Search

This paper presented Independent Component Analysis (ICA) Hard-IP integration in System on Programmable Chip (SOPC) platform. The ICA component can discover the main component for original signal in multiple fetching signal sources, and it has been used in biomedical signal processing such as electroencephalogram (EEG) analysis. The proposed system consists of a programmable CPU, ICA processing units, system bus, communication, and display interface. The experimental results showed that the proposed design implemented on Altera DE2 FPGA development board, can achieve real-time signal separation and display at 100 MHz. The whole design consists of 29,640 logic elements.

It is a significant task to extract market data from different Web pages for prediction and analysis. A prototype decision support system of an agricultural product market is designed and developed in this paper. It can extract online price information of a certain agricultural product from websites of agricultural wholesalse, predict the product price in the future months, and provide further decision support on such issues as which cities the product should be sent to for sale and which cities should be in the transport route. To achieve these goals, an algorithm named MDT-E (Market Data Table Eextraction) is proposed to extract the maximum data table in a Web page. Based on the common practice that “ the price data are usually displayed in the largest table on a Web page with the structure of “” and “” tags”, our market data extraction algorithm detects the largest table on a Web page at first, then transforms the table into a DOM tree, and further obtains the node values of the “” tags. This algorithm can automatically detect market data without an assigned data extraction region. The designed system uses a quadratic forcasting model of linear time series to predict the price, and compares the prediction results by using different time series and different sample data to find the best forecasting model to forecast the price in cites. In addition, it provides the decision support to determine the transport route based on the transport costs and product prices.

An effective and economic process for removing arsenic in waste water which is acuminating in the process of etching, cutting and washing in semiconductor industry has been developed in this paper. The proposed technique of arsenic removal is in three steps: first pretreatment step is to oxidize arsenite to arsenate by potassium permanganate, second key step is precipitation based on arsenic compound solubility with ferric sulfate and slaked lime under pH adjustment, and the last complementary step is followed by the adsorption of the bentonite with enhanced by activated carbon and organic adsorbent. Experimental results show that under the best condition the removal efficiency of arsenic in the waste water is better than 99.99%, or the concentration of arsenic is from its original 100 mg/l reduced to less than 10 mug/l accordingly.

Polychlorinated dibenzo-p-dioxins and dibenzofurans (PCDD/Fs) in surface sediment and mussel samples from Jiaozhou Bay were analyzed using HRGC-HRMS (High Resolution Gas Chromatograph-High Resolution Mass Spectrometer) technique. The total PCDD/Fs content was in the range of 100 ~ 910 pg/g dw, and that of total 2,3,7,8-substituted PCDD/Fs contents was 4.1 ~ 232.4 pg/g dw. The distribution of the total 2,3,7,8-substituted PCDD/Fs in the sediments formed an east-west pattern under the action of cycle-flow. The highest level occurred near the river mouth was an order of magnitude higher than at other stations which was believed to be pollutant source. Except for the area near the river mouth, the relative contributions of individual 2,3,7,8-substituted congeners at each station were similar, i.e., the lower chlorinated congeners dominated over the higher chlorinated PCDD/Fs., and the source was believed to be the atmospheric deposition from long-range transport. The source of PCDD/Fs was relatively constant as evidenced by the PCDDs/PCDFs ratios and the OCDD %Sigma. The total 2,3,7,8-PCDD/Fs contents in the mussel sample ranged from 5.1 to 2.9 pg/g dw (52.1 ~ 35.4 pg/g lipid). The main sources of TEQ (Toxicity Equivalency Quantity) were 2,3,4,7,8-PeCDF (35%), 2,3,7,8-TCDF (30%), and 2,3,7,8-TCDD (30%).

Concurrent manipulation control scheme is an important problem for cooperative design. Consistency, responsiveness and concurrency are three important while opposite requirements for concurrent manipulation control. Furthermore, the essence of concurrent manipulation control is to solve manipulation confliction and provide confliction-free manipulation. Hence, the focus of concurrent manipulation control scheme is how to solve manipulation confliction with minimum response time and maximum concurrency under consistency. This paper proposed several design considerations for concurrent manipulation control scheme. Based on these considerations, this paper presented a new entity-centric concurrent manipulation scheme based on a series of formalized concepts. This new scheme implements relax WYSIWIS, and adopts confliction elimination rather than confliction avoidance. The new scheme is divided into three steps: local manipulation, server-side manipulation validation and replica synchronization. Local manipulation improves responsiveness; server-side manipulation validation maximizes concurrency; and replica synchronization guarantees consistency. Moreover, an overall implementation for the new entity-centric concurrent manipulation control scheme is depicted in details.

Combination of kinds of artificial intelligence theories in application area of pattern recognition has become one of the most important ways of research of intelligent information processing. Neural network shows us its strong ability to solve complex problems for patter recognition. But neural network can't tell the redundant information from huge amount of data, which will easily lead to some problems such as too complex network structure, long training time, low converging speed and much computation. Focusing on these problems, this paper proposes a method to combine rough set theory with neural network theory and uses it in pattern recognition. Experiments show the potential of this method.