This study proposes a new representation method of an input ground motion model as a non-stationary stochastic process based on the short-time Fourier transform (STFT). The inverse STFT of the product of a deterministic amplitude function concerning time and frequency and the Fourier transform of a Gaussian white noise gives the input ground motion model. This definition implements the ground motion model as an evolutionary process and delivers the expression of the non-stationary power spectral density function. The sample waveform functions generated numerically from the stochastic model keep expected well-designed characteristics. Furthermore, a method of random response analysis of an elastic SDOF system excited by the input ground motion model is carried out and used to estimate its maximum response. Some numerical simulations demonstrate the effectiveness and accuracy of the proposed method.

抄録全体を表示

代表的な手術時の鎮静度指標として脳波を利用したBispectral IndexおよびEntropyがあげられる。これらの算出にはが用いられているが、利用する脳波の時間幅は、Bispectral Indexについては一定であり、Entropyについても解析周波数帯域に応じて段階的に変化させているだけで、いずれも時間分解能が悪い。本研究では、より良い鎮静度指標を作ることを目的として、時間分解能の良いウェーブレット変換を用い、を用いた場合と比較検討する。

抄録全体を表示

近年我が国を筆頭にインフラ維持管理の効率化を目指す動きが活発である．著者は現在まで高解像度カメラ，レーザー，レーダーなどのデバイスを，またソフトについてはディジタル信号処理とバスワードとなったAI を活用し，インフラ維持管理情報を自動で，効率的に，高速で構築するための技術開発を行っている．若手研究者として実際にハード・ソフトを自ら開発してきた経験から，近い将来に現実的にインフラ維持管理に革新をもたらすと期待している技術の開発動向について述べる．

抄録全体を表示

心電はストレス検出や血圧推定等に用いられる重要な人体モニタリング指標である。本研究では指向性マイクを用いて非接触で測定した心音から心電を推定することを目指す。基礎的な研究として、心音の特性の調査を行い、

を用いて周波数成分に分解した心音から心電のピークの推定を行った。その結果、心音から心電のピークが推定され、その精度を評価した。

抄録全体を表示

自動コード推定は，楽曲分析の分野における重要な研究課題である．既存研究において，DCGANを用いた手法が提案されているが，自動コード推定の正解率は73.2%に留まっている．そのため本研究では，DCGANを用いた手法の改良と，CGANを用いた手法を提案した．DCGANを用いた手法では，全結合層と出力層にtanh関数を使用することで，値域の統一を行い，Minibatch Discrimination層とすべての畳み込み層にL2正則化を追加して，Mode Collapseと過学習を抑制するように改良した．実験の結果，自動コード推定の正解率は91.7%となった．また，CGANを用いた手法では，条件の入力にOne-hotベクトルを採用し，MIDIデータの末尾に付加する手法と，MIDIデータの先頭を置換する手法を検証した．実験の結果，2つの手法の自動コード推定の正解率は，それぞれ90.5%，90.9%となった．本研究により，自動コード推定において，DCGANを用いた手法の正解率を向上させることが可能となった．さらに，CGANを用いた手法も，DCGANを用いた手法にはやや劣るものの，高い精度の正解率となることが示された．

抄録全体を表示

In non-destructive testing using ultrasound, a reliable signal processing method which separates the superimposed echoes is needed to detect defects. In this study, we propose a new signal processing method using the phase of ultrasonic echoes. When defects are located near the surface of the inspected object, it is difficult to detect defects because the echoes reflected by the defect are superimposed on ones reflected by the surface of the inspected object. To separate the superimposed echoes, the authors focused on points on the phase distribution of the wavelet transform of the echoes. This point can be used to estimate the location and size of defects, but its applicability is limited. However, there is not theoretically the limitation when using the short-time Fourier transform. In this paper, some samples with a hole were prepared and superimposed echoes were measured by using ultrasonic waves. The depth and diameter of the hole were estimated by applying wavelet transform and short-time Fourier transform to the superimposed echoes. It provides that the use of the short-time Fourier transform enabled estimation over a wider range.

抄録全体を表示

　　　Rheological behavior of interstitial fluid in epidermal tissue, including blood micro-circulation, can vary skin mechanics in micro scale, i.e. visco-elasticity. Therefore, an in vivo quantitative measurement of fluid velocity is necessary to clarify their properties. In this paper, we constructed OCDV (Optical Coherence Doppler Velocigraphy), which is a method of tomographic micro-visualizing interstitial fluid velocity using a high frequency modulated low coherence interferometer. The tomographic detection of modulated Doppler frequency can determine fluid velocity profile within tissue. An optimal signal processing of OCDV was evaluated alternately on short time FFT and Hilbert transform, and validated by measuring the fluid velocity of Intralipid solution (2%) in micro channel. Consequently, Hilbert transform has further advantages in the accuracy even with a rapid scanning system than short time FFT. OCDV was suggested to be effective to diagnose flow velocity profiles in tissue non-invasively and tomographically.

抄録全体を表示

Rheological behavior of interstitial fluid in epidermal tissue, including blood micro-circulation, can vary skin mechanics in micro scale, i.e. visco-elasticity. Therefore, an in vivo quantitative measurement of fluid velocity is necessary to clarify their properties. In this paper, we propose OCDV (Optical Coherence Doppler Velocigraphy), which is a method of tomographic micro-visualizing interstitial fluid velocity using a high frequency modulated low coherence interferometer. The tomographic detection of modulated Doppler frequency can determine fluid velocity profile within tissue. OCDV system was constructed on the signal processing of Hilbert transform, and validated by measuring the fluid velocity of Intralipid solution (2%) in micro channel. Consequently, Hilbert transform has further advantages in the accuracy even with a rapid scanning system than short time FFT. OCDV was suggested to be effective to diagnose flow velocity profiles in tissue.

抄録全体を表示

This paper presents a time-frequency distribution which is suitable for a signal analysis. It is difficult to obtain a high resolution in both time and frequency domains in the time-frequency analysis of the signal. However, the multiple short-time Fourier transform (NISTFT) is proposed as a means which obtains high resolution in both time and frequency domains. Further, it is extended to achieve higher resolution signal analysis using multiple expression of the MSTFT. The diagnosis of rotating condition of the pumps is attempted with driving sound using several expressions of time-frequency distributions based on the MSTFT.

抄録全体を表示

In playing the golf, using suitable golf club is important to put out high scores and to hit the correct shot. So, various studies have been conducted on golf clubs and swings to select an appropriate golf club. However, only a few studies are available on how the swing changes when the physical characteristics of the club are changed. The purpose of this study was to analyze the torque data during the swing by using the short-time Fourier transform (STFT) to clarify how the mass change of the shaft affects the swing. First, we measured the golfer's swing at the time of change in shaft mass using a motion capture system and created a three-dimensional rigid body dynamics model consisting of arms and clubs. Next, we calculated the torque data exhibited during the swing using inverse dynamics calculation. After we applied the STFT, confirmed the main frequency component and extracted common points between testers by ANOVA and multiple comparison test. Finally, we confirmed the effect of shaft weight on the spectrogram in each tester.

抄録全体を表示

呼吸器疾患の診断方法としては，聴診器を用いた呼吸音の聴診が長年用いられてきた．これは簡便で安全な診断方法である一方，聴診音の診断には定量的な評価基準がないため，医師の診断支援を行うシステムの開発が必要である．そこで本論文では，畳み込みニューラルネットワーク（CNN： convolutional neural network）を用いた呼吸音の自動分類手法の提案を行う．おもな手法の流れとしては，呼吸音データに対して

と連続ウェーブレット変換を適用し，スペクトログラム画像およびスカログラム画像を生成する．その後，生成した画像を用いてCNN による正常呼吸音，連続性ラ音，断続性ラ音の識別を行う．提案手法を呼吸音データ22 症例に適用した結果，分類性能として，全体正解率は 79.44［％］，ROC（receiver operating characteristic）曲線に基づくAUC（area under the curve）は0.942 を得た．

抄録全体を表示

1．音楽と物理学

音は物理学の対象だが，音楽は芸術であり，物理学で扱おうとしても一筋縄ではいかない．ヘルムホルツの歴史的な著書¹⁾は音楽を聴くことを出発点として，まず聴覚の生理を物理に還元するこ

抄録全体を表示