本稿は，極超長波帯にて観測された環境電磁波信号に含まれる異常な電磁放射信号の検出を目的とし，外積展開法による背景雑音除去および異常電磁放射信号の推定精度について検討を行うものである．著者らは，複数のセンサから観測された観測信号に共通する背景雑音と，低頻度で観測される局所成分の分離という視点に立ち，外積展開法と非線形フィルタによる背景雑音の推定手法を提案した．外積展開法は，観測点に独立な白色雑音が含まれる多点観測信号に適用した場合，背景雑音の推定精度が低下する問題がある．しかしながら，観測雑音を想定した白色ガウス雑音と観測信号のSN 比に対する推定精度の傾向は明らかとされていない．さらに，実際の観測データを用いた背景雑音除去に関する検討は皆無と言える．本稿では，疑似信号を用いて外積展開法が有するSN 比と背景雑音除去精度の関連を明らかにし．さらに極超長波帯環境電磁波信号に含まれる異常電磁放射信号の推定精度について検討を行う．%This paper shows the denoising performance of the outer product expansion with non-linear filters for the background noise included in the ELF (Extremely Low Frequency) electromagnetic waves. We have proposed novel source separation techniques based on an outer product expansion with non-linear filter. This technique is the signal separation that the background noise, which is observed in almost all input signals, can be estimated. The effectiveness of outer product expansions for artificial signals is represented. These methods have a problem; the estimated result is strongly affected by Gaussian random noise. However, the performance of background noise estimation in the noisy input signal has not been discussed enough. Moreover, the denoising accuracy for real data has not been shown in the conventional researches. In this paper, 3 background noise estimations using an outer product expansion are applied to the noise reduction problem in the artificial signals and electromagnetic wave analysis to evaluate the denoising performance for noisy signals and global noises.
展开▼
