Tetsuya Tabaru

A Combined Approach using Subspace and Beamforming Methods for Time-Frequency Domain Blind Source Separation

Blind source separation (BSS) is an approach to estimate original source signals only by measurements through multiple sensors without any prior knowledge about their mixture process. BSS is currently applied to a broad range of fields, such as pre-processing of speech recognition, analysis of bio-signals, etc. In BSS of convolutive mixtures, the signals must be expanded into time-frequency domains. In this case, permutation between frequencies becomes indefinite. This paper proposes a method for solving the indefiniteness problem by using directional peaks in addition to directional nulls in directivity patterns. This method makes it possible to combine the beamforming and subspace methods and to improve the accuracy in separatability of BSS. The effectiveness of the proposed method is demonstrated by computer simulation

Dead Time Detection Based on Wavelet Analysis of Cross Correlation Data

Abstract This paper proposes a dead time detection method for a linear system. Around the dead time, an impulse response of the system has self-similarity, which can be detected by the wavelet analysis. The proposed method detects the dead time from an wavelet transform of a cross correlation function between an input and an output, if the input satisfies a certain condition and a disturbance is small. A numerical simulation shows efficiency of the proposed method.

Reconsideration of Dead Time Measurement by Wavelet from Phase Property of Frequency Response

Abstract This paper provides another view of the dead time measurement method, which we have already proposed. The method measures a dead time of a linear system from a wavelet transform of a cross correlation function between an input and an output. On the other hand, we notice that the wavelet transform of the cross correlation function has a close relation to the frequency response of the system. The objective of this paper is to derive this relation and present another view of the dead time measurement method. In order to achieve the objective, we consider a relation between the wavelet transform of the cross correlation function and the corresponding cross spectrum.

INPUT-OUTPUT RELATION OF LINEAR SYSTEMS BY USING COMPLEX WAVELET PACKET TRANSFORM

This paper is concerned with an input-output relation of linear systems when using the complex wavelet packet transform. In general, the linear relation between input-output is guaranteed by the following two conditions: (1) the mother wavelet has a better frequency resolution, and (2) the real and imaginary parts in the mother wavelet consist of a Hilbert transform pair. The complex wavelet satisfying the above conditions has been used for demonstrating the linear relation. In this paper, a complex wavelet packet transform is applied for improving the frequency resolution. The validity of our approach is shown through a numerical experiment.

Dead Time Detection by Wavelet Transform of Cross Spectrum Data

Abstract This paper proposes a dead time detection method for a linear system. At the dead time, an impulse response of the system has self-similarity, which can be detected by the wavelet analysis. Hence, the dead time can be detected by estimating the impulse response and applying the wavelet analysis to it. The proposed method obtains the estimate of the impulse response from cross spectrum data and power spectrum data. A numerical simulation is shown to validate the method. An experimental result on an air conditioner plant is also given.

Bayes estimation of road surface using road noise

This paper proposes a novel and simple method of estimating a road surface by means of a road noise. The noise is measured with an IC voice recorder, and a Bayesian estimator decides a sort of surface based on the noise. It is assured that the estimator works well on actual cars which run on dry asphalt, wet asphalt, turf, and so on.

Dead time measurement of closed loop system by wavelet

This paper shows that the wavelet based dead time measurement method, which has been already studied for open loop systems. is also applicable to closed loop systems. The method uses a wavelet transform of a cross correlation function between an input and an output. To achieve our objective, we derive the wavelet transform for the. closed loop case.

WAVELET-BASED IDENTIFICATION FOR CONTROL OF WATER-TUBE DRUM BOILER

Abstract The procedure for identifying linear time-varying systems with dead times by using the wavelet analysis has been already proposed by the authors. This paper reports an extension of the identification procedure to MIMO systems and its application to actual control systems of water-tube drum boiler in a thermal power plant. The control systems are required to supply steadily high-quality steam against disturbances due to unexpected changes in load. Therefore, it is necessary to make a model which can take the disturbances into consideration for control system design. The transfer function models identified in stationary and non-stationary states are evaluated by the prediction errors in the process outputs; steam-pressure and water-level, in comparison with an ARX model.

Wavelet-Based Identification for Transfer Function with Dead Time and Its Application

Abstract Wavelet analysis is used in identifying a continuous-time system with dead time. First, we present a dead-time identification technique by using the wavelet transform of the cross-specta of input/output (I/O) data. Secondly, we present a method for obtaining the frequency response of the system to be considered by directly using the wavelet transform of the auto-/cross-eorrelation functions of the I/O data. Thirdly, we determine the parameters in the transfer function by using the curve-fitting in the least squares sense. The adventage of our identification procedure is that the estimates of parameters can be obtained separately from the delay-time estimate. Finally, we demonstrate the feasibility and validity of our procedure through experiments with an actual boiler process.