Chakarida Nukoolkit

Human gesture recognition using Kinect camera

In this paper, we propose a comparison of human gesture recognition using data mining classification methods in video streaming. In particular, we are interested in a specific stream of vector of twenty body-joint positions which are representative of the human body captured by Kinect camera. The recognized gesture patterns of the study are stand, sit down, and lie down. Classification methods chosen for comparison study are backpropagation neural network, support vector machine, decision tree, and naive Bayes. Experimental results have shown that the backpropagation neural network method outperforms other classification methods and can achieve recognition with 100% accuracy. Moreover, the average accuracy of all classification methods used in this study is 93.72%, which confirms the high potential of using the Kinect camera in human body recognition applications. Our future work will use the knowledge obtained from these classifiers in time series analysis of gesture sequence for detecting fall motion in a smart home system.

Heterogeneous ensemble approach with discriminative features and modified-SMOTEbagging for pre-miRNA classification

An ensemble classifier approach for microRNA precursor (pre-miRNA) classification was proposed based upon combining a set of heterogeneous algorithms including support vector machine (SVM), k-nearest neighbors (kNN) and random forest (RF), then aggregating their prediction through a voting system. Additionally, the proposed algorithm, the classification performance was also improved using discriminative features, self-containment and its derivatives, which have shown unique structural robustness characteristics of pre-miRNAs. These are applicable across different species. By applying preprocessing methods—both a correlation-based feature selection (CFS) with genetic algorithm (GA) search method and a modified-Synthetic Minority Oversampling Technique (SMOTE) bagging rebalancing method—improvement in the performance of this ensemble was observed. The overall prediction accuracies obtained via 10 runs of 5-fold cross validation (CV) was 96.54%, with sensitivity of 94.8% and specificity of 98.3%—this is better in trade-off sensitivity and specificity values than those of other state-of-the-art methods. The ensemble model was applied to animal, plant and virus pre-miRNA and achieved high accuracy, >93%. Exploiting the discriminative set of selected features also suggests that pre-miRNAs possess high intrinsic structural robustness as compared with other stem loops. Our heterogeneous ensemble method gave a relatively more reliable prediction than those using single classifiers. Our program is available at http://ncrna-pred.com/premiRNA.html.

Identification of non-coding RNAs with a new composite feature in the Hybrid Random Forest Ensemble algorithm

To identify non-coding RNA (ncRNA) signals within genomic regions, a classification tool was developed based on a hybrid random forest (RF) with a logistic regression model to efficiently discriminate short ncRNA sequences as well as long complex ncRNA sequences. This RF-based classifier was trained on a well-balanced dataset with a discriminative set of features and achieved an accuracy, sensitivity and specificity of 92.11%, 90.7% and 93.5%, respectively. The selected feature set includes a new proposed feature, SCORE. This feature is generated based on a logistic regression function that combines five significant features—structure, sequence, modularity, structural robustness and coding potential—to enable improved characterization of long ncRNA (lncRNA) elements. The use of SCORE improved the performance of the RF-based classifier in the identification of Rfam lncRNA families. A genome-wide ncRNA classification framework was applied to a wide variety of organisms, with an emphasis on those of economic, social, public health, environmental and agricultural significance, such as various bacteria genomes, the Arthrospira (Spirulina) genome, and rice and human genomic regions. Our framework was able to identify known ncRNAs with sensitivities of greater than 90% and 77.7% for prokaryotic and eukaryotic sequences, respectively. Our classifier is available at http://ncrna-pred.com/HLRF.htm.

Survey of Smart Technologies for Fall Motion Detection: Techniques, Algorithms and Tools

The aging population has become a world-wide social concern. The number of people living alone and experiencing falls is increasing. This is a major health risk, especially among the elderly; thus, the early detection of fall motion is of great significance. A smart home care system is needed to monitor abnormal events. This paper first conducts a survey of existing smart systems and techniques in detecting fall motion in human movement, including the emergence of new natural user interface (NUI) devices and systems in the consumer market. Secondly, the paper categorizes smart technologies for fall motion detection into three main technological groups: acoustic and ambient sensor-based, kinematic sensor-based, and lastly the computer vision and NUI. An insightful discussion of each category’s advantages and disadvantages is provided. The findings show a promising research direction of integrating the computer vision with the novel consumer-grade NUI device, such as Kinect, in achieving of an affordable and practical smart home fall motion detection system.

Optical Music Recognition on Android Platform

This paper describes the concept and algorithms used in an optical music recognition application on the Android mobile platform. The application can recognize a scanned image or an image taken from a camera phone of sheet music to be interpreted and exported as a playable melody in both MIDI and MusicXML formats while handling resource utilization on an Android mobile phone platform. Limited processing performance and memory capacity, including the lack of image processing and other related APIs, are major issues that cause the algorithms used in the application to be different from traditional approaches applied in software on a PC platform. The proposed system performed with a 76.03% accuracy rate for the scanned sheet music and 71.43% for the sheet music captured by a mobile phone’s camera, which are quite significant values for a mobile platform with limited resources.

Office workers syndrome monitoring using kinect

Many office workers today sit and work at computers for extended periods of time, which can result in a group of symptoms called “Office Workers Syndrome”. To help prevent these symptoms, we propose a novel system to monitor computer users by using a Kinect camera. Firstly, data mining classification is applied for detection of prolonged sitting, while mathematics that include a spherical coordinate system and geometry as well as threshold models is applied to detect unhealthy postures. Secondly, the system gives an alert to the user when unhealthy postures are detected, via simple popup/voice messages or via an alerting device developed by using a microcontroller. Moreover, this research also focuses on enhancing the user experience, various user-interfaces and data visualization techniques for generating useful summary reports are provided.

Detect the Daily Activities and In-house Locations Using Smartphone

Falls are a key cause of significant health problems, especially for elderly people who live alone. Falls are a leading cause of accidental injury and death. To help assist the elderly, we propose a system to detect daily activities and in-house location of a user by means of a smartphone’s sensor and Wi-Fi access points. We applied data mining techniques to classify activity detection (e.g., sitting, standing, lying down, walking, running, walking up/downstairs, and falling) and in-house location detection. Health risk level configurations (threshold model) are applied for unhealthy activity detection with an alarm sounding and also short messages sent to those who have responsibility such as a caregiver or a doctor. Moreover, we provide various forms of easy to understand visualization for monitoring and include health risk level summary, daily activity summary, and in-house location summary.

Data mining approach for automatic discovering success factors relationship statements in full text articles

In the context of Business-to-Business (B2B), an understanding of inter-organizational success factors and their impacts is crucial for effective strategic management. Several studies regarding those success factors and their influences have been conducted and published as articles. We aim at applying existing techniques, especially data mining, to automatically classify relevant sentences describing an influencing relationship between success factors. This paper presents the experiment method and results to find the optimal data mining workflow for our classification task. In particular, we apply several well-known data mining techniques based on different control factors. Then all discovered models are evaluated and compared to find the optimal data mining workflow. The main contributions include (i) the application of data mining for discovering success factors and their relationships, and (ii) the optimal workflow as a standardized flow for further similar classification tasks. The major challenge of this work is that there exists no mature corpus in this context, and hence our approach is implemented without a supporting corpus. The result shows that the models derived from the workflows that consider a section where a sentence is located perform better than the others in term of average performance. Furthermore, we found that the Support Vector Machine (SVM) performs better than other classifiers.

Optical Music Recognition on Windows Phone 7

Optical Music Recognition (OMR) software currently in the market are not normally designed for music learning and ad hoc interpretation; they usually require scanned input of music scores to perform well. In our work, we aimed to remove this inconvenience by using photos captured by mobile phone’s camera as the input. With the cloud-based architecture and the design without the assumption of perfect image orientation and lighting condition, we were able to eliminate many of the software’s architectural and algorithmic problems while still maintaining an overall decent performance.

Text cohesion visualizer

In this paper, we describe the concept and design of a novel visualization tool to aid in academic writing of English as a Second Language students. The tool makes use of theory in classroom discourse, WordNet API, and linguistics rules given by a linguistics expert, by analyzing English language essays for their linguistic bond counts and links within and between paragraphs. These linguistic indicators reveal the structure and flow of essays, clusters of main ideas, as well as incoherent sentences, which obstruct essay unity. The lack of essay unity is one of the most common writing errors of English as a Second Language learners. The output of the system is shown as several kinds of visualizations that provide writing feedback to users, as well as an autocorrect functionality to improve essay unity. Novice English learners may benefit greatly from this system.