Using Wavelet Extraction for Haptic Texture Classification Waskito Adi 1 , Suziah Sulaiman 1 1 Computer and Information Sciences Department, Universiti Teknologi PETRONAS Bandar Seri Iskandar, 31750 Tronoh, Perak, Malaysia {waskito@gmail.com, suziah@petronas.com.my} Abstract. While visual texture classification is a widely-researched topic in image analysis, little is known on its counterpart i.e. the haptic (touch) texture. This paper examines the visual texture classification in order to investigate how well it could be used for haptic texture search engine. In classifying the visual textures, feature extraction for a given image involving wavelet decomposition is used to obtain the transformation coefficients. Feature vectors are formed using energy signature from each wavelet sub-band coefficient. We conducted an experiment to investigate the extent in which wavelet decomposition could be used in haptic texture search engine. The experimental result, based on different testing data, shows that feature extraction using wavelet decomposition achieve accuracy rate more than 96%. This demonstrates that wavelet decomposition and energy signature is effective in extracting information from a visual texture. Based on this finding, we discuss on the suitability of wavelet decomposition for haptic texture searching, in terms of extracting information from image and haptic information. Keywords: Texture recognition, supervised learning, machine learning, haptic texture search engine, wavelet decomposition. 1 Introduction Lately, computer vision has become one of the most popular research subjects. Computer vision contributes in various fields that include medical, engineering, and robotics. One of the most interesting topics in computer vision is texture recognition. Being the smallest entity, textures could be used as a parameter to recognize a particular object. There are many different types of applications involving texture analysis, including medical imaging, industrial inspection, remote sensing, document segmentation, and computer based image retrieval [1]. Feature extraction plays an important role in a classification process. The effectiveness of such classification relies greatly on the choice of this feature. In this case, a suitable extraction algorithm influences the process end result. Parallel to computer vision, computer haptic is also another area which is gaining its popularity among researchers. Computer haptic enable user to touch and interact with the virtual