Recent Patents on Computer Science (v.8, #1)

Editorial: by Hamid Mcheick (1-1).

A Review on Pedestrian Crossing Detection and Behavior Analysis: In Intelligent Transportation System by Ronghui Zhang, Fuliang Li, Jiali Zhou, Feng You, Wenxiao Zeng, Tonghai Jiang (2-12).
Pedestrian detection and behavior analysis is an active research area with challenge in ITS. This text mainly emphasizes the research methods and the pedestrian detection fruits. According to the frame of pedestrian detection system, the pedestrian detection method based on machine vision, the pedestrian detection method under multi-sensor information fusion, the principle for pedestrian tracking and pedestrian behavior analysis are summarized in turn. Classifying the research methods above, it divides machine vision detection method into static background and dynamic background. Considering the pedestrian detection principle with multi-sensor information fusion, it was divided into the one based on classic mathematics and the other one based on AI. Pedestrian tracking methods are divided into the matching step and the probability prediction step. For the Pros and cons, they are also analyzed and concluded. Finally, some currently problems are proposed. Besides, the pedestrian detection method patents combining the future tendency of communication, automation and computer technology are expected by authors.

Financial fraud investigation, detection and risk evaluation are essential to either bank (as well as other financial institutions) or business entities. This topic continuously receives significant amount of attention from different stakeholders, including the financial institutions and regulators, investors, researchers, community and merchants. Maximum level of decidability and elimination of uncertainty in fraud likelihood estimation are essential to ensure financial stability, evaluate risk and minimize possible financial loss. Artificial intelligence techniques prove to be efficient solution and worth of attention to any inventor. This paper surveys recent patents related to financial fraud detection, which apply one or more artificial intelligence, machine learning or statistics based techniques, identifies main features of modern intelligent solutions for fraud risk detection and summarizes main trends.

More and more researchers have been making effort on developing intelligent vehicle systems. In contrast with single vehicle operation, multivehicle cooperation can bring considerable benefits to intelligent vehicles that operate in common areas. In the spirit of multivehicle cooperation, this paper describes a method that realizes a function (cooperative augmented reality, shortly as CAR) via which a vivid visualization of occluded situation can be generated for driver assistance purpose. This paper details the knowledge on 3D perspective transform of the vehicle visual data and describes an improvement for the original CAR solution presented in a related conference publication along with recent patents related to it. Experimental results are given to demonstrate the effect of the CAR solution.

A Front Vehicle Detection Algorithm for Intelligent Vehicle Based on Improved Gabor Filter and SVM by Zhonghua Zhang, Xuecai Yu, Feng You, George Siedel, Wenqiang He, Lifang Yang (32-40).
Front Vehicle Detection is the key and difficult point of the key technology research for the intelligent vehicle. In this paper the digital image is firstly binarized through the image enhancement, threshold segmentation and noise eliminating along with recent patents described. Then hypothesis generation is done according to the structure, shape, aspect ratio of the vehicle and shadow at the bottom of the vehicle. On this basis, features extraction is performed with a Gabor features extraction method based on the improved features weightings for the selected vehicle samples and background samples. The extracted features vector is regarded as the input of the support vector machine (SVM) for training. Finally, the trained SVM classifier is used to conduct the vehicle classification and recognition. Thus the vehicle detection is completed. The experimental results show that the approach can improve the recognition rate and the robustness of preceding vehicle detection for the intelligent vehicle.

Traffic Incident Automatic Detection Algorithms by Using Loop Detector in Urban Roads by Weiwei Guo, Zhijian Wang, Wuhong Wang, Heiner Bubb (41-48).
This paper proposed an improved California Algorithm and recent patents for detecting traffic incidents in urban roads. Traffic incidents in urban roads contribute 50% to 60% of the total congestion delays. Incident detection involves the collection and the analysis of traffic data. The common loop detectors were selected as the collected tool in this research. The existing incident detection algorithms are more abundant, and mostly apply in freeways where the traffic flows are steady. The traffic incident detection algorithms in urban roads should be developed. Based on the existing detection algorithms, an improved California Algorithm was presented, which contained the adoptive dynamic threshold with the characters of traffic incidents and traffic flows in the road network. Further detection results of the improved California Algorithm were judged by the OD matching regularity of the volumes at the signal intersections to make the detection results fit with the actual flow feature. Finally, the data collected by loop detectors was used to validate the proposed algorithm. The results showed that the improved algorithm could predict the traffic incidents effectively, but there are still some false positives.

Real-Time 3D Reconstruction of Wood Logs in a Production Line by Lu Yang, Luis Diago, Ichiro Hagiwara (49-57).
The wood log is an important material in industry. These wood logs need to be cut for use. In wood cut production line, the wood information is necessary. Currently, laser is widely used to obtain the wood logs information for cutting. However, laser is usually very expensive. In this paper, we try to use camera to replace laser for reconstructing 3D information of wood logs and recent patents associated with it. For this purpose, we propose a vision system, which can provide the necessary information to the cutter machine. The obtained 3D information can be used to create aesthetically pleasant wooden patterns for accurately cutting the wood. Matching is the key technique for 3D reconstruction in our vision system. In this paper, we use area-based matching approach which has the result of dense disparity map and saves running time. Our vision system uses two cameras to collect original image of the wood logs. These images are preprocessed for matching with SAD (Sum of Absolute Differences) similarity operator. Several constraints are also proposed to obtain disparity map for 3D reconstruction. Finally, the experiment results indicate that the proposed approach can offer accurate 3D measurement to automatically cut the wood.

Modeling Uyghur Speech Phenomena with Morphological Rules by Huajian Xue, Ronghui Zhang, Haiwei Wang, Wenxiao Zeng, Chongke Bi (58-66).
Due to the morphology of Uyghur, it poses a challenge to state-of-the-art speech recognition systems. This paper describes our work on Uyghur speech recognition technology and a Uyghur voice search application with recent patents on them. Firstly, we introduce the morphology of Uyghur and Uyghur speech phenomena. Secondly, we investigate the use of morphemes in Uyghur auto speech recognition. When speech phenomena happen, variant surface forms of the morpheme are produced. Then, we describe a new approach utilizing the morphological rules to model speech phenomena. In this new approach, variant surface forms are replaced by their corresponding original stems. This creates a better pronunciation lexicon and a more robust language model in Uyghur speech recognition experiments. Finally, we apply the new morphemebased speech recognition approach to a Uyghur voice search application. Experimental results show that this new approach gives the best results in both speech recognition experiments and voice search experiments.

The memory types used in conventional computers are analyzed and a new patented ordered data access method and apparatus are proposed. Recent patented data access method and apparatus are described and challenging problems of their use in high-performance computers are discussed. Distinctive features of the new method compared to known data access methods are discussed. Input data, their indices and output data of the ordered access memory are described. The ordered access memory organization and interface as well as their advantages compared to the random, associative, and sequential access memories are discussed. Four variants of the ordered access memory are described, and the examples of the ordered access memory usage in computer systems are demonstrated. The results of the ordered access memory implementation in FPGA are presented.