※一部利用できない機能があります
1.
論文 
Nakayama, Kenji ; Katou, Haruo ; Hirano, Akihiro
概要:
金沢大学理工研究域 電子情報学系


2.
論文 
Nakayama, Kenji ; Kaneda, Yasuaki ; Hirano, Akihiro ; Haruta, Yasuhiro
概要:
金沢大学理工研究域 電子情報学系<br />Multilayer neural networks (MLNN) and the FFT amplitude of brain waves have been applied to
…
dasiaBrain Computer Interfacepsila (BCI). In this paper, a magnetoencephalograph (MEG) system, dasiaMEGvisionpsila developed by Yokogawa Corporation, is used to measure brain activities. MEGvision is a 160channel wholehead MEG system. Channels are selected from 8 main regions, a frontal lobe, a temporal lobe, a parietal lobe and a occipital lobe, located in the left and the right sides of the brain. The 8 channels, located at the central point in the 8 lobes, are initially selected. Optimum channels are searched for in the same lobe as the initial channels in order to achieve high classification accuracy. Two subjects and four mental tasks, including relaxed situation, multiplication, playing sport and rotating an object, are used. The brain waves are measured 10 times for one subject and one mental task. Among them, 8 data sets are used for training the MLNN, and the remaining 2 data sets are used for testing. 5 kinds of combinations of 2 data sets are selected for testing. Rates of correct classification by using the initial channels are 82:5 ~ 90%. By optimizing the channels, the accuracy is improved up to 85:0 ~97:5%, which is very high accuracy. Furthermore, contributions of the brain waves in the 8 lobes are analyzed.
続きを見る


3.
論文 
Ikeda, Kazushi ; Wang, Youhua ; Nakayama, Kenji
概要:
金沢大学理工研究域電子情報学系<br />The numerical property of the recursive least squares (RLS) algorithm has been extensively st
…
udied. However, very few investigations are reported concerning the numerical behavior of the predictorbased least squares (PLS) algorithms which provide the same least squares solutions as the RLS algorithm. In Ref. [9], we gave a comparative study on the numerical performances of the RLS and the backward PLS (BPLS) algorithms. It was shown that the numerical property of the BPLS algorithm is much superior to that of the RLS algorithm under a finiteprecision arithmetic because several main instability sources encountered in the RLS algorithm do not appear in the BPLS algorithm. This paper theoretically shows the stability of the BPLS algorithm by error propagation analysis. Since the timevariant nature of the BPLS algorithm, we prove the stability of the BPLS algorithm by using the method as shown in Ref. [6]. The expectation of the transition matrix in the BPLS algorithm is analyzed and its eigenvalues are shown to have values within the unit circle. Therefore we can say that the BPLS algorithm is numerically stable.
続きを見る


4.
論文 
Kuranishi, Yoshiaki ; Nakayama, Kenji ; Miyadera, Kazuyuki ; Okumura, Toshiyuki
概要:
金沢大学理工研究域 電子情報学系<br />A singlechip speech spectrum analyzer which contains a 20channel filter bank, a 9bitreso
…
lution analogtodigital converter, and a 396bit buffer memory is described. Several efficient design techniques were used to realize the equivalent 308thorder transfer functions on a single chip monolithic MOS circuit. A new timedivisionmultiplexed switchedcapacitor filter technique is introduced which can easily cancel DC offsets which appear in the multiplexed channel outputs. The LSI was fabricated in 3.5μm CMOS technology, with a 7.0×6.5 mm/SUP 2/ die size, a power consumption of 150 mW, with a single power supply of +5 V. Experimental results show that designed performance was realized.
続きを見る


5.
論文 
Nakayama, Kenji ; Chigawa, Y.


6.
論文 
Hara, Kazuyuki ; Nakayama, Kenji
概要:
金沢大学理工研究域 電子情報学系


7.
論文 
Rebolledo Méndez, Jovan David ; Nakayama, Kenji ; Méndez Morales, Adolfo
概要:
金沢大学理工研究域 電子情報学系<br />Visually impaired people face many obstacles to interact properly with the current screen /
…
visually oriented technologies, as monitors or in using Internet, being destined just for users with good visual ability. This paper refers to a specification of a proposal of an innovative way to safe that obstacle for visually impaired people and provide them a device that will make them understand the normal output from a computer. This method attempts to use the basic element: a wind element (wixel). It will be explained its technical specification, methodology, materials, usages, benefits, current development and future work as well.
続きを見る


8.
論文 
Nakayama, Kenji ; Katayama, Hiroshi


9.
論文 
Nakayama, Kenji ; Inagaki, Kiyoto
概要:
金沢大学理工研究域 電子情報学系<br />Brain Computer Interface (BCI) is one of hopeful interface technologies between human and ma
…
chine. However, brain waves are very weak and there exist many kinds of noises. Therefore, what kinds of features are useful, how to extract the useful features, how to suppress noises, and so on are very important. On the other hand, neural networks are very useful technology for pattern classification. Especially, multilayer neural networks trained through the error backpropagation algorithm have been widely used in a wide variety of field. In this paper, the neural network is applied to the BCI. Amplitude of the FFT of the brain waves are used for the input data. Several kinds of techniques are introduced in this paper. Segmentation along the time axis for fast response, nonlinear normalization for emphasizing important information with small magnitude, averaging samples of the brain waves for suppressing noise effects and reduction in the number of the samples for achieving a small size network, and so on are newly introduced. Simulation was carried out by using the brain waves, which are available from the web site of Colorado state university. The number of mental tasks is five. Ten data sets for each mental task are prepared. Among them, 9 data sets are used for training, and the rest one data set is used for testing. Selection of the one data set for testing is changed and accuracy of the correct classifications are averaged over the possible selections. Approximately, 80% of correct classification of the brain waves is obtained, which is higher than the conventional. © 2006 IEEE.
続きを見る


10.
論文 
Kaneda, Yasuaki ; Nakayama, Kenji ; Hirano, Akihiro
概要:
金沢大学理工研究域 電子情報学系<br />脳波のFFTと階層形ニューラルネットワークを用いるブレイン・コンピュータ・インタフェイス(BCI)に関して,以前に前処理の方法をいくつか提案し,メンタルタスクの分類性能を向上した.本稿
…
では,まず,階層形ニューラルネットワークでメンタルタスクを分類するために用いられる特徴の解析を行った.特徴は結合荷重の分布に基づいて解析した.隠れ層から出力層への結合荷重はメンタルタスクに対して独立になる傾向があった.従って,入力層から各メンタルタスクに対応する隠れユニットへの結合荷重分布がメンタルタスク毎の特徴を表している.次に,汎化能力を向上する2通りの学習法について検討を行った.一つは,ニューラルネットワークの入力データに乱数を加える方法であり,もう一つは,結合荷重を圧縮する方法する方法である.シミュレーションの結果,いずれの方法もテストデータに対する分類性能を向上することが出来たが,乱数を加える方法が有効であることが分かった. In this paper, a multilayer neural network is applied to 'Brain Computer Interface' (BCI), which is one of hopeful interface technologies between humans and machines. Amplitude of the FFT of the brain waves are used for the input data. Several techniques have been introduced for preprocessing the brain waves. They include segmentation along the time axis for fast response, nonlinear normalization to emphasize important information, averaging samples of the brain waves to suppress noise effects, reduction in the number of the samples to realize a small size network, and so on. In this paper, two kinds of generalization techniques, including adding small random noises to the input data and decaying connection weight magnitude, are applied. Their usefulness are analyzed and compared base on correct and error classifications. Simulation is carried out by using the brain waves, which are available from the web site of Colorado State University. The number of mental tasks is five. Some data sets are used for training the multilayer neural network, and the remaining data sets are used for testing. In our previous work, classification accuracy of 64%〜74% for the test data have been achieved. In this paper, by applying the generalization techniques, the accuracy can be improved up to 80%～88%.
続きを見る


11.
論文 
Nakayama, Kenji
概要:
金沢大学理工研究域電子情報学系


12.
論文 
Rebolledo Méndez, Jovan David ; Nakayama, Kenji
概要:
金沢大学理工研究域 電子情報学系<br />A Natural Language Generation (NLG) engine is proposed based on the combination of NLG and E
…
xpert Systems. The combination of these techniques paves the way to employ user defined behaviors in virtual worlds as inputs to an expert system. Adaptive Algorithms can then be used to retrieve information from the Internet to give feedback to the user via the NLG engine. The combination of these AI techniques can bring about some benefits such as believability in the interaction between AIdriven and humandriven avatars in virtual worlds. © 2008 SpringerVerlag Berlin Heidelberg.
続きを見る


13.
論文 
Nakayama, Kenji ; Hasegawa, H. ; Hernandez, C.A.
概要:
金沢大学理工研究域 電子情報学系


14.
論文 
Wang, Youhua ; Nakayama, Kenji
概要:
金沢大学理工研究域 電子情報学系<br />This paper proposes a new combined fast algorithm for transversal adaptive filters. The fast
…
transversal filter (FTF) algorithm and the normalized LMS (NLMS) are combined in the following way. In the initialization period, the FTF is used to obtain fast convergence. After converging, the algorithm is switched to the NLMS algorithm because the FTF cannot be used for a long time due to its numerical instability. Nonstationary environment, that is, time varying unknown system for instance, is classified into three categories: slow time varying, fast time varying and sudden time varying systems. The NLMS algorithm is applied to the first situation. In the latter two cases, however, the NLMS algorithm cannot provide a good performance. So, the FTF algorithm is selected. Switching between the two algorithms is automatically controlled by using the difference of the MSE sequence. If the difference exceeds a threshold, then the FTF is selected. Other wise, the NLMS is selected. Compared with the RLS algorithm, the proposed combined algorithm needs less computation, while maintaining the same performance. Furthermore, compared with the FTF algorithm, it provides numerically stable operation.
続きを見る


15.
論文 
Hirano, Akihiro ; Nakayama, Kenji
概要:
金沢大学理工研究域電子情報学系<br />This paper presents an implementation of a stereophonic acoustic echo canceller on nVIDIA GeF
…
orce graphics processor and CUDA software development environment. For ef.ciency, fast shared memory has been used as much as possilbe. A tree adder is introduced to reduce the cost for summing thread outputs up. The performance evaluation results suggest that Even a lowcost GPU's with a small number of shader processor greatly helps the echo cancellation for lowcost PCbased teleconferencing. ©2009 IEEE..
続きを見る


16.
論文 
Nakayama, Kenji ; Uriya, Susumu ; Matsuura, Takashi ; Mitama, Masataka
概要:
金沢大学理工研究域 電子情報学系


17.
論文 
Horita, Akihide ; Nakayama, Kenji ; Hirano, Akihiro ; Dejima, Yasuhiro
概要:
金沢大学理工研究域 電子情報学系<br />In Blind Source Separation (BSS), a eparation block is trained so as to make the output sign
…
als to be statistically independent. In this case, the independency is able to be increased by changing frequency response of the output signals, resulting in signal distortion. Especially, a feedforward BSS (FFBSS) has some degree of freedom in the separation block, and the signal distortion will be caused. The signal distortion is evaluated as difference between the output signal and the signal source in the measured signal. Some equations are derived from the conditions of complete separation and signal distortion free. They are used as the distortion free constraint in the conventional learning process [11]. On the other hand, a feedback BSS (FBBSS) has a solution, which can satisfy both complete separation and distortion free. In this paper, the learning algorithm with the distortion free constraint is applied to the FFBSS in time domain. Many kinds of signal sources are used in simulation in order to compare the proposed method and the conventional, in which difference between the output signals and the measured signals is included in the cost function [4]. Furthermore, the FBBSS is also evaluated.ブラインド信号源分離では(BSS) は分離回路がその出力信号が統計的に独立になるように学習される．この場合，出力信号の周波数特性が変化することにより，独立性が高まることもあるので，信号歪みが生じる可能性がある．特に，フィードフォワード形BSS(FFBSS)は分離回路における自由度が高く，信号歪みを生じる可能性がある．信号歪みの基準を観測信号に含まれる信号源と考え，完全分離の条件と信号無歪の条件から導かれた制約条件を学習に加味する信号歪み抑制学習法を提案した[11]．信号源をsi，観測信号をxi，出力信号をyi とするとき，信号を分離するとともにyi をxi におけるsi 成分に近づけることができる．これに対し，観測信号と出力信号の差を評価関数に追加する従来法では，観測信号に含まれる複数の信号源の影響で信号源分離が充分ではない．一方，フィードバック形BSS(FBBSS)では，信号源分離と信号歪み抑制の条件を同時に満たす回が存在する．本稿では信号歪み抑制学習法を時間領域で学習するFFBSS に適用し，種々の信号源を 使って従来方式[4] と比較することによりその特性を解 析する．同時に，FBBSS の有効性も検証する．
続きを見る


18.
論文 
Horita, Akihide ; Nakayama, Kenji ; Hirano, Akihiro
概要:
金沢大学理工研究域 電子情報学系


19.
論文 
Jansen, Boris ; Nakayama, Kenji
概要:
金沢大学理工研究域 電子情報学系<br />Over the years, many improvements and refinements to the backpropagation learning algorithm
…
have been reported. In this paper, a new adaptive penaltybased learning extension for the backpropagation learning algorithm and its variants is proposed. The new method initially puts pressure on artificial neural networks in order to get all outputs for all training patterns into the correct half of the output range, instead of mainly focusing on minimizing the difference between the target and actual output values. The upper bound of the penalty values is also controlled. The technique is easy to implement and computationally inexpensive. In this study, the new approach is applied to the backpropagation learning algorithm as well as the RPROP learning algorithm. The superiority of the new proposed method is demonstrated though many simulations. By applying the extension, the percentage of successful runs can be greatly increased and the average number of epochs to convergence can be well reduced on various problem instances. The behavior of the penalty values during training is also analyzed and their active role within the learning process is confirmed. Copyright © 2006 The Institute of Electronics, Information and Communication Engineers.
続きを見る


20.
論文 
Nakayama, Kenji ; Higashi, Shoya ; Hirano, Akihiro
概要:
金沢大学理工研究域 電子情報学系<br />A noise spectral estimation method, which is used in spectral suppression noise cancellers,
…
is proposed for highly nonstationary noise environments. Speech and nonspeech frames are detected by using the entropybased voice activity detector (VAD). An adaptive normalization parameter and a variable threshold are newly introduced for the VAD. They are very useful for rapid change in the noise spectrum and power. Furthermore, a recursive averaging method is applied to estimating the noise spectrum in the nonspeech frames. In this method, an adaptive smoothing parameter is proposed, based on speech presence probability. Simulations are carried out by using many kinds of noises, including white, babble, car, pink, factory and tank, which are changed from one to the other. The segmental SNR is improved by 2.0 ~3.8dB, and noise spectral estimation error is improved by 3.2 ~ 4.7dB for the white noise and the babble noise, which are changed from one to the other.
続きを見る


21.
論文 
Hara, Kazuyuki ; Nakayama, Kenji
概要:
金沢大学理工研究域 電子情報学系<br />Signal classification performance using multilayer neural network (MLNN) and the conventiona
…
l signal processing methods are theoretically compared under the limited observation period and computational load. The signals with N samples are classified based on frequency components. The comparison is carried out based on degree of freedom the signal detection regions in an Ndimensional signal space. As a result, the MLNN has higher degree of freedom, and can provide more flexible performance for classifying the signals than the conventional methods. This analysis is further investigated throught computer simulations. Multifrequency signals and the real application, a dial tone receiver, are taken into account. As a result, the MLNN can provide much higher accuracy than the conventional signal processing methods.
続きを見る


22.
論文 
Huq, Asadul ; Ma, Zhiqiang ; Nakayama, Kenji


23.
論文 
Nakayama, Kenji
概要:
理工研究域 電子情報学系<br />A method is reported for designing a cascaded Nyquist filter with zero intersymbol interference,
…
particularly when the same transfer function is used in both the transmitter and receiver. The proposed method is applicable to both IIR and FIR filters.
続きを見る


24.
論文 
Nakayama, Kenji ; Ikehara, K.


25.
論文 
25. An errorcorrecting learning algorithm using double hysteresis thresholds for associative memory
Nakayama, Kenji ; Nishimura, Katsuaki


26.
論文 
Nakayama, Kenji ; Suzuki, H. ; Hirano, Akihiro
概要:
金沢大学理工研究域 電子情報学系<br />In this paper, new approaches to noise spectrum estimation and spectral gain control are pro
…
posed for noise spectral suppressors. First, the speech absent frames are detected by using spectral entropy. In the speech absent frames, a weighting factor used in estimating the noise spectrum is modified so as to emphasize effect of the noisy speech signal. Next, a spectral gain is more reduced by multiplying a factor in order to suppress effects of the noise in the speech absent frames. Furthermore, in the speech present frames, in order to reduce signal distortion, the spectral gain is controlled to be unity based on an SNR calculated by using a ridgeline spectrum. Finally, the original noisy speech is added to the estimated speech in some ratio. This ratio is controlled by the long term averaged SNR of the estimated noise and the noisy speech. Computer simulations by using speech signals, the white noise, the car noise and the bubble noise, which are available in public, have been carried out for the conventional methods and the proposed method. The proposed method can improve a segmental SNR and speech quality compared to the conventional methods. Especially, it is useful for the bubble noise. ©2007 IEEE.
続きを見る


27.
論文 
Jansen, Boris ; Nakayama, Kenji
概要:
金沢大学理工研究域 電子情報学系<br />階層型ニュートラルネットワークの学習法としてよく用いられるバックプロパゲーション(BP)アルゴリズムに対して，学習の収束性を改善する多くの方法が提案されている. 本稿では，BPアルゴリ
…
ズムやBPROP法などの類似する学習法を対象として，新しい適応形ペナルティに基づく学習法を提案する.学習法で用いられる出力誤差をペナルティにより増減する.ペナルティは出力が目標値と同じ放物線上にあれば小さく，そうでなければ大きく制御される. これにより，局所解に陥ることを防ぐことが出来，最適解への収束性を高めることが出来る. 多くの例を用いてシュミュレーションを行った結果，提案方法の有効性が確認できた.
続きを見る


28.
論文 
Ma, Zhiqiang ; Nakayama, Kenji
概要:
金沢大学理工研究域電子情報学系<br />A new tap assignment method for subband adaptive filters is proposed. At the beginning of ad
…
aptation, the tap assignment is controlled by the meansquared error(MSE) of each band. After the residual errors of all the bands become the same, the tap assignment is controlled by both the MSE and the tail tap weights of each subband adaptive filter. The new method in contrast to the method which uses only the MSE, allows taps to be correctly assigned in the event that the subband adaptive filters have substantially different convergence rates. The efficiency of the proposed method has been confirmed through computer simulation. The performance of the proposed method in case of a colored input signal is investigated. It is confirmed that, in the colored input signal case, the total residual error can be minimized even though the unknown system is not exactly identified. Therefore, the proposed method is suitable for applications, such as, echo cancellation and noise cancellation.
続きを見る


29.
論文 
Nakayama, Kenji ; Hirano, Akihiro ; Kashimoto, Hiroaki
概要:
金沢大学理工研究域 電子情報学系<br />適応Volterraフィルタ(AVF)は一般的な非線形特性を表現できるが, 収束が遅いという問題がある.これに対して, 入力信号を白色化する方法が提案されている.その中で, ラチス形予
…
測器を用いる方法が収束性の点で優れている.しかし, 反射係数とフィルタ係数の更新における非同期の問題がある.線形AVFではこの問題を解決する同期形学習法が提案され, 収束特性を大幅に改善している.本稿では, ラチス形AVFにおける非線形部に対する同期学習法を新たに提案する.シミュレーションにより, 非定常有色信号に対して有効であることを確認した. The adaptive Volterra filters (AVF) can represent general nonlinearity. However, its convergence speed is very slow. In order to solve this problem, techniques of whitening the input signal have been proposed. Among them, the lattice prediction error filter is useful. The lattice predictor based AVF (LatticeAVF) has some problem. Updating the reflection coefficients and the filter coefficients is asynchronous, causing poor convergence properties. In order to solve this problem, the synchronized learning algorithm has been proposed for the linear adaptive filters. In this paper, a new synchronized learning algorithm is proposed for the nonlinear part of the LatticeAVF. Its usefulness is confirmed though simulation using nonstationary colored signals.
続きを見る


30.
論文 
Nakayama, Kenji ; Sato, Yayoi ; Kuranishi, Yoshiaki
概要:
金沢大学理工研究域 電子情報学系<br />This paper describes design techniques for a switched capacitor adaptive line equalizer whic
…
h is applied to high speed (200 kb/s) digital transmission over analog subscriber loops. An equalizer transfer function is approximated so as to minimize intersymbol interference of an isolated pulse response. Optimum polezero location, which is suited to line characteristics in a wide frequency band, is also discussed. In order to attain high accuracy capacitor ratios using a small unit capacitor, capacitor values are rounded off into equivalent integer values, and are discretely optimized using polezero deviation as an error criterion. The equalizer has a finite number of frequency responses which correspond to line lengths. Gain and delay time differences between the adjoining step responses are compressed. The designed switched capacitor line equalizer was fabricated using 3mum CMOS technology. Measured data were very close to designed performances.
続きを見る


31.
論文 
Kuraishi, Yoshiaki ; Makabe, Takayoshi ; Nakayama, Kenji
概要:
金沢大学理工研究域 電子情報学系<br />This paper presents a fully integrated analog frontend LSI chip which is an interface syste
…
m between digital signal processors and existing analog telecommunication networks. The developed analog LSI chip includes many high level function blocks such as A/D and D/A converters with 11 bit resolution, various kinds of SCF's, an agc circuit, an external control level adjuster, a carrier detector, and a zero crossing detector. Design techniques employed are mainly directed toward circuit size reductions. The LSI chip is fabricated in a 5 mu m line double polysilicon gate NMOS process. Chip size is 7. 14 multiplied by 6. 51 mm. The circuit operates on plus or minus 5 v power supplies. Typical power consumption is 270 mw. By using this analog frontend LSI chip and a digital signal processor, modern systems can be successfully constructed in a compact size.
続きを見る


32.
論文 
Hirano, Akihiro ; Nakayama, Kenji
概要:
金沢大学理工研究域 電子情報学系<br />This paper presents an computationally ef cient implementation of sparsetap FIR adaptive lt
…
ers with tapposition control on Intel IA32 processors with singleinstruction multipledata (SIMD) capability. In order to overcome randomorder memory access which prevents a ectorization, a blockbased processing and a reordering buffer are introduced. A dynamic register allocation and the use of memorytoregister operations help the maximization of the loopunrolling level. Up to 66percent speedup is achieved.<br />Organized by the Electrical Engineering/Electronics, Computer, Telecommunications, and Information Technology Association (ECTI) Coorganized by GCEONGIT, Hokkaido University Technical sponsored by IEEE Circuits and Systems Society In cooperation with the Institute of Electronics, Information and Communication Engineering (IEICE)
続きを見る


33.
論文 
Wang, Youhua ; Ma, Zhiqiang ; Nakayama, Kenji


34.
論文 
Horita, Akihide ; Nakayama, Kenji ; Hirano, Akihiro ; Dejima, Yasuhiro
概要:
金沢大学理工研究域 電子情報学系<br />Source separation and signal distortion in three kinds of BSSs with convolutive mixture are
…
analyzed. They include a feedforward BSS, trained in the time domain and in the frequency domain, and a feedback BSS, trained in the time domain. First, an evaluation measure of signal distortion is discussed. Second, conditions for source separation and distortion free are derived. Based on these conditions, source separation and signal distortion are analyzed. The feedforward BSS has some degree of freedom, and the output spectrum can be changed. The feedforward BSS, trained in the frequency domain, has weighting effect, which can suppress signal distortion. This weighting is, however, effective only when the source spectra are similar to each other. Since, the feedforward BSS, trained in the time domain, does not have any constraints on signal distortion free, its output signals can be easily distorted. A new learning algorithm with a distortion free constraint is proposed. On the other hand, the feedback BSS can satisfy both source separation and distortion free conditions simultaneously. Simulation results support the theoretical analysis. © 2005 IEEE.
続きを見る


35.
論文 
Nakayama, Kenji
概要:
金沢大学理工研究域 電子情報学系<br />A simultaneous frequency and timedomain approximation method for discretetime filters is
…
proposed in this paper. In the proposed method, transfer function coefficients are divided into two subsets,X_{1}andX_{2}, which are employed for optimizing a time response and a frequency response, respectively. Frequency and time responses are optimized through the iterative Chebyshev approximation method and a method of solving linear equations, respectively. At therth iteration step, the maximum frequency response error, which appeared at the(r  1)th step, is minimized, andX_{2}^{(r 1)}becomesX_{2}^{(r)})cdot X_{1}^{(r)}is obtained from linear equations includingX_{2}^{(r)}as a constant. The frequency response at the rth step is evaluated using the above obtainedX_{1}^{(r)}andX_{2}^{(r)}. This means the optimum time response is always guaranteed in the frequencyresponse approximation procedure. A design example of a symmetrical impulse response shows the new approach is more efficient than conventional methods from the filter order reduction viewpoint.
続きを見る


36.
論文 
Nakayama, Kenji ; Horita, Hiroki ; Hirano, Akihiro
概要:
金沢大学理工研究域 電子情報学系<br />FFT and Multilayer neural networks (MLNN) have been applied to 'Brain Computer Interfac
…
e' (BCI). In this paper, in order to extract features of mental tasks, individual feature of brain waves of each channel is emphasized. Since the brain wave in some interval can be regarded as a vector, GramSchmidt orthogonalization is applied for this purpose. There exists degree of freedom in the channel order to be orthogonalized. Effect of the channel order on classification accuracy is investigated. Next, two channel orders are used for generating the MLNN input data. Two kinds of methods using a single NN and double NNs are examined. Furthermore, a generalization method, adding small random numbers to the MLNN input data, is applied. Simulations are carried out by using the brain waves, available from the Colorado State University website. By using the orthogonal components, a correct classification rate P c can be improved from 70% to 78%, an incorrect classification rate P e can be suppressed from 10% to 8%. As a result, a rate R c ∈=∈P c /(P c ∈+∈P e ) can be improved from 0.875 to 0.907. When two different channel orders are used, P e can be drastically suppressed from 10% to 2%, and R c can be improved up to 0.973. The generalization method is useful especially for using a sigle channel order. P c can be increased up to 84~88% and P e can be suppressed down to 2~4%, resulting in R c ∈=∈0.957~0.977. © 2008 SpringerVerlag Berlin Heidelberg.
続きを見る


37.
論文 
Hirano, Akihiro ; Nakayama, Kenji
概要:
This paper presents efficient implementa tion of RLSbased adaptive filters with a large number of taps on nVIDIA GeForce graphics processing unit (GPU) and CUDA software development environment. Modification of the order and the
…
combination of calcu lations reduces the number of accesses to slow offchip memory. Assigning tasks into multiple threads also takes memory access order into account. For a 4096tap case, a GPU program is almost three times faster than a CPU program.
続きを見る


38.
論文 
38. Efficient Implementation of RLSBased Adaptive Filterson nVIDIA GeForce Graphics Processing Unit
Hirano, Akihiro ; Nakayama, Kenji
概要:
This paper presents efficient implementa tion of RLSbased adaptive filters with a large number of taps on nVIDIA GeForce graphics processing unit (GPU) and CUDA software development environment. Modification of the order and the
…
combination of calcu lations reduces the number of accesses to slow offchip memory. Assigning tasks into multiple threads also takes memory access order into account. Multiple shader pro cessor arrays are used to handle a large matrix. For a 8192tap case, a GPU program is almost 30times faster than a CPU program. Realtime processing is possible for an 8kHzsampling and 512tap case by us ing 32 shader processors, which is only 25% of GeForce 8800GTS.
続きを見る


39.
論文 
Horita, Akihide ; Nakayama, Kenji ; Hirano, Akihiro ; Dejima, Yasuhiro
概要:
金沢大学理工研究域 電子情報学系<br />Source separation and signal distortion are theoretically analyzed for the FFBSS systems im
…
plemented in both the time and frequency domains and the FBBSS system. The FFBSS systems have some degree of freedom, and cause some signal distortion. The FBBSS has a unique solution for complete separation and distortion free. Next, the condition for complete separation and distortion free is derived for the FFBSS systems. This condition is applied to the learning algorithms. Computer simulations by using speech signals and stationary colored signals are carried out for the conventional methods and the new learning algorithms employing the proposed distortion free constraint. The proposed method can drastically suppress signal distortion, while maintaining high separation performance. The FBBSS system also demonstrates good performances. The FFBSS systems and the FBBSS system are compared based on the transmission time difference in the mixing process. Location of the signal sources and the sensors are rather limited in the FBBSS system. © 2006 IEEE.
続きを見る


40.
論文 
Nakayama, Kenji
概要:
金沢大学理工研究域 電子情報学系


41.
論文 
Kuranishi, Yoshiaki ; Nakayama, Kenji ; Miyadera, Kazuyuki
概要:
金沢大学理工研究域 電子情報学系


42.
論文 
Atse, Yapi ; Nakayama, Kenji ; Ma, Zhiqiang
概要:
金沢大学理工研究域 電子情報学系<br />Singlereference and multireference noise canceller (SRNC and MRNC) performances are invest
…
igated based on correlation between signal and noise. Exact relations between these noise canceller performances and signalnoise correlation have not been well discussed yet. In this paper, the above relations are investigated based on theoretical, analysis and computer simulation. The normalized LMS (NLMS) algorithm is employed. Uncorrelated, partially correlated, and correlated signal and noise combinations are taken into account. Computer simulation is carried out, using real speech, white noise, real noise sound, sine wave signals, and their combinations. In the SRNC problem, spectral analysis is applied to derive the canceller output power spectrum. From the simulation results, it is proven that the SRNC performance is inversely proportional to the signalnoise correlation as expected by the theoretical analysis. From the simulation results, the MRNC performance is more sensitive to the signalnoise correlation than that of SRNC. When the signalnoise correlation is high, by using a larger number of adaptive filter taps, the noise is reduced more, and, the signal distortion is increased. This means the signal components included in the noise are canceled exactly.
続きを見る


43.
論文 
Rebolledo Méndez, Jovan David ; Nakayama, Kenji
概要:
金沢大学理工研究域電子情報学系<br />A Natural Language Generation (NLG) engine is proposed based on the combination of NLG and Ex
…
pert Systems. The combination of these techniques paves the way to employ user defined behaviors in virtual worlds as inputs to an expert system. Adaptive Algorithms can then be used to retrieve information from the Internet to give feedback to the user via the NLG engine. The combination of these AI techniques can bring about some benefits such as believability in the interaction between AIdriven and humandriven avatars in virtual worlds. © 2008 SpringerVerlag Berlin Heidelberg.
続きを見る


44.
論文 
Nakayama, Kenji ; Tonomura, Masanobu
概要:
金沢大学理工研究域 電子情報学系


45.
論文 
Xu, Qun ; Nakayama, Kenji


46.
論文 
Nakayama, Kenji ; Inagaki, Kiyoto
概要:
Brain Computer Interface (BCI) is one of hopeful interface technologies between human and machine. However, brain waves are very weak and there exist many kinds of noises. Therefore, what kinds of features are useful, how to extract the
…
useful features, how to suppress noises, and so on are very important. On the other hand, neural networks are very useful technology for pattern classification. Especially, multilayer neural networks trained through the error backpropagation algorithm have been widely used in a wide variety of field. In this paper, the neural network is applied to the BCI. Amplitude of the FFT of the brain waves are used for the input data. Several kinds of techniques are introduced in this paper. Segmentation along the time axis for fast response, nonlinear normalization for emphasizing important information with small magnitude, averaging samples of the brain waves for suppressing noise effects and reduction in the number of the samples for achieving a small size network, and so on are newly introduced. Simulation was carried out by using the brain waves, which are available from the web site of Colorado state university. The number of mental tasks is five. Ten data sets for each mental task are prepared. Among them, 9 data sets are used for training, and the rest one data set is used for testing. Selection of the one data set for testing is changed and accuracy of the correct classifications are averaged over the possible selections. Approximately, 80% of correct classification of the brain waves is obtained, which is higher than the conventional. © 2006 IEEE.
続きを見る


47.
論文 
Ritthipravat, P. ; Nakayama, Kenji


48.
論文 
Nakayama, Kenji
概要:
金沢大学理工研究域 電子情報学系<br />A new approach to the implmentation of sum of products (SP) and transposed sum of products (
…
TSP) using partitioned arithmetic is proposed. This aproach makes it possible to realize SP and TSP with recent advanced semiconductor memory technology without multipliers. It is also shown to offer signficant reductions in memory capacity requirement and computational complexity. Dynamic range constraints and output roundoff noise analysis are also discussed.
続きを見る


49.
論文 
Hara, Kazuyuki ; Nakayama, Kenji


50.
論文 
Nakayama, Kenji ; Horita, Akihide ; Hirano, Akihiro
概要:
金沢大学理工研究域 電子情報学系<br />FeedForward (FF) and FeedBack (FB) structures have been proposed for Blind Source Separa
…
tion (BSS). The FFBSS systems have some degrees of freedom in the solution space, and signal distortion is likely to occur in convolutive mixtures. On the other hand, the FBBSS structure does not cause signal distortion. However, it requires a condition on the propagation delays in the mixing process. In this paper, source separation performance in the FBBSS is theoretically analyzed taking the propagation delays into account. Simulation is carried out by using white signals and speech signals as the signal sources. The FFBSS system and the FBBSS system are compared. Even though the FBBSS can provide good separation performance, there exits some limitation on location of the signal sources and the sensors. © SpringerVerlag Berlin Heidelberg 2007.
続きを見る


51.
論文 
Hara, Kazuyuki ; Nakayama, Kenji


52.
論文 
Ma, Zhiqiang ; Shen, Jiantao ; Huq, Asadul ; Nakayama, Kenji
概要:
金沢大学理工研究域 電子情報学系


53.
論文 
Horita, Akihide ; Nakayama, Kenji ; Hirano, Akihiro ; Dejima, Yasuhiro
概要:
金沢大学理工研究域 電子情報学系<br />Source separation and signal distortion are theoretically analyzed for the FFBSS systems im
…
plemented in both the time and frequency domains and the FBBSS system. The FFBSS systems have some degree of freedom, and cause some signal distortion. The FBBSS has a unique solution for complete separation and distortion free. Next, the condition for complete separation and distortion free is derived for the FFBSS systems. This condition is applied to the learning algorithms. Computer simulations by using speech signals and stationary colored signals are carried out for the conventional methods and the new learning algorithms employing the proposed distortion free constraint. The proposed method can drastically suppress signal distortion, while maintaining high separation performance. The FBBSS system also demonstrates good performances. The FFBSS systems and the FBBSS system are compared based on the transmission time difference in the mixing process. Location of the signal sources and the sensors are rather limited in the FBBSS system. © 2006 IEEE.
続きを見る


54.
論文 
Hirano, Akihiro ; Nakayama, Kenji ; Arai, Shoji


55.
論文 
Nakayama, Kenji ; Higashi, Shoya ; Hirano, Akihiro
概要:
金沢大学理工研究域 電子情報学系<br />A noise spectral estimation method, which is used in spectral suppression noise cancellers,
…
is proposed for highly nonstationary noise environments. Speech and nonspeech frames are detected by using the entropybased voice activity detector (VAD). An adaptive normalization parameter and a variable threshold are newly introduced for the VAD. They are very useful for rapid change in the noise spectrum and power. Furthermore, a recursive averaging method is applied to estimating the noise spectrum in the nonspeech frames. In this method, an adaptive smoothing parameter is proposed, based on speech presence probability. Simulations are carried out by using many kinds of noises, including white, babble, car, pink, factory and tank, which are changed from one to the other. The segmental SNR is improved by 2.0 ~3.8dB, and noise spectral estimation error is improved by 3.2 ~ 4.7dB for the white noise and the babble noise, which are changed from one to the other.
続きを見る


56.
論文 
Nakayama, Kenji
概要:
金沢大学理工研究域 電子情報学系


57.
論文 
Hernandez, C.A. ; Espi, J. ; Nakayama, Kenji ; Fernandez, M.
概要:
金沢大学理工研究域 電子情報学系


58.
論文 
Hernandez, C.A. ; Espi, J. ; Nakayama, Kenji


59.
論文 
Huq, Asadual ; Ma, Zhiqiang ; Nakayama, Kenji
概要:
金沢大学理工研究域 電子情報学系<br />For system identification problems, such as noise and echo cancellation, FIR adaptive filter
…
s are mainly used for their simple adaptation and numerical stability. When the unknown system is a highQ resonant system, having a very long impulse response, IIR adaptive filters are more efficient for reduction in the order of a transfer function. One way to realize the IIR adaptive filter is a separate form, in which the numerator and the denominator are separately realized and adjusted. In the actual applications, the order of the unknown system is not known. In this case, it is very important to estimate the total order and the order assignment on the numerator and the denominator. In this paper, effects of the order estimation error on the residual error are investigated. In this form, indirect error evaluation called `equation error' is used. Through theoretical and numerical investigation, the following results are obtained. First, under estimation of the order of the denominator causes large degradation. Second, over estimation can improve the performance. However, this improvement is saturated to some extent due to cancellation of the redundant poles and zeros. Third, the system identification error is proportional to the equation error as the adaptive filter approaching the optimum. Finally, there is possibility of recovering from the unstable state as the order assignment approaches to the optimum in an adaptive process using the equation error. Computer solutions are provided to aid in gaining insight of the order assignment and stability problem.
続きを見る


60.
論文 
Nakayama, Kenji ; Horita, Akihide ; Hirano, Akihiro
概要:
金沢大学理工研究域 電子情報学系


61.
論文 
Ma, Zhiqiang ; Nakayama, Kenji


62.
論文 
Nakayama, Kenji ; Inomata, Satoru ; Takeuchi, Yokou


63.
論文 
Jansen, Boris ; Nakayama, Kenji
概要:
Over the years, many improvements and refinements of the backpropagation learning algorithm have been reported. In this paper, a new adaptive penaltybased learning extension for the backpropagation learning algorithm and
…
its variants is proposed. The new method initially puts pressure on artificial neural networks in order to get all outputs for all training patterns into the correct half of the output range, instead of mainly focusing on minimizing the difference between the target and actual output values. The technique is easy to implement and computationally inexpensive. In this study, the new approach has been applied to the backpropagation learning algorithm as well as the RPROP learning algorithm and simulations have been performed. The superiority of the new proposed method is demonstrated. By applying the extension, the number of successful runs can be greatly increased and the average number of epochs to convergence can be well reduced on various problem instances. Furthermore, the change of the penalty values during training has been studied and its observation shows the active role the penalties play within the learning process. © 2006 IEEE.
続きを見る


64.
論文 
Wang, Youhua ; Nakayama, Kenji
概要:
金沢大学理工研究域 電子情報学系<br />In this letter, we introduce a predictor based least square (PLS) algorithm. By involving bo
…
th order and timeupdate recursions, the PLS algorithm is found to have a more stable performance compared with the stable version (Version II) of the RLS algorithm shown in Ref. [1]. Nevertheless, the computational requirement is about 50% of that of the RLS algorithm. As an application, the PLS algorithm can be applied to the fast newton transversal filters (FNTF) [2]. The FNTF algorithms suffer from the numerical instability problem if the quantities used for extending the gain vector are computed by using the fast RLS algorithms. By combing the PLS and the FNTF algorithms, we obtain a much more stable performance and a simple algorithm formulation.
続きを見る


65.
論文 
Horita, Akihide ; Nakayama, Kenji ; Hirano, Akihiro ; Dejima, Yasuhiro
概要:
金沢大学理工研究域 電子情報学系<br />畳み込み混合過程におけるフィードフォワード(FF)形ブラインドソースセパレーション(BSS)では自由度が存在するため信号歪みが生じる.我々は,信号源センサーが2チャンネルの場合にお
…
いて,完全分離と信号無歪みの条件を制約条件として課す信号歪み抑制学習アルゴリズムを時間領域,周波数領域のFFBSSに対して提案してきた.本稿では,信号歪み抑制の制約条件を多チャンネルに拡張し,かつ,計算の複雑さを軽減するために制約条件を近似する方式を提案する.音声を用いたコンピュータシミュレーションによってその近似制約方式と厳密制約方式がほぼ同等の分離性能と信号歪み抑制が得られることを確認した.また,3チャンネルにおいても,従来方式より特性が改善されることを確認した. Feedforward Blind Source Separation (FFBSS) systems have some degree of freedom in the solution space, and signal distortion is likely to occur in convolutive mixtures. Previously, a condition for complete separation and distortion free has been derived for 2channel FFBSS. This condition has been applied to the learning algorithms as a distortion free constraint in both the time and frequency domains. In this paper, the condition is further extended to multiple channel FFBSSs. This condition requires the a high computational complexity to be applied to the learning process as a constraint. An approximate constraint is proposed in order to relax the high computational load. In comparison with the original constraint, computer simulations have demonstrated that the approximation can obtain similar performances with respect to source separation as well as signal distortion using speech signals. Furthermore, the performances can be improved compared to the conventionals for three channels.
続きを見る


66.
論文 
66. An errorcorrecting learning algorithm using double hysteresis thresholds for associative memory
Nakayama, Kenji ; Nishimura, Katsuaki
概要:
金沢大学理工研究域 電子情報学系


67.
論文 
Nakayama, Kenji
概要:
金沢大学理工研究域 電子情報学系


68.
論文 
Hernandez, C.A. ; Espi, J. ; Nakayama, Kenji ; Fernandez, M.
概要:
金沢大学理工研究域 電子情報学系


69.
論文 
Wang, Y. ; Nakayama, Kenji
概要:
The numerical property of the recursive least squares (RLS) algorithm has been extensively studied. However, very few investigations are reported concerning the numerical behavior of the predictor based least squares (PLS) algorithms
…
that provide the same least square solutions as the RLS algorithm. This paper studies the numerical property of the backward PLS (BPLS) algorithm. First, the stability of the BPLS algorithm is verified by using state space method. Then, finiteprecision arithmetic error effects on both the BPLS and the RLS algorithms are presented through computer simulations. Some important results are obtained, which demonstrate that the BPLS algorithm appears quite robust to roundoff errors and provides a much more accuracy and stable numerical performance than that of the RLS algorithm under finiteprecision implementation.
続きを見る


70.
論文 
Hirano, Akihiro ; Nakayama, Kenji ; Takebe, Kazue


71.
論文 
Ikeda, K. ; Wang, Y. ; Nakayama, Kenji


72.
論文 
Ma, Zhiqiang ; Nakayama, Kenji ; Sugiyama, K.
概要:
金沢大学理工研究域 電子情報学系


73.
論文 
Nakayama, Kenji
概要:
金沢大学大学院自然科学研究科情報システム<br />金沢大学工学部<br />A simultaneous frequency and timedomain approximation method for di
…
scretetime filters is proposed. In the method, transfer function coefficients are divided into two subsets, X//1 and X//2, which are employed for optimizing a time response and a frequency response, respectively. Frequency and time responses are optimized through the iterative Chebyshev approximation method and a method of solving linear equations, respectively. At the rth iteration step, the maximum frequency response error, which appeared at the (r minus 1)th step, is minimized, and X//2**(**r** minus **1**) becomes X//2**(**r**). X//1**(**r**) is obtained from linear equations including X//2**(**r**) as a constant. The frequency response at the rth step is evaluated using the above obtained X//1**(**r**) and X//2**(**r**). This means the optimum time response is always guaranteed in the frequencyresponse approximation procedure. A design example of a symmetrical impulse response shows the new approach is more efficient than conventional methods from the filter order reduction viewpoint.
続きを見る


74.
論文 
Hara, Kazuyuki ; Nakayama, Kenji
概要:
金沢大学大学院自然科学研究科知能情報・数理<br />A training data selection method is proposed for multilayer neural networks (MLNNs). Th
…
is method selects a small number of the training data, which guarantee both generalization and fast training of the MLNNs applied to pattern classification. The generalization will be satisfied using the data locate close to the boundary of the pattern classes. However, if these data are only used in the training, convergence is slow. This phenomenon is analyzed in this paper. Therefore, in the proposed method, the MLNN is first trained using some number of the data, which are randomly selected (Step 1). The data, for which the output error is relatively large, are selected. Furthermore, they are paired with the nearest data belong to the different class. The newly selected data are further paired with the nearest data. Finally, pairs of the data, which locate close to the boundary, can be found. Using these pairs of the data, the MLNNs are further trained (Step 2). Since, there are some variations to combine Steps 1 and 2, the proposed method can be applied to both offline and online training. The proposed method can reduce the number of the training data, at the same time, can hasten the training. Usefulness is confirmed through computer simulation.
続きを見る


75.
論文 
Nakayama, Kenji ; Kato, Takuo ; Katayama, Hiroshi


76.
論文 
Nakayama, Kenji
概要:
A bandlimited signal extrapolation algorithm is proposed. J. A. Cadzow's algorithm (1979) is modified to eliminate undesired outband spectra. An inverse filter stopband response is relaxed by adding small random numbers. A
…
constrained heuristic optimization is suggested that can use arbitrary signal properties as constraints. Through numerical examples, it is shown how regularization techniques can improve SNR (signaltonoiseratio) by 20 dB.
続きを見る


77.
論文 
Keeni, Kanad ; Shimodaira, Hiroshi ; Nakayama, Kenji
概要:
This paper presents an automatic coding scheme for representing the output layer of a neural network. Compared to local representation where the number of output unit is p, the number of output unit required for the proposed representation is
…
close to log p. The output of seven different printers were used for evaluating the performance of the system. The proposed automatic representation gave the average recognition rate of 98.7% for 71 categories.
続きを見る


78.
論文 
Wang, Youhua ; Ikeda, Kazushi ; Nakayama, Kenji


79.
論文 
Nakayama, Kenji ; Katayama, Hiroshi


80.
論文 
Ikeda, Kazushi ; Suzuki, Akihiro ; Nakayama, Kenji
概要:
The effects of the quantization of the parameters of a learning machine are discussed. The learning coefficient should be as small as possible for a better estimate of parameters. On the other hand, when the parameters are quantized, it
…
should be relatively larger in order to avoid the paralysis of learning originated from the quantization. How to choose the learning coefficient is given in this paper from the statistical point of view.
続きを見る


81.
論文 
Horita, Akihide ; Nakayama, Kenji ; Hirano, Akihiro
概要:
金沢大学大学院自然科学研究科情報システム<br />FeedForward (FF) and FeedBack (FB) structures have been proposed for Blind Source Se
…
paration (BSS). The FFBSS systems have some degrees of freedom in the solution space, and signal distortion is likely to occur in convolutive mixtures. On the other hand, the FBBSS structure does not cause signal distortion. However, it requires a condition on the propagation delays in the mixing process. In this paper, source separation performance in the FBBSS is theoretically analyzed taking the propagation delays into account. Simulation is carried out by using white signals and speech signals as the signal sources. The FFBSS system and the FBBSS system are compared. Even though the FBBSS can provide good separation performance, there exits some limitation on location of the signal sources and the sensors. © SpringerVerlag Berlin Heidelberg 2007.
続きを見る


82.
論文 
Hirano, Akihiro ; Nakayama, Kenji ; Arai, Shinya ; Deguchi, Masayuki
概要:
金沢大学大学院自然科学研究科知能情報・数理


83.
論文 
Nakayama, Kenji ; Imai, Kunihiko
概要:
金沢大学理工研究域電子情報学系<br />A neural demodulator is proposed for amplitude shift keying (ASK) signal. It has several impo
…
rtant features compared with conventional linear methods. First, necessary functions for ASK demodulation, including wideband noise rejection, pule waveform shaping, and decoding, can be embodied in a single neural network. This means these functions are not separately designed but unified in a learning and organizing process. Second, these functions can be selforganized through the learning. Supervised learning algorithms, such as the backpropagation algorithm, can be applied for this purpose. Finally, both wideband noise rejection and a very sharp waveform response can be simultaneously achieved. It is very difficult to be done by linear filtering. Computer simulation demonstrates efficiency of the proposed method.
続きを見る


84.
論文 
Miyoshi, Seiji ; Nakayama, Kenji
概要:
In this paper, the geometric learning algorithm (GLA) is proposed for an elementary perceptron which includes a single output neuron. The GLA is a modified version of the affine projection algorithm (APA) for adaptive filters.
…
The weights update vector is determined geometrically towards the intersection of the k hyperplanes which are perpendicular to patterns to be classified. k is the order of the GLA. In the case of the APA, the target of the coefficients update is a single point which corresponds to the best identification of the unknown system. On the other hand, in the case of the GLA, the target of the weights update is an area, in which all the given patterns are classified correctly. Thus, their convergence conditions are different. In this paper, the convergence condition of the 1st order GLA for 2 patterns is theoretically derived. The new concept `the angle of the solution area' is introduced. The computer simulation results support that this new concept is a good estimation of the convergence properties.
続きを見る


85.
論文 
Ma, Zhiqiang ; Nakayama, Kenji ; Yamamoto, G.


86.
論文 
Nakayama, Kenji ; Kuriki, Satoshi
概要:
A simplified digital tunedcircuit extractor is proposed. An A/D converter and multipliers are not required. An A/D converter and multipliers are not required. The tuning frequency is determined only by a master clock. The tank circuit
…
output noise is sufficiently suppressed by newly introduced error canceller. Although sampling frequency for the tank circuit is relatively low, the phase of the output signal is effectively detected. A computer simulation shows that phase adjusting for data sampling clock is stable. The proposed timing extractor is easily realized on digital integrated circuits.
続きを見る


87.
論文 
Nakayama, Kenji
概要:
A new difference coefficient FIR digital filter is proposed. Its coefficient is obtained as the difference between the successive values of the original coefficients permuted in large magnitude sequence. It is effectively applied to
…
every kind of filter responses. Computation complexity becomes about 18% and 13% for 99th and 299th order FIR filters, compared with direct FIR filter realization.
続きを見る


88.
論文 
Miyoshi, Seiji ; Nakayama, Kenji
概要:
金沢大学理工研究域電子情報学系<br />In this paper, probabilistic memory capacity of recurrent neural networks(RNNs) is investigat
…
ed. This probabilistic capacity is determined uniquely if the network architecture and the number of patterns to be memorized are fixed. It is independent from a learning method and the network dynamics. It provides the upper bound of the memory capacity by any learning algorithms in memorizing random patterns. It is assumed that the network consists of N units, which take two states. Thus, the total number of patterns is the Nth power of 2. The probabilities are obtained by discriminations whether the connection weights, which can store random M patterns at equilibrium states, exist or not. A theoretical way for this purpose is derived, and actual calculation is executed by the Monte Carlo method. The probabilistic memory capacity is very important in applying the RNNs to real fields, and in evaluating goodness of learning algorithms. As an example of a learning algorithm, the improved error correction learning is investigated, and its convergence probabilities are compared with the upper bound. A linear programming method can be effectively applied to this numerical analysis.
続きを見る


89.
論文 
Jansen, Boris ; Nakayama, Kenji
概要:
Nowadays, the integer primefactorization problem finds its application often in modern cryptography. Artificial Neural Networks (ANNs) have been applied to the integer primefactorization problem. A composed number N is applied to the ANNs,
…
and one of its prime factors p is obtained as the output. Previously, neural networks dealing with the input and output data in a decimal format have been proposed. However, accuracy is not sufficient. In this paper, a neural network following a binary approach is proposed. The input N as well as the desired output p were expressed in a binary form. The proposed neural network is expected to be more stable, i.e. less sensitive to small errors in the network outputs. Simulations have been performed and the results are compared with the results reported in the previous study. The number of required search times for the true prime number can be well reduced. Furthermore, the probability density function of the training patterns is investigated and the need for different data creation and/or selection techniques is shown. © 2005 IEEE.
続きを見る


90.
論文 
Nakayama, Kenji ; Ikehara, Keisuke


91.
論文 
Nakayama, Kenji ; Kimura, Yoshinori ; Katayama, Hiroshi
概要:
In this paper, quantization level increase in human face images using a multilayer neural network (NN) is investigated.
…
Basically speaking, it is impossible to increase quality without any other information. However, when images are limited to some category, image restoration could be possible, based on the common properties in this category. The multilayer NN is trained using human face images of 32ﾃ・2 pixels with 8levels as the input data, and 256level images as the targets. The standard backpropagation (BP) algorithm is employed. 20, 40 and 100 training data are examined. By increasing the training data, a general function of regenerating missing information can be achieved. The internal structure of the trained NN is analyzed using some special input images. As a result, it has been confirmed that the NN regards the input image as the human face, and extracts features of the face. The input image is transformed using these features and the common properties of the training data, extracted and held on the connection weights, to the human face image.
続きを見る


92.
論文 
Nakayama, Kenji ; Hirano, Akihiro ; Katoh, Shinya ; Yamamoto, Tadashi ; Nakanishi, Kenichi ; Sawada, Manabu
概要:
In training neural networks, it is important to reduce input variables for saving memory, reducing network size, and ach
…
ieving fast training. This paper proposes two kinds of selecting methods for useful input variables. One of them is to use information of connection weights after training. If a sum of absolute value of the connection weights related to the input node is large, then this input variable is selected. In some case, only positive connection weights are taken into account. The other method is based on correlation coefficients among the input variables. If a time series of the input variable can be obtained by amplifying and shifting that of another input variable, then the former can be absorbed in the latter. These analysis methods are applied to predicting cutting error caused by thermal expansion and compression in machine tools. The input variables are reduced from 32 points to 16 points, while maintaining good prediction within 6 ﾎｼm, which can be applicable to real machine tools.
続きを見る


93.
論文 
Kuraishi, Yoshiaki ; Makabe, Takayoshi ; Nakayama, Kenji
概要:
金沢大学大学院自然科学研究科情報システム<br />金沢大学工学部<br />This paper presents a fully integrated analog frontend LSI chip whic
…
h is an interface system between digital signal processors and existing analog telecommunication networks. The developed analog LSI chip includes many high level function blocks such as A/D and D/A converters with 11 bit resolution, various kinds of SCF's, an agc circuit, an external control level adjuster, a carrier detector, and a zero crossing detector. Design techniques employed are mainly directed toward circuit size reductions. The LSI chip is fabricated in a 5 mu m line double polysilicon gate NMOS process. Chip size is 7. 14 multiplied by 6. 51 mm. The circuit operates on plus or minus 5 v power supplies. Typical power consumption is 270 mw. By using this analog frontend LSI chip and a digital signal processor, modern systems can be successfully constructed in a compact size.
続きを見る


94.
論文 
Nakayama, Kenji ; Hirano, Akihiro ; Dejima, Yasuhiro
概要:
This paper analyzes signal separation and distortion in a feedforward blind source separation (BSS) applied to convolutive mixture environment. In the BSS learning, a separation block is adjusted so as to make the output signals
…
statistically independent. One direction for this adjustment is to extract only one signal source at a single output. The other direction is to make the output signal spectra white. The second direction causes signal distortion and conflicts with the first direction. This confliction prevent signal separation. These relations are also dependent on a length of the impulse responses in the mixture and the filter length in the separation block. These properties are theoretically analyzed and are investigated through simulations using the white signal sources and the speech signal sources.
続きを見る


95.
論文 
Xu, Q. ; Nakayama, Kenji
概要:
This paper investigates some possible problems of Cascade Correlation algorithm, one of which is the zigzag output mapping caused by weightillgrowth of the adding hidden unit. Without doubt, it could lead to deteriorate the
…
generalization, especially for regression problems. To solve this problem, we combine Cascade Correlation algorithm with regularization theory. In addition, some new regularization terms are proposed in light of special cascade structure. Simulation has shown that regularization indeed smooth the zigzag output, so that the generalization is improved, especially for functional approximation.
続きを見る


96.
論文 
Nakayama, Kenji ; Chigawa, Yasuhide ; Hasegawa, Osamu


97.
論文 
97. A weighted competitive learning method extracting skeleton pattern for Japanese Kanji characters
Nakayama, Kenji ; Kato, Takuo


98.
論文 
Tokui, N. ; Nakayama, Kenji ; Hirano, Akihiro
概要:
In order to achieve fast convergence and less computation for adaptive filters, a joint method combining a whitening process and the NLMS algorithm is a hopeful approach. However, updating the filter coefficients is not synchronized with the
…
reflection coefficient updating resulting in unstable behavior. We analyzed effects of this, and proposed the Synchronized Learning Algorithm to solve this problem. Asynchronous error between them is removed, and fast convergence and small residual error were obtained. This algorithm, however, requires O(ML) computations, where M is an adaptive filter length, and L is a lattice predictor length. It is still large compared with the NLMS algorithm. In order to achieve less computation while the fast convergence is maintained, a block implementation method is proposed. The reflection coefficients are updated at some period, and are fixed during this interval. The proposed block implementation can be effectively applied to parallel form adaptive filters, such as subband adaptive filters. Simulation using speech signal shows that a learning curve of the proposed block implementation a little slower than the our original algorithm, but can save the computational complexity.
続きを見る


99.
論文 
Hirano, Akihiro ; Nakayama, Kenji ; Ushimaru, S.


100.
論文 
Yukawa, Akira ; Maruta, Rikio ; Nakayama, Kenji
