LabVIEW 7.1 programming language for controlling some common equipment in the lecture hall, the lights, projectors, air conditioning and doors. Word recognition system consists of two subsystems, feature extraction subsystem and pattern matching subsystem. Feature extraction subsystem will convert input sound signal into a feature vector for recognition. The feature vector will then be compared with the characteristic vectors stored in a database through pattern matching subsystem. To obtain good performance, so in this study compared the performance of several feature extraction methods namely Mel Frequency Cepstrum Coefficients (MFCC), MFCC+Delta and MFCC+Delta+Double Delta. For pattern matching method also performed comparison of the performance of the Dynamic Time Warping (DTW) method and Vector Quantization (VQ) method. The word used is limited to 11 pieces of the Indonesian language. Speakers in this study is limited to two male speakers and two female speakers. The research focuses on the recognition accuracy at the word level. The results showed that the best word recognition is equal to 98.2% is obtained using the MFCC + Delta + Double Delta as feature extraction and DTW as pattern matching methods.