| Judul | Abstract | Halaman |
|---|
| Characters Segmentation of Cursive Handwritten Words based on Contour Analysis and Neural Network Validation | This paper presents a robust algorithm to identify the letter boundaries in images of unconstrained handwritten word . The proposed algorithm is based on vertical contour analysis. Proposed algorithm is performed to generate presegmentation by analyzing the vertical contours from right to left. The unwanted segmentation points are reduced using neural network validation to improve accuracy of segmentation. The neural network is utilized to validate segmentation points. The experiments are performed on the IAM benchmark database. The results are showing that the proposed algorithm capable to accurately locating the letter boundaries for unconstrained handwritten words. | 1-16 |
| Free Model of Sentence Classifier for Automatic Extraction of Topic Sentences | This research employs free model that uses only sentential features without paragraph context to extract topic sentences of a paragraph. For finding optimal combination of features, corpus-based classification is used for constructing a sentence classifier as the model. The sentence classifier is trained by using Support Vector Machine (SVM). The experiment shows that position and meta-discourse features are more important than syntactic features to extract topic sentence, and the best performer (80.68%) is SVM classifier with all features. | 17-34 |
| The Effectiveness of Chosen Partial Anthropometric Measurements in Individualizing Head-Related Transfer Functions on Median Plane | Individualized head-related impulse responses (HRIRs) to perfectly suit a particular listener remains an open problem in the area of HRIRs modeling. We have modeled the whole range of magnitude of head-related transfer functions (HRTFs) in frequency domain via principal components analysis (PCA), where 37 persons were subjected to sound sources on median plane. We found that a linear combination of only 10 orthonormal basis functions was sufficient to satisfactorily model individual magnitude HRTFs. It was our goal to form multiple linear regressions (MLR) between weights of basis functions acquired from PCA and chosen partial anthropometric measurements in order to individualize a particular listener’s H RTFs with his or her own anthropometries. We proposed a novel individualization method based on MLR of weights of basis functions by employing only 8 out of 27 anthropometric measurements. The experiments’ results showed the proposed method, with mean error of 11.21%, outperformed our previous works on individualizing minimum phase HRIRs (mean error 22.50%) and magnitude HRTFs on horizontal plane (mean error 12.17%) as well as similar researches. The proposed individualization method showed that the individualized magnitude HRTFs could be well estimated as the original ones with a slight error. Thus the eight chosen anthropometric measurements showed their effectiveness in individualizing magnitude HRTFs particularly on median plane. | 35-56 |
| Digital Dermatoscopy Method for Human Skin Roughness Analysis | In this study we propose a digital dermatoscopy method to measure the human skin roughness. By using this method we eliminate the use of silicon replica. Digital dermatoscopy consists of handheld digital microscope, image processing and information extraction of skin roughness level. To reduce the noise due to the variation of reflection factor on the skin we use median filter. Hence, by Fourier transform the skin texture is imaged in terms of 2D frequencyspatial distribution. Skin roughness is determined from its entropy, where the roughness level is proportional to the entropy. Three types of experiment have been performed by evaluating: (i) the skin replicas; (ii) young and elderly skin; and (iii) seven volunteers treated by anti wrinkle cosmetic in three weeks period. We find that for the first and second experiment that our system did manage to quantify the roughness, while on the third experiment, six of seven volunteers, the roughness are succeeded to identify. | 57-71 |
