Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2019-01-03 Cooperative journals: 《计算机应用研究》
Abstract： For the allocation of spatial tasks in mobile crowd sensing, the spatial distance between users and tasks directly affects the cost required to complete the tasks, while the existing research about this is considered inadequate. In order to minimize cost, the paper proposed a spatial task allocation mechanism. Firstly, based on genetic algorithm and greedy algorithm, it designed an efficient task allocation method to minimize the cost required to complete the tasks. Secondly, considering the randomness of user sensing quality, it designed a mechanism of user sensing quality updating based on the historical quality and the quality of the current task. To verify the effectiveness of the proposed mechanism, it conducted simulations compared with the two benchmarks. The results show that for spatial task allocation the proposed mechanism has better results in terms of total cost and spatial distance that the user needs to move to perform the task. Therefore, the allocation mechanism proposed has better foreground.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2019-01-03 Cooperative journals: 《计算机应用研究》
Abstract： At present, the existing convolutional neural network has complicated structure and bases on huge dataset. So it has difficulties in meeting the requirement of computing performance and limitation of energy consumption requested by some practical applications or computing platforms. We studied the binary algorithm based on ARM+ FPGA platform and designed a binary neural network aiming at these applications or platforms. This work reduces the demand for data storage units and simplifies the computational complexity. When implemented in the ARM+ FPGA platform, the convolution multiply-accumulate operation is converted into XNOR logic and popcount operation, which improves the overall operation efficiency and declines the consumption of energy and resources. At the same time, based on the characteristics of data storage in binary neural network, a new row processing algorithm is proposed to improve the throughput of the network. In a word, This implementation is superior to the existing FPGA neural network acceleration methods in terms of GOPS, energy and resource efficiency.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-11-29 Cooperative journals: 《计算机应用研究》
Abstract： In many applications of mobile crowd sensing, participants should collect continuous data over a period of time in this scenario on which existing research lacks consideration. For this scenario, the paper proposes a participant selection mechanism which is time window dependent. This mechanism includes: a participant selection method which is time window dependent based on dynamic programming algorithm. The target of the method is to maximize data benefits while cover time period of the task; A updating mechanism of participant’s reputation given the willingness and data quality of the participant. Finally, simulation results show that compared with two common selection mechanisms, the participant selection mechanism proposed has better performance in terms of data reliability, data benefits and cost and has better prospects.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-09-12 Cooperative journals: 《计算机应用研究》
Abstract： In the field of safe driving, the driver's physical and mental state is critical to traffic safety. It is an effective means to detect abnormal driving such as fatigue by detecting the driver's face video input network through a webcam. The previous method mainly analyzes the facial expressions such as the driver's mouth shape to analyze whether or not to yawn, thereby judging whether or not fatigue driving, and therefore many similar states such as speaking are also mistakenly detected as fatigue. Aiming at the above problems, a detection framework based on sequential facial motion information is proposed to detect the driver's state, thus improving the detection accuracy and reducing the false detection rate. The framework mainly consists of two key parts: (1) by detecting the contour of the face in the video, extracting various features of the face to form a facial action unit; and (2) forming a sequential facial action by training the corresponding LSTM network. The unit performs a plurality of action unit fusions according to its correlation to detect the state of the final driver. The test results on the public YAW-DD data set show that the accuracy rate is increased to 93.1% compared with the existing method, and the false detection rate of the fatigue state is greatly reduced.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-20 Cooperative journals: 《计算机应用研究》
Abstract： This paper aims to research the tetrahedron fission topology that is a new type of three dimensional Network-on-Chip topology, gives generation process of topology and the design of encoding and routing. Through a three dimensional extension of the gpNoCsim which is a simulator on chip network, the simulation experiment of the tetrahedron fission topology is conducted. The simulation results show, tetrahedron fission topology can achieve lower average latency and fewer average hops than Mesh under uniform traffic pattern, When the injection rate is 0.02, comparing with Mesh, the average latency degrades by 16.8 % and the average hops decrease 5.5% ; Under localized traffic pattern, the average delay of tetrahedral fission topology and the average hops are significantly improved when the injection rate is greater than 0.008, Comparing with Mesh , tetrahedron fission leads to 18.7% decrease of the average latency and 9.6% decrease of the average hops when the injection rate is 0.014. It is shown that the tetrahedron fission topology can be used in the design of three dimensional Network-on-Chip topology.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-12 Cooperative journals: 《计算机应用研究》
Abstract： To improve the accuracy of the pedestrian detection method, a rectangular input of convolution neural network enhance the new pedestrian detection method based on the depth residual network and YOLO object detection method. The rectangular input helped the model gain the pedestrian characteristics expression by analyzing the expression and distribution characteristics of pedestrians in the images. The depth residual network with pre-activation for YOLO object detection improved the feature extraction ability through more layers of convolution neural networks. Hybrid dataset training and cluster anchor boxes could also improve the pedestrian detection performance. The test results of INRIA dataset have proved that the method has better detection performance than the popular pedestrian detection methods, the index of False Positive per Image can reduce to 13.86%, improving ranging from 1.51% to 58.62% in varying degrees.
Hits
Hits
Downloads
Comment