Macario O. Cordel

SMO-based System for identifying common lung conditions using histogram

A radiograph is a visualization aid that physicians use in identifying lung abnormalities. Although digitized X-ray images are available, diagnosis by a medical expert through pattern recognition is done manually. Thus, this paper presents a system that utilizes machine learning for pattern recognition and classification of three lung conditions, namely Normal, Pleural Effusion and Pneumothorax cases. Using two histogram equalization techniques, the designed system achieves an accuracy rate of 76.19% and 78.10% by using Sequential Minimal Optimization (SMO).

Fast Emulation of Self-organizing Maps for Large Datasets☆

Abstract The self-organizing map (SOM) methodology does vector quantization and clustering on the dataset, and then projects the obtained clusters to a lower dimensional space, such as a 2D map, by positioning similar clusters in locations that are spatially closer in the lower dimension space. This makes the SOM methodology an effective tool for data visualization. However, in a world where mined information from big data have to be available immediately, SOM becomes an unattractive tool because of its time complexity. In this paper, we propose an alternative visualization methodology for large datasets that emulates SOM methodology without the speed constraints inherent to SOM. To demonstrate the efficiency and the potential of the proposed scheme as a fast visualization tool, the methodology is used to cluster and project the 3,823 image samples of handwritten digits of the Optical Recognition of Handwritten Digits dataset. Although the dataset is not, by any means large, it is sufficient to demonstrate the speed-up that can be achieved by using this proposed SOM emulation procedure.

iXray: A Machine Learning-Based Digital Radiograph Pattern Recognition System for Lung Pathology Detection

A radiograph is a visualization aid that physicians use in identifying lung abnormalities. Although digitized x-ray images are available, diagnosis by a medical expert through pattern recognition is done manually. Thus, this paper presents a system that utilizes machine learning for pattern recognition and classification of six lung conditions. Classified into two categories, namely histogram-based (normal, pleural effusion, and pneumothorax) and statistics-based (cardiomegaly, hyperaeration, and possible lung nodules). Using preprocessing and feature extraction techniques, the designed system achieves an accuracy rate of 92.59% for the histogram-based lung conditions using Sequential Minimal Optimization (SMO) and 67.22% for the statistics-based lung conditions using logic operations.

Vsion: Vehicle Occlusion Handling for Traffic Monitoring

Recently, the pervasiveness of street cameras for security and traffic monitoring opens new challenges to the computer vision technology to provide reliable monitoring schemes. These monitoring schemes require the basic processes of detecting and tracking objects, such as vehicles. However, object detection performance often suffers under occlusion. This work proposes a vehicle occlusion handling improvement of an existing traffic video monitoring system, which was later integrated. Two scenarios were considered in occlusion: indistinct and distinct - wherein the occluded vehicles have similar and dissimilar colors, respectively. K-means clustering using the HSV color space was used for distinct occlusion while sliding window algorithm was used for indistinct occlusion. The proposed method also applies deep convolutional neural networks to further improve vehicle recognition and classification. The CNN model obtained a 97.21% training accuracy and a 98.27% testing accuracy. Moreover, it minimizes the effect of occlusion to vehicle detection and classification. It also identifies common vehicle types (bus, truck, van, sedan, SUV, jeepney, and motorcycle) rather than classifying these as small, medium and large vehicles, which were the previous categories. Despite the implementation and results, it is recommended to improve the occlusion handling issue. The disadvantage of the sliding window algorithm is that it requires a lot of memory and is time-consuming. In case of deploying this research for more substantial purposes and intentions, it is ideal to enhance the CNN model by training it with more varied images of vehicles and to implement the system real-time. The results of this work can serve as a contribution for future works that are significant to traffic monitoring and air quality surveillance.

Measuring the contribution of filter bank layer to performance of convolutional neural networks

Object identification is essential in diverse automated applications such as in health, business, and national security. It relies on the ability of the image processing scheme to detect visual features under a wide variety of conditions such as the object rotation, translation and geometric transformation. Machine learning methods, in this case, play an important role in improving the object identification performance by resolving whether the extracted visual patterns are from the possibly distorted target object or not. In recent works, systems that employ a Convolutional Neural Network (CNN) as the primary pattern recognition scheme demonstrate superior performance over other object identification systems based on handpicked shape-based features. Several studies credit this to the invariance of CNN to small distortion and spatial translation which in turn is attributed to its filter bank layer or the convolution layer. However, there has been no study to carefully test this claim. Towards studying the source of CNN’s superior performance, a methodology is designed that tracks the CNN performance when spatial information for visual features (e.g. edges, corners and end points) are gradually removed. Using the MNIST dataset, results show that as the spatial correlation information among pixels is slowly decreased, the performance of the CNN in recognizing handwritten digits also correspondingly decreases. The drop in accuracy continues until the accuracy approximates the performance of the classifier that was obtained without the filter bank. Conducted using a more complex dataset consisting of images of land vehicles, a similar set of experiments show the same drop in classification performance as spatial information among pixels is slowly removed.

Self-organizing Maps as Feature Detectors for Supervised Neural Network Pattern Recognition

Convolutional neural network (CNN)-based works show that learned features, rather than handpicked features, produce more desirable performance in pattern recognition. This learning approach is based on higher organisms visual system which are developed based on the input environment. However, the feature detectors of CNN are trained using an error-correcting teacher as opposed to the natural competition to build node connections. As such, a neural network model using self-organizing map (SOM) as feature detector is proposed in this work. As proof of concept, the handwritten digits dataset is used to test the performance of the proposed architecture. The size of the feature detector as well as the different arrangement of receptive fields are considered to benchmark the performance of the proposed network. The performance for the proposed architecture achieved comparable performance to vanilla MLP, being 96.93 % using 4\(\times \)4 SOM and six receptive field regions.

Performance comparison of ASR classifiers for the development of an English CAPT system for Filipino students

Computer Assisted Pronunciation Training (CAPT) systems aim to provide immediate, individualized feedback to the user on the overall quality of the pronunciation made. In such systems, one must be able to extract features from a waveform and represent words in the vocabulary. This paper presents the performance of Hidden Markov Model (HMM), Support-Vector Machine (SVM) and Multilayer Perceptron (MLP) as automatic speech recognizers for the English digits spoken by Filipino speakers. Speech waveforms are translated into a set of feature vectors using Mel Frequency Cepstrum Coefficients (MFCC). The training set consists of speech samples recorded by native Filipinos who speak English. The HMM-trained model produced a recognition rate of 95.79% compared to 86.33% and 91.66% recognition rates of SVM and MLP, respectively1.