Pengenalan Bentuk Benda Berdasarkan Sinyal Suara dengan Transducer Mikrofon dan Teknologi Kinect

Abstract

Voice processing or speech recognition is growing rapidly hence it can be used for various applications such as moving a system or motion control and multimedia-based learning media. Implementation of speech recognition and image detection in this study using microphone transducer and kinect technology. This study aims to produce a system that can identify and recognize an object with word commands, such as circles, triangles, rectangles and many. In sound processing, sound feature extraction is carried out with Mel-Frequency Cepstrum Coeffecient (MFCC). Word modeling was done using statistical modeling, namely the Hidden Markov Model (HMM). HMM is able to provide an efficient mechanism for statistically modeling diversity in words or words. Data were collected with offline and online microphone transducers. This study matches the pattern of words through training and testing process. The output of this system is a recognizable word based on the highest probability and displaying the object shape based on the recognized word, namely circle, triangle and quadrilateral. Test results with microfon tranducers, for 85% trained sources, 81.5% untrained sources, and 84% untrained Kinect source testing hence that word recognition systems can be implemented with Kinect technology.

Keywords

• speech processing
• HMM
• MFCC
• Kinect