News ticker recognition is a vital area of research due to its applications such as information analysis, opinion mining and language translation for media regulatory authorities. Without automated systems, manual anatomizing is difficult. In this paper, we focus on the automatic Arabic and Urdu news ticker recognition system. It mainly consists of ticker segmentation and text recognition to generate textual data for various online services. Our work investigates character-wise explicit segmentation and syntactical models with Kufi and Nastaleeq fonts. Various network models anticipate learning of deep representations by homogenizing the classes regardless of inter-symbol correlations and linguistic taxonomy. The proposed learning model incorporates fairness by maximizing the balance among sensitive features of characters in a unified manner. Furthermore, we demonstrate the efficiency of the proposed model by carrying out experiments using customized news tickers datasets with accurate character-level and component-level labeling. Moreover, our method is evaluated on a challenging Urdu Printed Text Images (UPTI) dataset that only provides ligature based annotations. The proposed method attains 98.36%, outperforms the current state of the art method. Ablation investigations show that our technique enhances the performance of character classes with low symbol frequencies.

No abstract

Abstract-Information storage and retrieval is the fundamental requirement for many real-time applications. These systems demand that data should be sorted all the time, real-time insertion, deletion and searching should be supported and system must support dynamic entries. These systems require search operations to be performed from massive databases implemented by various data structures. The common data structures used by these systems are stack, queue or linked list all having their own limitations. The biggest advantage of using stack is that binary search can be performed on it easily while on the other hand insertion and deletion of nodes involves more processing overhead. In linked list, insertion and deletion of nodes is easier but searching operation involves more processing overhead as binary search cannot be performed efficiently on it. In this paper, a hybrid solution is presented for such systems, which provides efficient insertion, deletion and searching operations. Results show the effectiveness of the proposed approach as it outperforms the existing techniques used by these systems.

Detection and recognition of the moving objects in dynamic environment is difficult task. This paper presents a modified framework for the detection and recognition of moving people in videos. Detection part of the proposed method consists of average background model with supportive secondary model and an adaptive threshold selection model based on Gaussian distribution. The background model used for background modelling and adaptive threshold method is used to simultaneously update the system according to environment. Then feature extraction is performed by an established human model. This human model consists of five parts with robust features to facilitate recognition process. For recognition purpose, back propagation neural network has been used as a classifier. Experimental results show the effectiveness of proposed system.

No abstract