Gait recognition is a promising non-intrusive biometric method. A robust and compact gait model is desirable in many security applications from public facilities to personal devices. Shape cues are chosen in most curr...
详细信息
Gait recognition is a promising non-intrusive biometric method. A robust and compact gait model is desirable in many security applications from public facilities to personal devices. Shape cues are chosen in most current researches except a few adopting dynamical features exclusively. And most of these systems are velocity-dependent. In order to explore more features of gait and to fit the varying environments of different applications, a new gait recognition model which synthesizes dynamic model and statistical one is designed. A kind of dynamical features, angular variables with respect to ankle joint, are adopted as the model's input. The proposed model has a circular structure consisted of 2 pairs of correlated states. A constrained learning algorithm is proposed under the model's special structure configured according to a 2-link virtual passive walking model which plays an important role both in the initialization and in the updating step. By evaluating the recognition rates of different models, the velocity-robust characteristics of the new model and its low computational load compared with conventional HMM are verified.
In this paper, two modified constrained learning algorithms are proposed to obtain better generalization performance and faster convergence rate. The additional cost terms of the first algorithm are selected based on ...
详细信息
In this paper, two modified constrained learning algorithms are proposed to obtain better generalization performance and faster convergence rate. The additional cost terms of the first algorithm are selected based on the first-order derivatives of the activation functions of the hidden neurons and the second-order derivatives of the activation functions of the output neurons, while the additional cost terms of the second one are selected based on the first-order derivatives of the activation functions of the output neurons and the second-order derivatives of the activation functions of the hidden neurons. In the course of training, the additional cost terms of the proposed algorithms can penalize the input-to-output mapping sensitivity and the high frequency components simultaneously so that the better generalization performance can be obtained. Finally, theoretical justifications and simulation results are given to verify the efficiency and effectiveness of our proposed learningalgorithms. (c) 2007 Elsevier Inc. All rights reserved.
In this paper, a modified learningalgorithm to obtain better generalization performance is proposed. The cost terms of this new algorithm are selected based on the second-order derivatives of the neural activation at...
详细信息
In this paper, a modified learningalgorithm to obtain better generalization performance is proposed. The cost terms of this new algorithm are selected based on the second-order derivatives of the neural activation at the hidden layers and the first-order derivatives of the neural activation at the output layer. It can be guaranteed that in the course of training, the additional cost terms for this algorithm can penalize both the input-to-output mapping sensitivity and the high frequency components to obtain better generalization performance. Finally, theoretical justifications and simulation results are given to verify the efficiency and effectiveness of the proposed learningalgorithm.
This paper makes the detailed analyses of computational complexities and related parameters selection on our proposed constrainedlearning neural network root-finders including the original feedforward neural network ...
详细信息
This paper makes the detailed analyses of computational complexities and related parameters selection on our proposed constrainedlearning neural network root-finders including the original feedforward neural network root-finder (FNN-RF) and the recursive partitioning feedforward neural network root-finder (RP-FNN-RF). Specifically, we investigate the case study of the CLA used in neural root-finders (NRF), including the effects of different parameters with the CLA on the NRF. Finally, several computer simulation results demonstrate the performance of our proposed approach and support our claims. (c) 2004 Elsevier Inc. All rights reserved.
This paper proposes a novel recursive partitioning method based on constrainedlearning neural networks to find an arbitrary number (less than the order of the polynomial) of (real or complex) roots of arbitrary polyn...
详细信息
This paper proposes a novel recursive partitioning method based on constrainedlearning neural networks to find an arbitrary number (less than the order of the polynomial) of (real or complex) roots of arbitrary polynomials. Moreover, this paper also gives a BP network constrained learning algorithm (CLA) used in root-finders based on the constrained relations between the roots and the coefficients of polynomials. At the same time, an adaptive selection method for the parameter d P with the CLA is also given. The experimental results demonstrate that this method can more rapidly and effectively obtain the roots of arbitrary high order polynomials with higher precision than traditional root-finding approaches.
This paper proposes using recursive root moment method (RRMM) based on feedforward neural networks (FNN) trained by constrained learning algorithm (CLA) to find the roots of polynomials, which is of fewer computationa...
详细信息
ISBN:
(纸本)0780374908
This paper proposes using recursive root moment method (RRMM) based on feedforward neural networks (FNN) trained by constrained learning algorithm (CLA) to find the roots of polynomials, which is of fewer computational complexity than the root moment method (RMM) and the method using the relations between the roots and the coefficients (RRC) of polynomials. As a result, the RRMM is of faster training speed and higher accuracy than the latter two methods. The experimental results verify, our claims.
This paper proposes using recursive rootmoment method (RRMM) based on feedforward neuralnetworks (FNN) trained by constrainedlearningalgorithm (CLA) to find the roots of polynomials,which is of fewer computational co...
详细信息
This paper proposes using recursive rootmoment method (RRMM) based on feedforward neuralnetworks (FNN) trained by constrainedlearningalgorithm (CLA) to find the roots of polynomials,which is of fewer computational complexity than theroot moment method (RMM) and the method using therelations between the roots and the coefficients (RRC)of polynomials. As a result, the RRMM is of fastertraining speed and higher accuracy than the latter twomethods. The experimental results verify our claims.
暂无评论