@article{oai:niigata-u.repo.nii.ac.jp:00002674,
author = {菊池, 久和 and 中静, 真},
issue = {4},
journal = {シミュレーション, シミュレーション},
month = {Dec},
note = {In this paper, we introduce some extensions of popular wavelet transforms and their applications to signal analysis. Two applications are described : electro cardiogram (ECG) analysis by irregular sampling of the wavelet transform and wavelet packed basis construction for detonation diagnosis of combustion engines. In the characteristic points analysis of ECG waveforms, we apply the wavelet maxima representation to ECG. The wavelet maxima are defined as the modulus maxima of a dyadic wavelet transform. The entire signal is represented by the set of all wavelet maxima. In ECG analysis, we define a basic wavelet as the second-order derivative of a smoothing function. With this wavelet, wavelet, wavelet maxima indicate the location of turning points in ECG waveforms. The decaying behavior along scale in the wavelet maxima indicates the differentiability at turning points. By exploiting this multiscale property, we apply the wavelet maxima to the classification and recognition of characteristic points in ECG. ECG compression, analysis and transmission by the wavelet maxima representation are shown for possible new applications. Detonation detection is critical for clean emission of vehicle engine control, and wavelet packets are applied to it. A wavelet packet system is obtained from arbitrary combinations in wavelet filter banks. An entire set of particular orthogonal transforms are implemented by a single wavelet packet system, and the set covers the orthogonal transforms of which time-frequency decomposition approximates the various decompositions produced by the wavelet, Fourier and short-time Fourier transforms. The best basis for a particular application is selected from the entire set of the wavelet packet bases. To detect detonation, we define a Maharanobis distance criterion for the best basis search. Successful detection was demonstrated in viewing the discriminant score of 96%.},
pages = {261--270},
title = {ウェーブレット変換による異常探知 : 広がるウェーブレット変換の世界},
volume = {16},
year = {1997}
}