Rangkuti, A.H., Harjoko, A., and Putra, A.E.

This research evolves feature extraction algorithms in overcoming difficulties in classifying batik images that encounter changes in scale and rotation. the algorithm is multiscale and multilevel extended center symmetric local binary pattern (M2ECS-LBP). In utilizing this algorithm using several types of windows to obtain optimal feature extraction results, ranging from the size of 6×6, 9×9, 12 x 12 and 15×15 or a combination of several windows. However, for the use of algorithm carried out sequentially, it also requires a special strategy to obtain optimal image feature extraction results to support performance accuracy in the classification. The results of classification accuracy using K-Nearest neighborhood had reached up until the percentage to 78,4 – 81.7 percent of the image undergoing changes in scale and rotation. However, if the batik image undergoes a change in scale but the rotation is the same then the accuracy percentage can reach 98-99 percent. This algorithm is a very powerful breakthrough with lighter computing techniques than other algorithms. This research can be continued to recognize moving images, expected with maximum accuracy.

[click here]

Tags: , ,

Random Forest (RF) is known as one of the best classifiers in many fields. They are parallelizable, fast to train and to predict, robust to outlier, handle unbalanced data, have low bias, and moderate variance. Apart from these advantages, there are still opportunities to increase RF efficiency. The absence of recommendations regarding the number of trees involved in RF ensembles could make the number of trees very large. This can increase the computational complexity of RF. Recommendations for not pruning the decision tree further aggravates the condition. This research attempts to build an efficient RF ensemble while maintaining its accuracy, especially in problem activity. Data collection is performed using an accelerometer sensor on a smartphone device. The data used in this research are collected from five peoples who perform 11 different activities. Each activity is carried out five times to enrich the data. This study uses two steps to improve the efficiency of the classification of the activity: 1) Optimal splitting criteria for activity classification, 2) Measured pruning to limit the tree depth in RF ensemble. The first method in this study can be applied to determine the splitting criteria that are most suitable for the classification problem of activities using Random Forest. In this case, the decision model built using the Gini Index can produce the highest accuracy. The second method proposed in this research successfully builds less complex pruned-tree without reducing its classification accuracy. The research results showed that the method applied to the Random Forest in this study was able to produce a decision model that was simple but yet accurate to classify activity.

[https://dx.doi.org/10.14569/IJACSA.2019.0100658]

Tags: , , , , ,

Heriyanto, Hartati and Putra, 2018

The recitation of the Qur’an has its own uniqueness, among others having a special rule in reading and pronunciation, which is called tajwid science. At the time of the Qur’an is recited, there are often mistakes due to the limitations of knowledge of Tajwid. Therefore, the availability of tools to facilitate in checking the appropriateness of recitation is very much needed by those who recite the Qur’an and face limitations in understanding the science of tajwid. Checking the Qur’an reading is a problem that must be solved according to the rules. So far, voice identification studies have problems with feature extraction, compatibility or suitability testing, and accuracy. The issue of feature extraction, suitability, and impermanence testing have been improved in this study, which consists of two stages. The first stage is the extraction of the sound character of the Qur’an reading and the second stage is the testing of the conformity of the Qur’anic recitation and accuracy. In the first stage feature extraction is handled using MFCC and Normalization of Dominant Weight (NDW). Characteristics of reading the Qur’an as reference table is taken from one reader of Al-Qur’an who has competence in the field of science tajwid, for sampling 5-7 people as a source for testing. The process of the second stage of conformity testing of Qur’an reading is done starting from filtering, sequential multiplication of reference table and Conformity Uniformity Pattern (CUP). The sample of reading conformity test is taken from 11 Qur’anic letters containing 8 reading laws and 886 records. The test is performed on the dominant frame, the number of cepstral coefficient and the number of frames. The reading conformance test provides an average accuracy of 91.37% on the nine dominant frames. The test for the number of cepstral coefficients in the c-23 can be an average of 96.65%, while the number of frames on the F-10 is the best average of 96.65%.

[https://doi.org/10.14738/aivp.62.4268]

Tags: , ,

Oleh Agfianto Eko Putra, Unggul Adhi Nugroho, Bakhtiar Alldino Ardi Sumbada dan Catur Atmaji

Abstrak

Telah dirancang-bangun penerima  paket APRS  berbasis  Raspberry  Pi  2  untuk  stasiun bumi.Tuner  TV  digunakan  sebagai  penerima  sinyal,  alat  penerima  diakses  melalui  laptop secara nirkabel. Antena Yagi dengan sebuah pengendali digunakan agar dapat secara otomatis mengarahkanke  satelit.Ujicoba  dilakukan  dengan  menerima  paket  APRS  yang  dipancarkan digipeater  satelit  International  Space  Station  (ISS)  dan  satelit  LAPAN-A2.  Hasil  penelitian menunjukkan  bahwa alat ini mampu  mendapatkan paket  APRS  satelit  ISS  dengan  jumlah  6 paket dari 10 paket yang dipancarkan. Paket yang diterima memiliki rata-rata amplitudo pada frekuensi 1.200 Hz dan 2.200 Hz yang bernilai jauh lebih kecil dibandingkan amplitudo audio keseluruhan.  Hal  ini  menunjukkan  bahwa  terdapat  derau  yang  tinggi  pada  sinyal.  Sedangkan paket APRS dari satelit LAPAN-A2 belum berhasil diperoleh.

(informasi lebih lanjut klik https://doi.org/10.22146/ijeis.44299)

Tags: , , , , ,

by Endang Anggiratih and Agfianto Eko Putra

Abstract

Ship identification on satellite imagery can be used for fisheries management, monitoring of smuggling activities, ship traffic services, and naval warfare. However, high-resolution satellite imagery also makes the segmentation of the ship difficult in the background, so that to handle it requires reliable features so that it can be identified adequately between large vessels, small vessels and not ships. The Convolution Neural Network (CNN) method, which has the advantage of being able to extract features automatically and produce reliable features that facilitate ship identification. This study combines CNN ZFNet architecture with the Random Forest method. The training was conducted with the aim of knowing the accuracy of the ZFNet layers to produce the best features, which are characterized by high accuracy, combined with the Random Forest method. Testing the combination of this method is done with two parameters, namely batch size and a number of trees. The test results identify large vessels with an accuracy of 87.5% and small vessels with an accuracy of not up to 50%.

(for more information please click https://doi.org/10.22146/ijccs.37461)

Tags: , , , ,

by Agfianto Eko PUTRA, Wiwit SURYANTO, Agung Nugraha SULISTYANA

Seismic data analysis of the 2006 Merapi volcano eruption has been carried out using the Continuous Wavelet Transform (CWT) and the Wavelet-based Decomposition and Correlation (WAVEDECOR) combined with the Fast Fourier Transform (FFT). The CWT is used to show the frequency pattern of the event while the WAVEDECOR is used to denote the frequency band of the signal. The CWT and the WAVEDECOR are supported by the FFT to ensure the dominant frequency of the observed signals. The results show that visual patterns and dominant frequency distribution of certain events, including the VT-A, the Low Frequency (LF), the VT-B, tremor, multiphase and lava avalanche. The result from this analysis was then compared with related eruption signal of Merapi in 1996 to determine the pattern similarity. The comparison results show almost identical results for dominant frequencies in VT-A events as well as MP events. The findings in the VT-B event showed that the dominant frequency pattern was slightly different from the 1996 data which showed at medium to high-frequency while for the 2006 data showed only at a medium frequency.

[click here]

Tags: , ,

Achmad Solichin, Agus Harjoko, and Agfianto Eko Putra

This study proposed a model for determining the movement direction of the object based on the optical flow features. To increase the speed of computational time, optical flow features derived into a Histograms of Oriented Optical Flow (HOOF). We extracted them locally on the grid with a certain size. Moreover, to determine the movement direction we also analyzed multiple frames at once. Based on the experiment results, showing that the value of accuracy, precision, and recall of the movement detection is good, amounting to 93% for accuracy, 73.07% for precision and 84.25% for recall. Furthermore, the results of testing using the best parameter shows the value of accuracy of 98.1%, 35.6% precision, 41.2% recall, and direction detection error rate (DDER) 25,28%. The results of this study are expected to provide benefits in video analysis studies such as riots detection and abnormal movement in public places.

[click here]

Tags: , , , ,

Agfianto Eko Putra, Catur Atmaji, and Fajrul Ghaleb

In the area of affective computing technology, the classification of emotions can be used for a variety of things such as health, entertainment, education, etc. This study determined the classification of emotions based on EEG (Electroencephalography) signals, which is emotions are classified according to the 2-dimensional graphics modeling of arousal and valence. This research uses a wavelet decomposition method to get features from the EEG signal. Features taken from the signal is a power signal decomposition of sub-band theta, alpha, beta, and gamma. These features are derived from the 5 levels decomposition using Coiflet2 and Daubechies2 mother wavelet. Classification is done using k-Nearest Neighbor (kNN) with the closest neighbor calculated based on correlation distance. Data validation is done using 5-folds cross-validation for validation of test data and training data. The highest accuracy obtained by using the mother wavelet Coiflet 2 with kNN parameter k=21. Valence classification accuracy is 57.5%, and accuracy of arousal is 63.98%.

[click here]

Tags: , , ,

Catur Atmaji, Agfianto Eko Putra, and Irvan Albab Tontowi

Many research on how the human brain works have been done in the last century. The use of electroencephalogram signal generated from quantifying the brainwave has been developed in many areas including the development of brain-computer interface (BCI) concept. One type of BCI that interesting for the future use is motor imagery (MI) based-BCI which only requiring imagination of a person to control an object. This study proposed a feature extraction in eight different channels using discrete wavelet (DWT) coefficients. The wavelet coefficient is transformed to a frequency domain using discrete Fourier transform (DFT) and then average power spectrum is calculated. Level 5 of detail component of the DWT is chosen because, from 512Hz sampling frequency (8 - 16Hz), it resembles mu rhythm of brain wave (8 - 12Hz) which affected from motor imagery activity. The classification of three classes, which are the imagination of right body movement, left the movement, and random word using multiclass support vector machine (SVM) shows a promising result with a sensitivity of 96.88%, 86.12% and 52.78% from three different subjects.

[click here]

Tags: , , , ,

Suwanto Raharjo, Retantyo Wardoyo, Agfianto Eko Putra

Proper nouns (often abbreviated PN or NNP) are a class of words important in labeling and subsequent text processing, especially in natural language processing (NLP). Name entity recognition (NER) is one study that requires PN. The lack of labeled data for Indonesian text, especially the PN label, may be attributed for the lack of NER research in the Indonesian language. This study aims to detect PN in Indonesian-language translations of the Quran guided by deriving location information from the Quran as its source text. In the Indonesian language, PN is written using initial capital letters, which are used to determine and guide PN location. This article proposes that PN in Indonesian-language translations of the Quran can be determined based on PoS Information of Quranic text by developing a certain production rule. The results of this research showed that the proposed method has promising results for further research.

[click here]

Tags: , , , , ,