Pakinee Aimmanee

Vessel transform for automatic optic disk detection in retinal images

Precise localisation of an optic disk (OD) in the retinal images is one of the most important problems in the ophthalmic image processing. Although a considerable progress has been made towards a computerised solution of the problem, the numerical algorithms often fail on retinal images characterised by poor quality. Therefore, the authors propose a new method suitable for low-quality images based on exploiting the convergence of the blood vessels to the OD. The novelty of the proposed techniques includes clustering the vessels endowed with a novel correction procedure and the vessel transform (VT) which measures the distance to the main clusters. The algorithm is integrated into the scale-space (SS) analysis to detect the boundary of the OD. The integrated method is called SS algorithm with VT (SSVT). SSVT has been tested on retinal images from two databases with fair and poor images against the fuzzy convergence (FC) method and a modification of the circular transform proposed by Lu. The absolute improvement on sensitivity of SSVT against FC and Lus are up to 12.37% and 8.18%. Bigger improvements of SSVT in terms of positive predictive value are up to 37.46% and 30.84% against FC and Lus, respectively.

An Audio Watermarking Scheme Based on Singular-Spectrum Analysis

This paper proposes a blind audio watermarking scheme based on singular-spectrum analysis (SSA) which relates to several techniques based on singular value decomposition (SVD). SSA is used to decompose a signal into several additive oscillatory components where each component represents a simple oscillatory mode. The proposed scheme embedded a watermark into a host signal by modifying scaling factors of certain components of the signal. Test results show that the proposed scheme satisfies imperceptibility criterion suggested by IHC with the average ODG of 0.18. It is robust against many attacks, such as MP3 and MP4 compression, band-pass filtering, and re-sampling. This paper does not only propose a new watermarking scheme, it also discusses the singular value and reveals its meaning, which has been deployed and played an important role in all SVD-based schemes.

An audio watermarking scheme based on automatic parameterized singular-spectrum analysis using differential evolution

This paper proposes an audio watermarking scheme based on singular-spectrum analysis (SSA) and differential evolution. In our framework, a watermark is embedded into an audio signal by modifying the amplitude of some oscillatory components which are decomposed by SSA, and a parameter set for the modification is determined by differential evolution. Test results showed that, although there is a trade-off between inaudibility and robustness, the sound quality of watermarked signal could be improved considerably while the bit error rate could be satisfied. Our proposed scheme is inaudible and robust. Furthermore, based on analyzing the second derivative of singular spectrum, it was found that our proposed scheme can be completely blind.

Automatic optic disk detection in retinal images using hybrid vessel phase portrait analysis

We propose vessel vector-based phase portrait analysis (VVPPA) and a hybrid between VVPPA and a clustering method proposed earlier for automatic optic disk (OD) detection called the vessel transform (VT). The algorithms are based primarily on the location and direction of retinal blood vessels and work equally well on fine and poor quality images. To localize the OD, the direction vectors derived from the vessel network are constructed, and points of convergence of the resulting vector field are examined by phase portrait analysis. The hybrid method (HM) uses a set of rules acquired from the decision model to alternate the use of VVPPA and VT. To identify the OD contour, the scale space (SS) approach is integrated with VVPPA, HM, and the circular approximation (SSVVPPAC and SSHMC). We test the proposed combination against state-of-the-art OD detection methods. The results show that the proposed algorithms outperform the benchmark methods, especially on poor quality images. Specifically, the HM gets the highest accuracy of 98% for localization of the OD regardless of the image quality. Testing the segmentation routines SSVVPPAC and SSHMC against the conventional methods shows that SSHMC performs better than the existing methods, achieving the highest PPV of 71.81% and the highest sensitivity of 70.67% for poor quality images. Furthermore, the HM can supplement practically any segmentation model as long as it offers multiple OD candidates. In order to prove this claim, we test the efficiency of the HM in detecting retinal abnormalities in a real clinical setting. The images have been obtained by portable lens connected to a smart phone. In detecting the abnormalities related to diabetic retinopathy (DR), the algorithm provided 94.67 and 98.13% for true negatives and true positives, respectively.

An Empirical Study on Multi-dimensional Sentiment Analysis from User Service Reviews

Online reviews on a service are important sources for service providers to improve their service delivery and service consumers to obtain information for decision making before their service acquisition. However, in the real situation, there are several points of view (dimensions) in service assessment using online reviews. This paper shows an empirical study to apply classification-based sentiment analysis on online reviews with multiple dimensions using natural language processing techniques. The aim of this study is to find the most influential part-of-speech on the sentimental analysis and the performance of the multi-dimensional classification methods. By the experiments on reviews of restaurants with five dimensions, i.e., taste, environment, service, price, and cleanness, we find out that adjective (JJ) has the most influential part-of-speech on the sentimental analysis and BRplus is the most efficient one with the classification accuracy of 85.89%.

A comparability approach to item reduction in Computerized Adaptive Testing

Computerized Adaptive Testing (CAT) is a widely used computer based testing that can classify examinees according to their abilities. The question sent to each examinee depends on his/her answers to previous questions. One drawback is that CAT is not suitable for a small examination pool since it may not guarantee to cover all topics. The purpose of this study is to introduce a new adaptive testing method called Comparable Item Reduction CAT which has many attractive features such as the ability to operate on an item pool whose size is fixed and to assure that all topics are distributed equally and efficiently. The result is the test that is shorter in length and accurate in each examineepsilas ability estimation.

Fast Hemorrhage Detection in Brain CT Scan Slices Using Projection Profile Based Decision Tree

Detection of a hemorrhage in CT scan slices is one of the crucial steps for a neurosurgeon to diagnose any abnormality and severity in the brain of a patient. It is usually time consuming as there are as many as 256 produced slices from a CT scan machine for each patient. In this paper, we introduce an automatic hemorrhage detection in brain CT slices using features-based approach. We employ decision tree based on 8 features to classify slices to two classes- with and without the sign of hemorrhage. The proposed method is tested on 1,451 CT scan slices and achieves a classification accuracy for up to 99% and it takes 0.12 second to detect slices.

OD Localization Using Rotational 2D Vessel Projection with Decision Tree Classification

Automatic Optic Disc (OD) localization is an important problem in ophthalmic image processing. Knowing its location helps doctors with the early detection of preventable eye diseases. Inspired by a fast and accurate OD localization algorithm utilizing the vessel projection technique that is usually inefficient when the OD in the image is unusually pale, we employed the decision tree with 5 features to improve the accuracy of the existing algorithm. Also to overcome the problem of poor accuracy when the image is tilted, we repeatedly run this improved algorithm on a series of images tilted at different degree from the original image to obtain the voted location of the OD. The proposed method has been tested on different starting angles between 0 to 180 degrees from Structured Analysis of the Retina (STARE) and retinopathy of prematurity (ROP) datasets. We achieve an average accuracy of up to 86% with an average computation time per image of only 13 seconds per image. Our approach outperforms two other based approaches, Mahfouz and Rotational 2D Vessel Projection (RVP), by up to 34% and 12%, respectively.

Tampering Detection in Speech Signals by Semi-Fragile Watermarking Based on Singular-Spectrum Analysis

To solve the problem of unauthorized modification in speech signals, this paper proposes a novel speech-tampering-detection scheme by using the semi-fragile watermarking based on the singular-spectrum analysis (SSA). The SSA is used to analyze the speech signals of which the singular spectra are extracted. The watermark (e.g., signature in-formation) is embedded into those signals by modifying some parts of the singular spectra according to the watermark bit. By comparing the extracted watermark with the original one, the tampered segments of the speech signals are identified and located. The evaluation results show that the proposed scheme is fragile to several malicious attacks but robust against other signal-processing operations. It also satisfies the inaudibility criteria. The proposed scheme not only can locate the tampered locations, but it also can make a prediction about the tampering types and the tampering strength.

SSA-based audio-information-hiding scheme with psychoacoustic model

This paper proposes an inaudible and robust audio-information-hiding scheme based on the singular-spectrum analysis (SSA) and a psychoacoustic model. SSA is used to decompose the host signals into several additive oscillatory components. The hidden information is embedded into the host signals by modifying amplitudes of some oscillatory components. To satisfy the inaudibility, we propose a novel method which is based on the psychoacoustic model to choose the oscillatory components that distort the host signal only slightly after the modification. Accordingly, to associate the output from the psychoacoustic model with the SSA-based scheme, the relationship between the frequency components and singular-value indices is established. The test results show that the proposed scheme achieves both inaudibility and robustness. In addition, compared with the previously proposed scheme based on automatic parameterized SSA using the differential evolution, this proposed scheme has more advantage in terms of reducing the computational time.