Keisuke Inoue

Time frequency analysis of biological signals during sleep

We have developed a relevance between sleep stage (Stage MT, W, 1, 2, 3, 4, REM) and rhythm of autonomic nerve activity computed by frequency analysis of heartbeat interval. Heartbeat during sleep was changed by sleep state. Extraction of relationships between sleep stage and heartbeat interval from signal of pressure sensor located in bed is goal. Then, we tried to apply continuous wavelet analysis to ECG in order to extract more precise information for the relationships. Moreover, we applied wavelet transform of impulse as mother wavelet that is impulse response of signal of pressure sensor on one heartbeat. In this paper we had six all-night sleep experiments, and compared result each subjects. As a result, we were able to estimate sleep state freely from sensor of pressure signal.

Influence of undetected ECG R waves on the heart rate variability analysis

As the sleep disorders become the issue, we need to know the quality of ourselves sleep. In general, the quality of sleep is measured by polysomnography (PSG). But, this method is a time-consuming process. Therefore an easy method is needed. The estimation of sleep stages from a heart rate variability for 1 minute in sleep is a more easy method. A deepness of sleep and heart-rate variability is influenced by automatic nerve. They are closely related. Therefore, if we can know heart-rate in sleep, we can know the deepness of sleep. The algorism that estimate of sleep stage form heart-rate are known. But, when a subject turned over, the R waves of heart-rate is undetected. In general, the sleep stages in the 1 minutes are not estimated. It is possible to estimate sleep stages if there are several undetectable R waves. LF and HF obtained from ECG are used to estimate sleep stage. And so, in this paper, we verify whether to estimate LF and HF from heart rate, when heart-rate is undetected.

Tracking a Moving Object Employing a Silicon Retina and FPGA Circuits

A real time image processing system composed of a neuromorphic silicon retina and FPGA circuits was designed. Using the system, target tracking was carried out under indoor illumination. In the tracking operation, the silicon retina distinguished a moving object from background objects by subtracting consecutive image frames after smoothing. Then, the image of the moving object was binarized and the medians of the objects horizontal and vertical histograms were found. The system was found to be able to execute robust image processing under indoor illumination using compact hardware and low power consumption. Therefore, it is considered to be suitable for various engineering applications such as robot vision.

A target tracking employing a silicon retina system

We have developed a silicon retina system, and applied it to target tracking in natural scenes. In this system, a silicon retina, which is capable of smoothing and contrast enhancement on input images, is used to detect edges of an object in real time. Experimental results show that this system can track the target in real time.

Multiresolution representation of 3D objects using B-spline surface model based on projection expression

This paper presents a method for multiresolution representation of 3D objects using an extended B-spline surface model based on a projection expression. In this method, a multiresolution representation based on multiresolution wavelet transform is implemented with the corresponding B-wavelets. The surface models at each resolution level are obtained by performing a decomposition algorithm, after the surface model at the maximum level is estimated.

Investigation on the applicability of a quick identification-based diagnosis method to various fault modes

The paper deals with identification-based fault diagnosis for nonlinear continuous-time dynamic systems. As a suitable way for monitoring physical parameters which specify fault modes in the system, a quick identification method has been proposed. In the paper, the applicability of the method to various fault modes is investigated through simulation studies on a ship propulsion system.

A clustering method based on quadric surface for segmentation of range data

In this paper, a clustering method is proposed to segment a range image of closed object. The data points are classified into domains with similar quadric surface. This method consists of two stages. First, new clusters are created on demand. Next, dispensable clusters are vanished by the competition of clusters, and remaining clusters are agglomerated. Consequently, optimal clusters are determined.