Applied Mechanics and Materials Vols. 303-306

Abstract: Based on the swarm behaviors in shoal of fish, this paper analyzes motion behavior of fish school when fish are foraging. This paper has simulated the pattern of motion behaviors for the fish school quantificationally. Using iterative and updating algorithm, it has modeled the 1-D procedure when fish school gradually approach food. These individual fish that constitute shoal of fish have finite perception and obey simple behavior rules. This paper further adopts modified intelligent optimization algorithm to model the process that individual fish interact with each other, changing position and velocity to gradually be close to food for the 2-D case. Experiment result shows the modified model successfully simulates the swarm behavior in shoal of fish when fish are looking for food, with small relative and absolute error.

Abstract: The authors proposed a trainable formant synthesis method based on the multi-channel Hidden Trajectory Model (HTM). In the method, the phonetic targets, formant trajectories and spectrum states from the oral, nasal, voiceless and background channels were designed to construct hierarchical hidden layers, and then spectrum were generated as observable features. In model training, the phonemic targets were learned from one-hour training speech data and the boundaries of phonemes were also aligned. The experimental results showed that the speech could be reconstructed with the formant trainable model by a source-filter synthesizer.

Abstract: Despite the appearance of high-tech human computer interface (HCI) devices, pattern recognition and gesture recognition with single camera are still playing vital role in research. A real-time human-body based algorithm for hand gesture recognition is proposed in this paper. The basis of our approach is a combination of moving object segmentation process and skin color detector based on human body structure to obtain the moving hands from input images, which is able to deal with the problem of complex background and random noises, and a rotate correction process for better finger detection. With ten fingers detected, more than 1000 gestures can be recognized before concerning motion paths. This paper includes experimental results of five gestures, which can be extended to other conditions. Experiments show that the algorithm can achieve a 99 percent recognition average rate and is suitable for real-time applications.

Abstract: Common spatial pattern (CSP) is based on sample covariance matrix estimation, whose estimation accuracy is severely overfit with small training sets. To address the drawback, the regularized CSP (R-CSP) was proposed that adds regularization information into the CSP learning process. In the algorithm, all samples of each generic subject were used for training sample covariance matrices. When only a part of the samples of each generic subject are allowed as generic training set, this R-CSP algorithm wouldn’t work. To solve this problem, an improved method is proposed in this paper. The new algorithm was applied to a brain-computer interface (BCI) data set containing five subjects and a mean improvement of 2.5% in classification rate was achieved.

Abstract: Directing to the dispersiveness and faintness failure characteristics of hydraulic excavator, the fault diagnosis method was presented based on the fault tree and fuzzy neural network. On the basis of analysis of the hydraulic excavator system works, the fault tree model of hydraulic excavator was built by using fault diagnosis tree. And then, utilizing the example of hydraulic excavator fault diagnosis, the method of building neural network, obtaining training samples and neural network learning in the process of intelligent fault diagnosis are expounded. And the status monitoring data of hydraulic excavator was used as the sample data source. Using fuzzy logic methods the samples were blurred. The fault diagnosis of hydraulic excavator was achieved with BP neural network. The experimental result demonstrated that the information of sign failure was fully used through the algorithm. The algorithm was feasible and effective to fault diagnosis of hydraulic excavator. A new diagnosis method was proposed for fault diagnosis of other similar device.

Abstract: The underwater moving targets classfication are important for the underwater surveillance system.This paper presents the classification algrithms based on the multi-feature fusion, including the target echo highlight features,temporal features from the assocition images interframe,and the moving features after tracking.The principal compoment analysis was used to reduce the feature dimension and the k-means algorithm was used for classification. At last,the experiment results of the classification between the divers and underwater vehicles are given, which show that the multifeature fusion can improve the classification performance.And the PCA algorithm can reduce the feature dimension without lower the identification probability.

Abstract: Due to the huge loss to the bank caused by credit risk in the financial debt crisis of the international Banking industry in 1980s, the research on Credit Assessment Methods is becoming the central issue of the study of financial theory in China and abroad. This paper builded the assets financial evaluation system of credit risk level based on the association rules-Apriori algorithm of data mining technology, which aimed the problems and the serious shortage of risk quantification study in domestic banks credit risk management. At the same time, taking into account the actual situation of our country, this paper analyzed that there are certain difficulties to use modern credit risk measurement models to evaluation the credit risk of commercial banks. And it suggests building a credit portfolio risk measurement model suitable for China's commercial banks with using logistic regression model of data mining technology.

Abstract: The observation on motion and behaviors of the animal is still important in scientific research fields, and various behaviors of animals are tracked and analyzed for many different purposes. The paper concentrates on the technologies employed for animal tracking in video, and reviews the tracking models related to animal motions and behaviors including shape-based model, contour-based model, articulated model, Bag-Of-Feature based model, Markov Model, etc. and gives a brief summary of these models respectively.

Abstract: The basic and improved algorithms of PSO focus on how to effectively search the optimal solution in the solution space using one of the particle swarm. However, the particles are always chasing the global optimal point and such points currently found on their way of search, rapidly leading their speed down to zero and hence being restrained in the local minimum. Consequently, the convergence or early maturity of particles exists. The improved PSO is based on the enlightenment of BP neural network while the improvement is similar to smooth the weight through low-pass filter. The test of classical functions show that the PSO provides a promotion in the convergence precision and calculation velocity to a certain extent.

Abstract: This paper studies an improved artificial bee colony algorithm, and two problems have been solved when the artificial colony algorithm is applied to objective optimization: the problem of slow convergence and premature aging problem. When the improved artificial bee colony algorithm is applied to land resources optimization problems, studies show the following two points. First, compared with the genetic algorithm, particle swarm optimization algorithm, and differential evolutionary algorithm, artificial bee colony algorithm has better adaptability and robustness in solving multivariate and multi peak global optimization problems. Second, compared with artificial bee colony algorithm, the improved artificial bee colony algorithm converges faster, the overall fitness increases by 8.9%, the maximum error is no more than 1%, and the short and medium term optimization has a high precision.