Yaya Heryadi

A method for dance motion recognition and scoring using two-layer classifier based on conditional random field and stochastic error-correcting context-free grammar

This paper presents a unified framework for recognizing and scoring dance motion using 2-layer classifier so that computation complexity is distributed into two layers. This research examines the performance of sliding window, hidden Markov Model (HMM) and conditional random field (CRF) as the first layer classifier to segment the input video into a sequence of motion primitive label. The second layer classifier is stochastic error-correcting context-free grammar, built based on dance master knowledge, to parse the sequence of labels, builds a parse tree, and computes the accumulated dance score. The dataset for this research is captured using one Kinect camera. The training dataset is: 212 samples of 12 motion primitive samples and seven videos of Pendet dance performance. From 5-fold cross-validation, accuracy of sliding window, HMM, and CRF are 0.63, 0.79, and 0.86 respectively. This result shows that CRF achieves higher performance as a dance motion primitive recognizer than HMM as proposed by [1]. The CRF model achieves 0.88 of accuracy when motion feature is all skeleton joint angular coordinates as proposed by [2] but increases to 0.93 if the motion feature is only upper-body joint coordinates. Stochastic error-correcting context-free grammar is chosen as dance choreography model. The experiment using synthetic sequence label with cost factor ci=1 and error-sequence labels up to 50 percent shows the grammar can tolerate the input label sequence error up to 25 percent. The experiment using Pendet dance performances show that the average dance score is 79.3. The low dance score is due to several factors including: dance skill variation, unstable basic gesture repetition, high cost contributed by replacing deletion and substitution of local error by insertion operation, duration variation due the absence of timing guideline of body part motions, and limited training dataset to capture possible basic gesture variations.

Weather forecasting using deep learning techniques

Weather forecasting has gained attention many researchers from various research communities due to its effect to the global human life. The emerging deep learning techniques in the last decade coupled with the wide availability of massive weather observation data and the advent of information and computer technology have motivated many researches to explore hidden hierarchical pattern in the large volume of weather dataset for weather forecasting. This study investigates deep learning techniques for weather forecasting. In particular, this study will compare prediction performance of Recurrence Neural Network (RNN), Conditional Restricted Boltzmann Machine (CRBM), and Convolutional Network (CN) models. Those models are tested using weather dataset provided by BMKG (Indonesian Agency for Meteorology, Climatology, and Geophysics) which are collected from a number of weather stations in Aceh area from 1973 to 2009 and El-Nino Southern Oscilation (ENSO) data set provided by International Institution such as National Weather Service Center for Environmental Prediction Climate (NOAA). Forecasting accuracy of each model is evaluated using Frobenius norm. The result of this study expected to contribute to weather forecasting for wide application domains including flight navigation to agriculture and tourism.

Stochastic regular grammar-based learning for basic dance motion recognition

This paper presents a simple and computationally efficient framework for 3D dance basic motion recognition based on syntactic pattern recognition. In this research, a class of basic dance motions is modeled by a stochastic regular grammar (SRG), inferred from training dataset, in which key body poses that are learned from training dataset are selected as gesture primitives. To represent a dance motion, body pose of a dancer is represented by angular coordinate of 15 skeleton joints. This feature is later compacted into one-dimensional string of labels for grammar inference which makes the recognition process is considerably fast compared to statistical pattern classifier such as k-nearest neighbor (kNN). A single test using the learned grammar in average takes only about 5 ms compared to around 20s using kNN whilst the overhead time to build all grammars takes only about 3.4s. This compacting process, however, leads to information loss which is observed in slightly degraded recognition performance for low articulated motions but quite large degradation for high articulated dance motions. To overcome this, we investigate several reliable feature selection methods such as Sequential Feature Selection (SFS), Principal Component Analysis (PCA), and Heuristic Sequential Feature Selection (HSFS) compared to the use of whole features. Based on our experiment, the HSFS is the most suitable feature selection to overcome this problem.

Weather Forecasting Using Merged Long Short-Term Memory Model (LSTM) and Autoregressive Integrated Moving Average (ARIMA) Model

Weather forecasting is an interesting research problem in flight navigation area. One of the important weather data in aviation is visibility. Visibility is an important factor in all phases of flight, especially when the aircraft is maneuvering on or close to the ground, i.e., during taxi-out, take-off and initial climb, approach and landing and taxi-in. The aim of these study is to analyze intermediate variables and do the comparison of visibility forecasting by using Autoregressive Integrated Moving Average (ARIMA) and Long Short Term Memory (LSTM) Model. This paper proposes ARIMA model and LSTM model for forecasting visibility at Hang Nadim Airport, Batam Indonesia using one variable weather data as predictor such as visibility and combine with another variable weather data as moderating variables such as temperature, dew point and humidity. The models were tested using weather time series data at Hang Nadim Airport, Batam Indonesia. This research compares the Root Mean Square Error (RMSE) resulted by LTSM model with the RMSE resulted by ARIMA model. The results of this experiment show that LSTM model with/or without intermediate variable has better performance than ARIMA Model.

Quality measurement of android messaging application based on user experience in Microblog

There are many options of android messaging application which give opportunity to user in order to choose which one as best or famous android messaging application and make it become suitable for them. Usually, people used to look at the information about best or famous android messaging application by texting in search engine such as google and get some link information from user/blogger reviews, and based on that reviews they will make decisions which one as suitable for them. We proposed the other way how to measure the quality of each android messaging application based on user experience which they text in Microblog such as Twitter. The unstructured data in the Microblog will be processed with 2 operators for sentiment analysis method in RapidMiner such as AYLIEN and ROSETTE. AYLIEN sentiment analysis has 3 categories such as positive, negative, and neutral, whilst ROSETTE sentiment analysis has 2 categories such as positive and negative sentiments. Finally, the finding sentiment analysis with these 2 operators will be compared with PlayStore review.

Smartphone sensors selection using decision tree and KNN to detect head movements in Virtual Reality Application

There are a lot of Virtual Reality applications on smartphone because nowadays smartphones are equipped with varies sensors. These sensors can be used to detect head movements in order to navigate the virtual world. Selecting the combinations of smartphone sensors that are suitable for detecting head movements is still in the grey area because there is no guidance for selecting the sensors and different researchers use different sets of sensors in smartphone. If the selection of relevant sensors has not been chosen wisely, the detection accuracy will not optimal. Detection Knowledge Algorithm with the combination of machine learning are the perfect combination for detecting head movement. Decision Tree and KNN methods are chosen because these methods are able to run on smartphone. Based on the experiment, accelerometer, gyroscope, and magnetometer combination has the highest accuracy result compared with other and suitable to be used in virtual reality application.

Recognizing debit card fraud transaction using CHAID and K-nearest neighbor: Indonesian Bank case

This paper presents a preliminary study on debit card fraud transaction recognition, whose cards are issued by Indonesian bank, based on actual ATM transaction records. The premise of this research is fraudulent transaction contains ‘anomaly’ from the pattern of non-fraudulent transactions so that the anomalous pattern can be detected and separated at some point using classification models. Less availability dataset for research, non-stationary distribution of the data, highly imbalanced class distributions, and continuous streams of transactions become the main driven of using CHAID and k-NN classification method. Empiric result using actual debit card transaction using ATM services shows that Accuracy of CHAID model is 0.8 and F = 0.7; and k-NN model (for k=3) is 0.7 and F = 0.6 These results are comparable to previous studies using Hidden Markov Models.

Smart city's context awareness using social media

The increasing number of inhabitants of a city, there will be more challenges in the management of the city. Many events that can not be controlled by either causing the slow response of the relevant institutions. Sensing the smart city through social media is offered for such a solution. Text mining is done to analyze the social media posts based on events that occurred and the emotion that follows is based on text, hashtag and geo-tagging. Methodology used is text mining approach kernel methods, particularly the support vector machine (SVM). Results are expected with this concept is the city that can listen to the aspirations and desires of the population quickly and accurately.

User experience evaluation of virtual reality-based cultural gamification using GameFlow approach

This paper presents some empiric results on the effect of some game player profiles to their user experience. The cultural gamification prototype for this study is developed using Virtual Reality technology on Google Cardboard in which some elements of Indonesian culture is used in part of the game story. Dataset for the game evaluation is collected using survey method. Data collection is implemented using self-administered questionnaire which is developed based on GameFlow framework. The respondent population is the college students who like to play games. The survey samples are chosen using purposive sampling technique. The results of this study indicate that there is a strong association of playing frequency and personality traits to game user experiences. Also, there is a weak association between gender and age to game user experiences. These research findings are valuable for game designer to design enjoyable and educational games.

Gamification of M-learning Mandarin as second language

This paper presents some empiric results on teaching basic Mandarin as second language to college students using gamification approach. This study shows some evidences that gamification outperform non-gamification teaching method in related to learning concentration, skills, feedback, and immersion. However, gamification method has some drawbacks in related to compliance of quiz difficulty with the level/learning progress, clarity of overall learning goal, and clarity of goal at each level. These strength and drawbacks of gamification design are expected to give valuable information for teachers in delivering Mandarin as second language.