Segmentation is the most difficult problem in a handwritten character recognition system and often contributes major errors to its performance. To reach a balance of speed and accuracy, a filter distinguishing a conne...
详细信息
ISBN:
(纸本)0769503187
Segmentation is the most difficult problem in a handwritten character recognition system and often contributes major errors to its performance. To reach a balance of speed and accuracy, a filter distinguishing a connected image from an isolated image is required for multi-stage segmentation. The Fourier spectrum is promising in this problem. Since it is influenced by the stroke width, we propose a Fourier spectrum standardization method. Based on the standardized Fourier spectrum, a set of features and a fine-tuned criterion are presented to classify connected/isolated images. A theoretical analysis proves their rationality. Experimental results demonstrate that this criterion is better than other methods.
Cache coherence protocols are important for operating a shared-memory multiprocessor system with efficiency and correctness. Cache coherence protocols have become increasingly complex because physical memory is logica...
详细信息
Cache coherence protocols are important for operating a shared-memory multiprocessor system with efficiency and correctness. Cache coherence protocols have become increasingly complex because physical memory is logically distributed, so that it is difficult for programmers to understand the view of logical shared-memory systems. Since random testing and simulations are not enough to validate the correctness of these protocols, it is necessary to develop efficient and reliable verification methods. Through the use of the Symbolic State Model (SSM) of Fong Pong (1995), we verified a directory-based protocol called the RACE (Remote-Access Cache coherence Enforcement) protocol. The protocol is verified for any system size, without state-space explosion.
Typically, two's complement is chosen to represent numbers since arithmetic operations of addition and subtraction are easy to perform. However, due to sign extension, an arithmetic operation for small dynamic dat...
详细信息
Typically, two's complement is chosen to represent numbers since arithmetic operations of addition and subtraction are easy to perform. However, due to sign extension, an arithmetic operation for small dynamic data ranges may require switching power in the entire word length. Herein, we develop a two's complement adder with the dynamic-range determination and sign-extension units to reduce power consumption. According to the actual data range by minimizing sign-extension bits, only partial functional blocks of an adder are active to generate a final result of which sign bit is then extended to match the original word length. Experimental results demonstrate that our 32-bit carry-lookahead adder has 22.9% power reduction than the conventional one while dynamic ranges of input data are the Gaussian distribution with a mean of 16 bits and a standard deviation of 8 bits.
We describe direct extraction techniques for the most important parameters of a new physics-based polysilicon (poly-Si) TFT model, suitable for circuit simulation. The physics-based model covers all operating regimes ...
详细信息
We describe direct extraction techniques for the most important parameters of a new physics-based polysilicon (poly-Si) TFT model, suitable for circuit simulation. The physics-based model covers all operating regimes using continuous functions, includes short-channel effects and has been validated for devices of channel lengths down to 2 /spl mu/m. In spite of a small parameter set, the model includes the necessary dependencies on channel length.
The relation between the number of nodes in a ROBDD and the number of implicants in the disjoint cover of the function represented by that ROBDD is studied. We identify a class of functions for which there are disjoin...
详细信息
ISBN:
(纸本)0780355830
The relation between the number of nodes in a ROBDD and the number of implicants in the disjoint cover of the function represented by that ROBDD is studied. We identify a class of functions for which there are disjoint covers such that a cover of a larger size can be represented by a ROBDD with a smaller number of nodes. This shows that the size of a ROBDD is not a monotonically increasing function of the size of the disjoint cover.
Classification, which involves finding rules that partition a given dataset into disjoint groups, is one class of data mining problems. Approaches proposed so far for mining classification rules from databases are mai...
详细信息
Similarity Indexing is very important for content-based retrieval on large multimedia databases, and the "tightness" of data set envelope is a factor that influences the performance of index. For equidistant...
详细信息
In the recognition of Chinese handwritten characters,it is a pattern matching process with large number of standard *** is the bottleneck of the recognition *** this paper,a multi-layered pipeline architecture is devi...
详细信息
In the recognition of Chinese handwritten characters,it is a pattern matching process with large number of standard *** is the bottleneck of the recognition *** this paper,a multi-layered pipeline architecture is devised to solve this bottleneck. The technology of multi-bank storage,parallel computing,*** also implemented to optimize the ***,a high recognition speed is *** experimental system is implemented on a Xilinx XC4013E FPGA *** will be migrated to a custom VLSI chip in the future.
The explosive popularity of wireless data networking has been continued over the years. Furthermore, the strong requirement to access the Internet while on the move via mobile devices is growing more and more, since l...
详细信息
ISBN:
(纸本)0818672250
The explosive popularity of wireless data networking has been continued over the years. Furthermore, the strong requirement to access the Internet while on the move via mobile devices is growing more and more, since lots of useful resources are residing in the wired net, Internet. DataTAC manufactured by Motorola Inc. provides an infrastructure bridging Internet and mobile devices. On top of it, we built a middleware to make it possible for existing network software to surf on the Internet with a mobile device. The middleware provides such high application transparency that it does not need to be recompiled to adapt the middleware. We call that system Brown (Bridge over wireless net to Internet). In this paper, we introduce Brown, and then addresses the core technology, Vision (Virtual socket interfaces over wireless net), a communication layer to support multiple TCP/IP socket-based communication channels.
Software reengineering for object-oriented rearchitecturing offers an exciting opportunity in migrating old legacy systems to evolvable systems in a disciplined manner. In the twofold-strategy software reengineering f...
详细信息
ISBN:
(纸本)0818685859
Software reengineering for object-oriented rearchitecturing offers an exciting opportunity in migrating old legacy systems to evolvable systems in a disciplined manner. In the twofold-strategy software reengineering for object-oriented rearchitecturing, one of the problems to be solved is to derive a proper object-oriented model from the output of reverse engineering and the output of forward engineering. In general, the outputs of forward and reverse engineering can be inconsistent in their abstract levels, the amount of design information, naming conventions, and structures. In this paper we present an Object-oriented model Refinement Technique (ORT) to build a final object model in the twofold-strategy software reengineering. We first organize the information gained from reverse engineering into specification information tree, and then compare the entities in the specification information tree with the information from forward engineering using tree-structured data dictionary to produce the final model. We demonstrate the usability of ORT by an example.
暂无评论