Advanced Materials Research Vols. 605-607

Abstract: In order to overcome the disadvantage of neural networks that their structure and parameters were decided stochastically or by one’s experience, an improved BP neural network training algorithm based on genetic algorithm was proposed.In this paper,genetic algorithms and simulated annealing algorithm that optimizes neural network is proposed which is used to scale the fitness function and select the proper operation according to the expected value in the course of optimization,and the weights and thresholds of the neural network is optimized. This method is applied to the stock prediction system.The experimental results show that the proposed approach have high accuracy,strong stability and improved confidence.

Abstract: The study was conducted to identify three types of non-touching grain kernels using a colour machine vision system. Images of individual cereal grain kernels were acquired using an camera. Shape feature was extracted from binary and edge images of cereal grain kernels obtained by iamge processing for classification. A total of 13 shape feature parameters, including region area, perimeter, length, width, the maximum radius, the smallest radius etc, were extracted from each kernel to use as input to the Bayesian classifier. Experimental results showed that the Bayesian classifier gave better classification with a calssificaiton accuracy of 99.67% for indica type rice, followed by 98.67% and 78.33% for japonica rice and glutinous rice using training set, respectively. The classification system was developed with Bayesian classifier that achieved an overall recognition rate of 92.22% with training data set and furthermore, a classification accuracy of 90% for the testing data set.

Abstract: This study was conducted to discriminate the weed from the corn in a field combined neural network classifier with image processing technology. The corn and weed images were scanned using a colour imaging system. In the first step, an approximate location of the object of interest was determined by minimum enclosing rectangle, in which image processing was done to obtain the binary image. In the second step, the seven invariant moments were extracted from binary images and used as input to the back propagation neural network (BPNN) classifier. The training set was used to construct shape model representing the objects. The detection accuracy was enhanced by adjusting the number of neurons in the network. Experimental results showed that the BPNN classifier achieved overall detection accuracy of 94.52% with 7-28-1.

Abstract: The implementation of a software acoustic echo canceller (AEC) on personal computers (PC) is much more difficult than just migration of the well defined algorithms. Software AECs have to deal with the diverse hardware environments, which are usually low-cost and inaccurate, and the software environment, which generally is committed to sharing resources through multi-tasking yet providing no real-time guarantees. Hardware differences lead to sampling rate differences and echo path nonlinearity, while the software environment can cause discontinuity of the audio stream. In this paper, these factors are investigated, and addressed by improved software AEC based on the coloration effect filter algorithm.

Abstract: The paper proposed a modified LMS algorithm of variable step size based on a brief analysis of traditional LMS,variable step size LMS algorithm and its improved algorithm.The novel algorithm based on nonlinear functional relationship between the step-size and the error ,increases adaptively at the beginning of the algorithm or when the channel is varying with time ,and it would be smaller during the steady state.So the algorithm has the excellences of faster constringency,little steady error ,tracking the change of the system and avoiding the effects of the noise. The theoretical analysis and computer simulation prove that the algorithm is better than traditional LMS algorithm.

Abstract: Stochastic resonance (SR) enhances the nonlinear system behavior with the assistance of noise. The energy-transfer mechanism makes the weak information revealed in the output spectrum, while the time-waveform is distorted. In this article, The Principle of stochastic resonance and recovery method are proposed and based on which, The nonlinear circuit is formed.At last, some examples are given to identify the performance of this nonlinear circuit.

Abstract: According to the shortcomings of the traditional fuzzy enhancement algorithms, several improvements are proposed. In this improved algorithm, the membership grade functions and fuzzy enhancement operator are made up of piecewise and continuous functions, the image is divided into two regions by OTSU method, one is high grey region, the other is low grey region, pixels in the high grey region are enhanced, and pixels in the low grey region are reduced. Simulation results show that this algorithm has good ability to enhance blur and little edges, and it is an effective and efficient way to increase image’s contrast.

Abstract: The blind source separation (BSS) using a two-stage sparse representation approach is discussed in this paper. We presented the algorithm based on linear membership function to estimate the unknown mixing matrix precisely, and then, the optimization algorithm based on integral to get the max value of the function is proposed. Another contribution described in this paper is the discussion of the impact of noise on the estimating the mixing matrix. Given the impact of noise, we set weights to put more emphasis on the more reliable data. Several experiments involving speech signals show the effectiveness and efficiency of this method.

Abstract: In this paper, a robust image watermarking method with S-Hough transformation is proposed which is robust against geometric distortion. This watermarking is detected by a linear frequency change. The chirp signals are used as watermarks and this type of signals is resistant to all stationary filtering methods and exhibits geometrical symmetry. In the two-dimensional Radon-Wigner transformation domain, the chirp signals used as watermarks change only its position in space/spatial-frequency distribution, after applying linear geometrical attack. But the two-dimensional Radon-Wigner transformation needs too much difficult computing. We propose a modified Hough transformation to provide improved energy concentration of the Dopplerlet transform. The proposed scheme can resolve the time-frequency localization in a better way than the standard Dopplerlet transformation. The watermark is embedded in the 1D improved Dopplerlet transformation domains. The watermark thus generated is invisible and performs well in test and is robust to geometrical attacks. Compared with other watermarking algorithms, this algorithm is more robust, especially against geometric distortion, while having excellent frequency properties.

Abstract: Chaos particle swarm optimization (CPSO) can not guarantee the population multiplicity and the optimized ergodicity, because its algorithm parameters are still random numbers in form. This paper proposes a new adaptive chaos embedded particle swarm optimization (ACEPSO) algorithm that uses chaotic maps to substitute random numbers of the classical PSO algorithm so as to make use of the properties of stochastic and ergodicity in chaotic search and introduces an adaptive inertia weight factor for each particle to adjust its inertia weight factor adaptively in response to its fitness, which can overcome the drawbacks of CPSO algorithm that is easily trapped in local optima. The experiments with complex and Multi-dimensional functions demonstrate that ACEPSO outperforms the original CPSO in the global searching ability and convergence rate.