Jianrong Wang

Trust-aware query routing in P2P social networks

Query efficiency is a significant challenge in peer-to-peer (P2P) social networks. In this paper, we build a trust-aware semantic-based query routing strategy to improve the searching efficiency and enhance the security. Based on the vector space model, similar semantic ordinary peers set up an unstructured semantic social community. Super peers in each community constitute a structured P2P network and store the keywords information in a global inverted file. We exploit Bloom filters to complete multikeyword query. Furthermore, malicious behaviors may take place in P2P social networks. We then build a community-based trust mechanism. The intracommunity reputations of ordinary peers in a semantic social community are calculated, and the intercommunity reputations of super peers in the structured P2P network are evaluated. This mechanism combines the local trust and the global trust. The results show that the proposed query scheme is able to increase query efficiency and also decrease query cost with significantly reduced malicious behaviors. Copyright

Audio-visual speech recognition integrating 3D lip information obtained from the Kinect

Abstract Audio-visual speech recognition (AVSR) has shown impressive improvements over audio-only speech recognition in the presence of acoustic noise. However, the problems of region-of-interest detection and feature extraction may influence the recognition performance due to the visual speech information obtained typically from planar video data. In this paper, we deviate from the traditional visual speech information and propose an AVSR system integrating 3D lip information. The Microsoft Kinect multi-sensory device was adopted for data collection. The different feature extraction and selection algorithms were applied to planar images and 3D lip information, so as to fuse the planar images and 3D lip feature into the visual-3D lip joint feature. For automatic speech recognition (ASR), the fusion methods were investigated and the audio-visual speech information was integrated into a state-synchronous two stream Hidden Markov Model. The experimental results demonstrated that our AVSR system integrating 3D lip information improved the recognition performance of traditional ASR and AVSR system in acoustic noise environments.

Community-based trust mechanism in P2P social networks

Malicious behaviors is always the problem to solve in P2P social networks. Building the trust model is effective method. In this paper a community-based trust mechanism is presented in the semantic-based P2P Social Networks. The ordinary peers form a community based on similar semantic and intra-community reputation of peers is calculated. The super peers constitute the structured P2P network and intercommunity reputation is evaluated. This mechanism combines the local trust and the global trust. The experiments show that this trust mechanism can be effective in reducing effects caused by a variety of malicious behavior.

l 0 sparsity for image denoising with local and global priors

We propose a l0 sparsity based approach to remove additive white Gaussian noise from a given image. To achieve this goal, we combine the local prior and global prior together to recover the noise-free values of pixels. The local prior depends on the neighborhood relationships of a search window to help maintain edges and smoothness. The global prior is generated from a hierarchical l0 sparse representation to help eliminate the redundant information and preserve the global consistency. In addition, to make the correlations between pixels more meaningful, we adopt Principle Component Analysis to measure the similarities, which can be both propitious to reduce the computational complexity and improve the accuracies. Experiments on the benchmark image set show that the proposed approach can achieve superior performance to the state-of-the-art approaches both in accuracy and perception in removing the zero-mean additive white Gaussian noise.

Multimodal sensory fusion for soccer robot self-localization based on long short-term memory recurrent neural network

Self-localization is a fundamental requirement for autonomous mobile robots. With the rapid development in sensor technology, the sensor suites of robot provide multimodal information that naturally ensures perception robustness, multimodal sensory fusion are able to provide a better solution for enhance the capability of self-localization. This paper proposes a multimodal sensory fusion method based on Long Short-Term Memory (LSTM) Recurrent Neural Network (RNN) for RoboCup 3D Simulation league. This approach fuses Inertia Navigation System (INS) and vision perceptor information from different sensors at feature level instead of raw data. The experiment results demonstrate that the proposed approach makes an improvement in predictive accuracy and efficiency compared with the standard Extended Kalman Filter (EKF) and the static Particle Filter (PF) methods.

Image Driving Speed Control for a Smart Car System

In the Seventh Freescale Cup National College Smart Car Competition, we designed and implemented a smart car to realize the autonomous driving based on road tracking method. This paper presents three main aspects of the smart car, the mechanical structure; hardware system; software framework and algorithms employed for road tracking and PID speed control. The image-based road tracking system was used to control the steering angle and the speed of the smart car. The images captured by an analog CMOS camera were preprocessed by a monochrome binarization circuit, and then the binarization signals were fed into the MCU as well as the speed feedback signals obtained by a rotary photoelectric encoder, as input variables for controlling the motor, steering gear etc, by the MCU. Experimental results showed that this smart car can accurately recognize the images of various road elements to precisely adjust the steering angle, and steadily run at a maximum controllable speed.

Sound Source Localization Based on Microphone Uniform Linear Array

In order to locate the sound source based on the microphone uniform linear array and reduce the impact of noise and reflection, improved multiple signal classification algorithm and a kind of weighted average filters be used in this paper. According to the microphone array speech processing characteristics, we improved the traditional multiple signal classification algorithm and designed a kind of weighted average filters. Then, we made the computer simulation experiments. The experimental results show that the location of the sound source is the peak with the highest power in the spatial spectrum. Besides, the frequency domain diagram is more smoothly and the power of the noise and reflection is effectively reduced except sound source through the weighted average processing. Therefore, improved multiple signal classification algorithm can achieve sound source localization based on the microphone uniform linear array. And the impact of the noise and reflection is effectively reduced by processing of the weighted average filters.

Eliminating False Matches Using Geometric Context

In this paper, we present a novel approach to encode strong geometric information with the final purpose of eliminating false matches. This approach aims to build an invariant coordinate as the spatial information representation of features set, which can improve accuracy of matches. Firstly, a set of reliable MSER pairs between two images are obtained by a MSER detector and matching. Secondly, all matched MSER pairs are fitted into elliptical regions to achieve rotation invariance, scale invariance, and affine invariance. Finally, for each elliptical region, we build an invariant coordinate system to represent spatial layout of features. We use Full-Geometry-Constraint (FGC) to eliminate false matches without rapidly decreasing on correct matches. The experimental evaluation results show that our method is robust against most geometric and photometric transformations including rotation, scale change, blur, viewpoint change, JPEG compression and illumination changes.

Semantic-based query routing in P2P social networks

In P2P social networks, query efficiency is always a significant challenge. In this paper we build a semantic-based query routing strategy to improve the searching efficiency. Based on Vector Space Model (VSM), similar semantic ordinary peers set up an unstructured semantic community. Super peers in each community constitute a structured P2P network and store the keywords information in a global inverted file. We exploit Bloom Filters to complete multi-keyword query. Simulation results show that the proposed scheme is able to increase query efficiency and also reduce query cost.

Research on the DHT algorithm chord

Chord which used in structured P2P network is a successful routing algorithm based on DHT (Distributed Hash Table). In Chord nodes locate along the Chord ring by being assigned the node identifiers and data store in corresponding nodes with key identifier. Finger table is built in each node and is maintained when nodes join and depart. The problems in Chord are unidirectional clockwise routing along ring and information redundancy in routing table. An improved method is presented which is bidirectional routing table. Routing can perform along clockwise and anticlockwise according to the locations of the current node and destination node. The next hop direction is optimization that the next node is the nearest one apart from the destination node. This strategy limits the search area in half-ring, reduces the average hops and enhances the search efficiency. The redundancy routing information is deleted in order to decrease the added storage space in bidirectional routing table, thus the items share the routing information.