Tadaaki Shimizu

An algorithm to determine neighbor nodes for Automatic Human Tracking System

The determination algorithm of neighbor video cameras by deployed position and view distance of video camera is introduced in this paper. This algorithm is utilized to determine the neighbor video cameras in the Automatic Human Tracking System. The Automatic Human Tracking System is enhanced Video Monitoring System utilizing the Mobile Agent Technologies. Information of the neighbor video cameras is needed in the Video Monitoring System during the transition of video display. And the mobile agent in the Automatic Human Tracking System also utilizes such information to pursue the human efficiently. On the event of modifying the installed cameras in the system the following complications occur: Updating of the camera information and computation of the neighbor cameras. As such the Automatic Human Tracking System have inherited the above problems. For this reasons, the algorithm is developed to solve the mentioned problems.

An adaptive noise reduction filter for discrete signal by use of sandglass-type neural network

An adaptive noise reduction filter composed of a sandglass-type neural network (SNN) noise reduction filter (RF) is proposed in this paper. SNN was originally devised to work effectively for information compression. It is a hierarchial network and is symmetrically structured. SNN consists of the same number of units in the input and output layers and a smaller number of units in the hidden layer. It is known that SNN has signal processing performance which is equivalent to Karhunen–Loeve expansion after learning. We proved the theoretical suitability of SNN for an adaptive noise reduction filter for discrete signals. The SNNRF behaves optimally when the number of units in the hidden layer is equal to the rank of the covariance matrix of the signal components included in the input signal. Further we show by applying the recursive least squares method to learning of the SNNRF that the filter can process signals for on-line adaptive noise reduction. This is an extremely desirable feature for practical application. In order to verify the validity of SNNRF, we performed computer experiments examining how the noise reduction ability of SNNRF is affected by altering the properties of the input pattern, learning algorithm, and SNN. The results confirm that the SNNRF acquired appropriate characteristics for noise reduction from the input signals, and markedly improved the SNR of the signals.

TPUnit neural network and simple ensemble for abnormal shadow detection in lung X-ray images

We have constructed systems that detect abnormal areas of lung X-ray images from one-dimensional numeric sequences using neural networks. In these systems, the neural network consists of neurons that use trigonometric polynomials as activation functions, or TPUnit neural networks. The TPunit neural network has a high generalization ability in a smaller number of hidden units. Several TPUnit neural networks are placed in parallel and their outputs are processed as a simple ensemble. ROC curves denoted performance greater than that of previous reports. In addition, the AUC (area under curve) value was 0.9998 and the EER (equal error rate) was 0.5363%. Experimental results indicate that this proposed system is useful for medical imaging diagnosis.

EEG Signals Classification by Using an Ensemble TPUnit Neural Networks for the Diagnosis of Epilepsy

The electroencephalogram (EEG) is necessary for the diagnosis of epilepsy. To make a diagnosis of epilepsy exactly, a full EEG recording for a long stretch of time is needed. The observation for a long record is a big burden for a doctor. To reduce this burden, a computer aid is important. This paper presented classifications of EEG patterns using the ensemble TPunit NNs for the diagnosis of epilepsy. The classification accuracy rates of the proposed classifiers were found to be higher than that of stand alone neural network. In addition, the classification accuracy was higher than previous study. The ensemble of the TPUnit neural networks is highly effective in classification problem.