Search

Your search keyword '"Blind source separation"' showing total 18 results
Start Over Descriptor "Blind source separation" Language japanese
18 results on '"Blind source separation"'

1. Neural network-based selective noise control with sound source separation

Author

Takumi SUNOHARA and Toshihiko SHIRAISHI

Subjects
noise cotrol, sound source separation, blind source separation, neural network, network design, experiment, time-variant system, acoustic environment, Mechanical engineering and machinery, TJ1-1570, Engineering machinery, tools, and implements, TA213-215
Abstract

In noisy environments such as factories or construction sites, noise should be reduced and the speech of workers moving in the environments should be recognized. Noise control is a method for reducing existing sounds by producing a secondary sound with the same amplitude but with antiphase, and effective on low frequency noise reduction. However, conventional noise control systems reduce all existing sounds but cannot preserve a specific target sound such as speech in noise. In this study, we propose a novel system which selectively reduces noise and preserves target sound by combining a noise control system with a sound source separation system using multilayer neural networks. The sound source separation part of the system was designed to separate mixture into noise for reduction and target sound for preservation and the use as a teaching signal of supervised learning of the neural networks. The noise control part of the system was designed to reduce noise at fixed and moving evaluation points. The whole system was designed to realize high noise reduction performance preserving target sound and low computational cost for real-time control in experiments. The performance of the proposed system was experimentally verified under the conditions that the noise and target waveforms were set as both sine waves, both superposed sine waves, and band noise and a sine wave, respectively. The results indicate that the proposed system preserves target sound to some extent and reduces noise by more than 10 dB and approximately 5–6 dB at the fixed and moving evaluation points, respectively.

Published
2023
Full Text
View/download PDF

2. Operational modal analysis of under-determined system based on Bayesian CP decomposition

Author

Sunao TOMITA and Tomohiko JIMBO

Subjects
operational modal analysis, modal properties extraction, blind source separation, tensor decomposition, candecomp/parafac decomposition, bayesian inference, under-determined problem, second-order blind identification, Mechanical engineering and machinery, TJ1-1570, Engineering machinery, tools, and implements, TA213-215
Abstract

Modal properties such as natural frequencies, modal shapes and damping ratio are useful to understand structural dynamics of mechanical systems. To use the modal properties for structural health monitoring, they need to be estimated under operational conditions. Therefore, operational modal analysis (OMA), extraction of the modal properties without input signals, has been proposed to easily extract the modal properties under operational conditions. Recently, OMA for underdetermined systems, i.e. number of measurements is less than that of active modes, has been paid attention to reduce the number of sensors. This paper proposes the OMA framework for the underdetermined systems based on Bayesian tensor decomposition of second-order statistics data. The proposed method enables us to extract the modal properties from underdetermined systems without tuning the number of active modes because rank of the tensor data corresponding to the number of the active modes is automatically determined via Bayesian inference. To show advantage of the method, the modal properties are extracted from artificial vibration data obtained from a mass-spring system under the operational and the underdetermined conditions.

Published
2021
Full Text
View/download PDF

5. Study on a permutation problem of Blind Source Separation

Subjects
echo canceller, independent component analysis, blind source separation
Abstract

Substitution problem in blind signal isolating circuit was examined on the system in the echo canceller that blind signal separation proposed in advance was pretreated. As FDSOS using actual sound condition and voice sample the algorithm of the blind signal separation the basic performance it evaluates. Next, the combination of the parameter that it arranges parameter in the algorithm and relationship between the required condition variously and maximizes the separating function by the experiment it investigates investigate.

Published
2007

6. An analysis of source separation performances in BSS based on transmission delays in convolutive mixture

Subjects
ブラインドソースセパレーション, 遅延時間, 畳み込み, Dalay time, 収束性, Network structure, Blind source separation, 学習アルゴリズム, Convergence, Learning algorithm, Convolutive, 回路構成
Abstract

金沢大学理工研究域 電子情報学系, ブラインドソースセパレーション(BSS)において現在,フィードフォーワード形(FF-)とフィードバック形(FB-)という2つの回路構造が提案されている.FF-BSSでは自由度が存在するため信号歪みが生じる可能性がある.一方,FB-BSSは本質的に信号歪みを発生しにくい構造であるが,信号源からセンサーまでの遅延時間に条件が課せられる.この観点から,FF-BSSとFB-BSSの信号源分離性能についていろいろな条件下で解析と比較を行い,各々が有効に適用できる範囲を明らかにした. Feed-Forward (FF-) and FeedBack (FB-) structures have been proposed for Blind Source Separation (BSS). The FF-BSS systems have some degree of freedom in the solution space, and signal distortion is likely to occur in convolutive mixtures. On the other hand, FB-BSS structure is hard to cause signal distortion. However, the FB-BSS system requires a condition for the transmission delay time in the mixing process. The FF-BSS systems and the FB-BSS system are compared based on the transmission time delay in the mixing process. Even though the FB-BSS can provide good separation performance, there exit some limitations on location of the signal sources and the sensors.

Published
2007

7. Learning Algorithm Improvement and Source Separation Analysis in Cascade BSS Applied to Nonlinear Mixture

Subjects
継続形構成, 線形化, nonlinear mixture, number of sensors, blind source separation, センサー数, 分離性能, cascade form, linearization, separation performances, ブラインド信号源分離, 非線形混合
Abstract

金沢大学理工研究域 電子情報学系, 信号源の統計的性質を利用して,混成信号から信号源を分離する方法をブラインド信号原分離という.実際の混合過程では非線形歪みを生じる場合も多く,分離がより複雑になる.このような場合に対して分離と線形化を縦続に構成する方法が提案されている.本研究では,この縦続形構成における線形化に対して安定な学習法,すなわち,初期値の設定法及びパラメータの学習法に関して新しい方法を提案し,非線形成分を大幅に抑制できた.次に,センサー数と分離性能の関係をクロス項s_is_j,i=⃥jの大きさに基づいて解析した.観測信号において主信号が干渉信号の2倍程度あればセンサー数を低減できること,干渉信号が大きい場合は分離が難しい問題であり,センサー数の低減も難しいこと,センサー数を低減した場合は学習が遅くなることを確認した. A cascade form blind source separation (BSS) has been proposed for nonlinear mixture. It consists of a signal separation block and a linearization block in this order from the input to the output. In this paper, firstly, a learning process for the linearization block is improved. New ideas are introduced for initialization and adjusting of parameters. Performance of linearization is well improved. Furthermore, by ordering the new linearization block and the signal separation block in this order, the number of sensors can be well reduced. For instance, 3 sensors are enough to 3 sources. Secondly, a relation between separation performance and the number of sensors is analyzed based on the cross term components s_is_j, i=⃥j. As the learning makes a progress, separation performances for the different number of sensors approach to the same value. In difficult problems, separation performances are affected by the number of sensors.

Published
2007

8. Improvement of A Learning Algorithm for Feedback Over-Complete BSS

Subjects
音声信号, スペクトルサプレッション ヒストグラム, Suppression, Over-complete, Spectral, Histogram, Speech signal, フィードバック, ブラインド信号源分離, オーバーコンプリート, Feedback, Blind source separation, 学習アルゴリズム, Learning algorithm
Abstract

金沢大学理工研究域電子情報学系, 信号源数がセンサ数より多いオーバーコンプリート形ブラインド信号源分離(BSS)で,信号源の完全分離は困難という問題に対して,我々はフィードバック形の回路構成と学習法を既に提案している.まず,1個の信号源が複数の出力に含まれないことを分離の条件として,1巡目の信号源分離を行ない,少なくとも1個の出力に単一信号源を分離する.この出力を観測信号にフィードバックして,観測信号から単一信号源をキャンセルすることにより,等価的に信号源数を低減する.単一信号源のキャンセル法として,当該出力と推定した混合過程の情報を使ってキャンセルする他,観測信号のヒストグラムを使ってキャンセルする方法を組み合わせる.本稿ではさらに,フィードバックにより生じる信号歪みの問題に対して,スペクトルサプレッション法を導入することにより,信号歪みを抑制する.2巡目の信号源分離では,観測信号に含まれる信号源が1個少ない状態で1巡目とは異なる学習法により処理を行なう.信号源として音声を用いたシミュレーションにより,提案方法の有効性を確認した. A feedback structure and its learning algorithm were proposed for over-complete blind source separation, where the number of sources is larger than that of sensors. In the first phase, the signal sources are separated so as to satisfy the condition, under which one signal source is not included in different outputs. By setting the number of the sensors to be more than a half of the sources, at least one output includes a single source. This output is selected based on property of a single source. Furthermore, this output is fed back and cancelled from the observed signals. The selected single source, information about a mixing process and histogram of the observations and the separated source are used for this cancellation. In this paper, a spectrum suppression method is introduced for reducing signal distortion caused in the cancellation process based on the histogram. Furthermore, a learning algorithm in the second source separation is also modified. Compared to the conventional method we proposed, the signal to interference is well improved.

Published
2006

9. Improvement of A Learning Algorithm for Feedback Over-Complete BSS

Subjects
音声信号, ヒストグラム, Over-complete, Histogram, Spectral Suppression, Speech signal, フィードバック, ブラインド信号源分離, オーバーコンプリート, Feedback, Blind source separation, スペクトルサプレッション, 学習アルゴリズム, Learning algorithm
Abstract

金沢大学理工研究域 電子情報学系, 信号源数がセンサ数より多いオーバーコンプリート形ブラインド信号源分離(BSS)で,信号源の完全分離は困難という問題に対して,我々はフィードバック形の回路構成と学習法を既に提案している.まず,1個の信号源が複数の出力に含まれないことを分離の条件として,1巡目の信号源分離を行ない,少なくとも1個の出力に単一信号源を分離する.この出力を観測信号にフィードバックして,観測信号から単一信号源をキャンセルすることにより,等価的に信号源数を低減する.単一信号源のキャンセル法として,当該出力と推定した混合過程の情報を使ってキャンセルする他,観測信号のヒストグラムを使ってキャンセルする方法を組み合わせる.本稿ではさらに,フィードバックにより生じる信号歪みの問題に対して,スペクトルサプレッション法を導入することにより,信号歪みを抑制する.2巡目の信号源分離では,観測信号に含まれる信号源が1個少ない状態で1巡目とは異なる学習法により処理を行なう.信号源として音声を用いたシミュレーションにより,提案方法の有効性を確認した. A feedback structure and its learning algorithm were proposed for over-complete blind source separation, where the number of sources is larger than that of sensors. In the first phase, the signal sources are separated so as to satisfy the condition, under which one signal source is not included in different outputs. By setting the number of the sensors to be more than a half of the sources, at least one output includes a single source. This output is selected based on property of a single source. Furthermore, this output is fed back and cancelled from the observed signals. The selected single source, information about a mixing process and histogram of the observations and the separated source are used for this cancellation. In this paper, a spectrum suppression method is introduced for reducing signal distortion caused in the cancellation process based on the histogram. Furthermore, a learning algorithm in the second source separation is also modified. Compared to the conventional method we proposed, the signal to interference is well improved.

Published
2006

10. Study on an Improvement of Echo Canceller by using Blind Source Separation

Subjects
echo canceller, independent component analysis, blind source separation
Abstract

We have proposed the adaptive filter combined with the blind source separation circuit which improve the performance of echo cancelation at presence of speech signal. In this report, we present the convergence speed and the statical performance of the separation circuit. Then, we will propose a method using the correlation between the recieving signal and the separated signal. By adjusting the coefficient of separation circuit to increase or reduce the correlation, we can expect the better performance of separation circuit than usual.

Published
2006

11. A Learning Algorithm Reducing Signal Distortion for Frequency Domain Feedforward Blind Source Separation

Subjects
信号歪, フィードフォワード, 学習, Signal distortion, ブラインド, 周波数領域, Blind source separation, Learning, 信号源分離, Frequency domain, Feedforward
Abstract

金沢大学理工研究域電子情報学系, ブラインド信号源分離(BSS)では、出力を独立にするように学習が行われるため信号歪みが発生する可能性があることが知られている. また, フィードフォワード形ブラインド信号源分離(FF-BSS)は分離回路における自由度が高く, 出力信号を互いに独立にする学習が信号歪みを生じる可能性がある. そこで信号歪みの低減のために完全分離の条件と無歪の条件から導かれた式を学習に加味するという手法がすでに提案されている. 本稿では周波数領域におけるFF-BSSにおいてその手法を適用し, 様々な信号源に対しての信号歪み抑制法の性能をシミュレーションにより解析する.

Published
2005

12. A Learning Algorithm Reducing Signal Distortion for Frequency Domain Feedforward Blind Source Separation

Subjects
信号歪, フィードフォワード, 学習, Signal distrtion, 周波数領域, Blind source separation, Learning, ブラインド信号源分離, Frequency domain, Feedforward
Abstract

金沢大学理工研究域 電子情報学系, ブラインド信号源分離(BSS)では、出力を独立にするように学習が行われるため信号歪みが発生する可能性があることが知られている.また, フィードフォワード形ブラインド信号源分離(FF-BSS)は分離回路における自由度が高く, 出力信号を互いに独立にする学習が信号歪みを生じる可能性がある.そこで信号歪みの低減のために完全分離の条件と無歪の条件から導かれた式を学習に加味するという手法がすでに提案されている.本稿では周波数領域におけるFF-BSSにおいてその手法を適用し, 様々な信号源に対しての信号歪み抑制法の性能をシミュレーションにより解析する. In the blind source separation (BSS), a separation block is trained so as to make its output signals to be statistically independent. Since the frequency domain BSS has some degree of freedom, the wave form and frequency response of the separated signals may be changed from the originals, resulting in signal distortion. In this paper, a learning algorithm, suppressing the signal distortion, for the feedforward BSS, implemented in the frequency domain, is proposed. A condition on both complete separation and signal distortion free has been proposed. This condition is included in the learning process as constraint. Usefulness of the proposed learning algorithm is analyzed by using several kinds of the signal sources.

Published
2005

13. A Study on the Independent Component Analysis on Blind Source Separation

Subjects
independent component analysis, blind source separation, Comon algorithm
Abstract

The author proposes a new method for the permutation problem on the Comon algorithm. The Comon algorithm is one of the effective Independent Component Analysis which is used for the Blind Source Separation. In this paper, permutation problem and conventional method for the problem are described. Then, a new method is proposed, and the performance of it is evaluated by using noise reduction rate and auditory test.

Published
2005

14. A Study on the Stable Control of Blind Source Separation

Subjects
blind source separation, stability of iterative algorithm
Abstract

The aouthor has proposed the adaptive filter combined with the blind source separation circuit which improve the performance of echo cancelation at presence of speech signal. In this paper, the convergence speed and the statical performance of the separation circuit are presented. Then, a new method are proposed. The method varing the coefficient of updating to get the first convergence and good separation simultaneously. The method use a cumulant of the circuit output to know the convergence of the circuit. The method are compared to the performance of the usual system, The relation between the function form of getting the coefficient and the performance of the separation circuit is studied. Finally, this paper deal with the stable control of update coefficient to get fast convergence.

Published
2004

15. A study of the stability of blind source separation algorithm

Subjects
blind source separation, stability of iterative algorithm
Abstract

This paper descirbe the stabihty analysis of the algorithm for the blind separation ciurcuit which is to be used to avoid the infiuence of near-end speech in the echo canceller system. The alogrithm was proposed by H,Jutten et al, but they have not considered the stability of it. They have proposed the two type of algorithm. The first one using the 3rd and lst order statistics is not stable in the usual case of source signal and mixing system. The seconde one use the lst and 3rd order statistics, and the algorithm is stable. The stability of the second algorithm is ensured by the computer simulation in some case.

Published
2003

16. Studies on BSS in a real environment and EEG analysis based on ICA

Subjects
independent component analysis, blind source separation, EEG analysis, signal processing
Abstract

第1章 序論||第2章 ブラインド信号分離と独立成分分析||第3章 瞬時混合過程における独立成分分析の不定性問題の解決||第4章 実環境下での音響信号の分離と目的音声の選択||第5章 単一試行視覚認識時脳波の特徴抽出と変動識別||第6章 結論||謝辞||参考文献, 九州工業大学博士学位論文 学位記番号:情工博甲第193号 学位授与年月日:平成19年3月23日, 平成18年度

18. A study on Blind Source Separation and blind Source Extraction that preserve signal quality

Subjects
Blind Source Separation, independent component analysis, blind Source Extraction, signal quality, convolutive mixture
Abstract

第1章 序論||第2章 問題設定||第3章 信号の性質を保存するブラインド信号分離||第4章 信号の性質を保存するブラインド信号抽出||第5章 音声信号を対象とした実験||第6章 結論||参考文献||謝辞, 九州工業大学博士学位論文 学位記番号:生工博甲第141号 学位授与年月日:平成22年3月24日, 平成21年度

Catalog

Books, media, physical & digital resources
See catalog results