Volume 20, Issue 2

Amblyopia, or lazy eye, is a neurological vision disorder that studies have shown to affect two to five percent of the population. Current methods of treatment produce the best visual outcome, if the condition is identified early in the patient's life. Several early screening procedures are aimed at finding the condition while the patient is a child, including an automated vision screening system. This paper aims to use artificial intelligence techniques to automatically identify children who are at risk for developing the amblyopic condition and should therefore be referred to a specialist, i.e., pediatric ophthalmologist. Three techniques, namely, decision tree learning, random forest, and artificial neural network, are studied in this paper in terms of their effectiveness, using metrics of sensitivity, specificity, and accuracy. The features used by the techniques are extracted from images of patient eyes and are based on the color information. The efficacy of pixel color data is investigated with respect to the measurement of the rate of change of the color in the iris and pupil, i.e., color slope features. A 10-fold stratified cross validation procedure is used to compare the effectiveness of the three AI techniques in this medical application domain.

Speed is one of the major factors by which the traffic safety is affected. If the speed limit traffic signs on the road are recognised and displayed to a driver, this will be a motivation to keep the vehicle's speed within the permitted range. The purpose of this paper is to investigate Eigen-based traffic sign recognition which can aid in the development of Intelligent Speed Adaptation. This system is based on invoking the PCA technique to detect the unknown speed limit traffic sign and computes its best effective Eigen vectors. The traffic sign is then recognized and classified by using the shortest Euclidean distance to the different speed limit traffic sign classes. The system was trained using 24 037 images which were collected in different light conditions. To check the robustness of this system, it was tested against 1429 images and it was found that the accuracy of recognition was 97.5% which indicates clearly the high robustness targeted by this system.

Inventory prediction and management is a key issue in a retail store. There are a number of inventory prediction techniques. However, it is difficult to identify a time series prediction model for inventory forecasting that provides uniformly good results for all the products in a store. This paper uses data from a small retail store to demonstrate the variability of results for different modeling techniques and different products. We demonstrate inadequacy of a generic inventory model. Stability and seasonality analysis of the time series is used to identify groups of products (local groups) exhibiting similar sales patterns. Different clustering techniques are applied to determine reasonable local groups. With the help of Mean absolute percentage error (MAPE), the effectiveness of dataset partitioning for better inventory management is demonstrated. Appropriate inventory management strategies are proposed based on the stability and seasonality analysis.

Center of pressure (COP) measurements are often used to identify balance problems. A new method for COP signal analysis using mean frequency is proposed in this paper. The computation of mean frequency is based on the Fourier–Bessel (FB) expansion that is suitable for use in non-stationary COP signals. In addition, FB expansion provides better frequency resolution and low frequency detectability. Seventeen subjects were tested under eyes open and eyes closed conditions, with different vibration frequencies applied for the eyes closed condition in order to further perturb sensory information. More accurate estimates of mean frequency were obtained for 200-point segments, with significant increase in mean frequency observed over time as subjects recovered from the initial perturbation of stepping on to the force plate. Subjects were significantly more stable in the mediolateral direction compared to anteroposterior one. Mean frequency as calculated by FB expansion was able to distinguish between eyes open and eyes closed conditions ( p < 0.05), but no effect of vibration could be detected.

This paper presents a new boosting algorithm called NormalBoost which is capable of classifying a multi-dimensional binary class dataset. It adaptively combines several weak classifiers to form a strong classifier. Unlike many boosting algorithms which have high computation and memory complexities, NormalBoost is capable of classification with low complexity. Since NormalBoost assumes the dataset to be continuous, it is also noise resistant because it only deals with the means and standard deviations of each dimension. Experiments conducted to evaluate its performance shows that NormalBoost performs almost the same as AdaBoost in the classification rate. However, NormalBoost performs 189 times faster than AdaBoost and employs a very little amount of memory when a dataset of 2 million samples with 50 dimensions is invoked.