Search

Conventionally, by mixed respiratory sound and heart sound in auscultation, appropriate diagnosis often prevented. In order to solve this problem, a method of separating respiratory sound and heart sound based on blind source separation (BSS) has been proposed. However, there is the problem that sufficient separation precision cannot be obtained. In the BSS method, it is necessary to use an optimal non-linear function for the signal distribution. However, there is no non-linear function corresponding to the distribution of respiratory sound and the distribution of heart sound. This was a factor that made it impossible to obtain a sufficient separation precision. Therefore, in this study, we propose three non-linear functions appropriately corresponding to super Gaussian distribution through sub-Gaussian distribution to uniform distribution. We evaluate the effectiveness of this method by computer simulation in section 3 and confirm its effectiveness.

In the communication system, received signals include channel distortion due to channel characteristics. Generally, it is difficult to equalize when noises are added to received signals. In order to solve this problem, an online blind equalization method with a noise removal function was proposed, however the equalization performance is not sufficient. Therefore, in this paper, we propose a method that can improve the equalization performance for received signals with noises. In the proposed method, first, we adaptively estimate the channel using the Total Least Squares method(TLS method). Because it is known that the TLS method can accurately estimate the channel even in a noisy environment. Second, the result is used to remove noises. Next, we adaptively and accurately calculate equalization parameters from estimated channel characteristics. Last, it is considered that high equalization performance can be obtained by giving these parameters to the equalizer every sample time. The proposed method is evaluated by computer simulation.

Recently, the handle of driver's health condition while driving a car are actively studied. Especially, Pulse Wave (PW) is gathering attention as an indicator to grasp to one's health condition. However, extracting PW while driving without wearing, many noises such as running vibration and driver's body movement will be included. In this paper, we propose PW extraction method using BSS based on Wavelet Transform in order to improve separation precision. Because Wavelet Transform is superior to process nonstationary signals. We evaluate the advantage of proposed method with computer simulation compared to conventional method based on FFT.

Recently, in order to compensate for distortion of the received signals, the blind equalization method has been studied. The blind equalizer is designed by using only received signals. However, equalization precision is lower when noise is included in the received signals [1]. Therefore, in order to solve the problem, we propose a method to equalize accurately even in the noisy environment by using Total Least Squares (TLS). Specifically, the procedures of equalization with noise reduction function are as follows

Conventionally, there is a method to use auscultation when medical doctors diagnose a disease from heartbeats with heart noise. However, the respiratory sound and the heart sound are mixed. Further, it is difficult for the desired heart sound to be heard. For solving this problem, we propose a method to separate respiratory sound and heart sound based on blind source separation (BSS). In BSS method, we use a non-linear function corresponding to the probability distribution of source signals in order to separate. A distribution of the respiratory sound is a high-order sub-Gaussian distribution close to uniform distribution, a distribution of the heart sound is a super Gaussian distribution. In the conventional method, there is no non-linear function corresponding to both distribution. Therefore, in this paper, in order to separate the respiratory sound and the heart sound, we propose new non-linear function accurately separated from super Gaussian distribution through sub-Gaussian distribution to uniform distribution. Then, we build the BSS method based on the proposed non-linear function. We evaluate the proposed method by computer simulation.

Recently, monitor for the health condition of a driver during driving a car is actively studied [1]. Especially, pulse wave (PW) is studied to monitor the health condition of a driver during driving a car. Conventionally, the observed signals are assumed instantaneous mixture of PW and noise, and the separation precision with conventional method is lower [3]. Therefore, the observed signal may be convolutive mixtures. In this paper, in order to improve separation precision, we propose the method based on frequency domain Blind Source Separation (BSS) for convolutive mixture. We evaluate the advantage of the proposed method by computer simulation.

A Blind Source Separation (BSS) can separate the source signals from only the observation signals mixed the plural signals under the condition that the source signals and the mixing process are unknown. On the other hand, a block signal processing is often used for a reduction of a calculation or a frequency domain processing and a BSS has been adopted in a biosignal processing field as well as an audio signal processing field. In this paper, we propose a BSS method with locally mixed nonstationary noise reduction based on block processing for extracting a pulse wave(PW). The aim of this research is, using a block BSS, to remove nonstationary noise within a block that is included in a short interval. To show the performance of the proposed method for measured data, we used a pulse wave data while is sensored while driving. It was confirmed that the proposed method is effective for extraction of PW.

The auscultatory sounds are mixed heart sound, lung sound and other noises. If an auscultatory sound can be separated to a heart sound and a lung sound, it is possible that it may be useful for automatic diagnosis disease of the heart and lungs. However, a conventional method using the blind signal separation based on an ICA has lower separation precision. In the conventional method, a nonlinear function based on the separated signals is used in the separation algorithm. It is supposed to be an appropriate function in terms of separation precision corresponding to the probability distribution of source signals. However, because the conventional method has not been given an appropriate nonlinear function, it is a cause of lower separation precision. In this paper, we estimate a probability distribution of the source signals from the high order statistics of the separated signals and propose a better method which uses an appropriate nonlinear function in term of separation precision. Moreover, we evaluate a proposed algorithm to improve separation precision.

We propose the new suppression method of a click noise. A conventional suppression method of a click noise is as follows: 1) detection of a click noise 2) removal of a noise 3) interpolation using a linear interpolation or a spline interpolation. When we use the conventional method, the precision of the suppression is low. On the other hand, in fields of a digital signal processing and a system identification, there are some studies on an estimation method of missing data. Then, in this paper, we regenerate the original signal for the missing data by assuming a click noise to be missing data and applying the estimation method of missing data to a linear predictive model. The proposed method is as follows: 1) detection of a click noise 2) regarding that noise as missing data 3) replacing the missing part with the estimation signal of the original signal for the missing data, and regenerating the original signal. The third processing 3) of our method is different from the conventional method. We describe the detailed procedure of the proposed method and the principle.

Blind signal separation method is the method that the source signals are obtained by separating only the mixed signals. Generally, separation precision is lower when the basic method of blind signal separation is used for the sound signals. It is the cause that the power of the sound signals is concentrated in the low-frequency band. For the problem, we already considered that the method of the blind signal separation using QMF. Specifically, the mixed signals are divided into two frequency bands, and we perform suitable separation for each two bands. As the result, the conventional method has higher separation precision than the basic method. However, in the low-frequency band, power spectrum is not constant for frequency. Therefore, it is expected that the separation precision is higher than the conventional method by performing the suitable separation for each of signals based on the division to several narrow frequency bands on the low-frequency band. Then, in this paper, we propose the method that has higher separation precision and we evaluate the proposed method by computer simulation.

Blind signal separation method is the method that the source signals are obtained by separating only the mixed signals. Generally, separation precision is lower when blind signal separation method is used for the sound signals. It is the cause that the power of the sound signals is concentrated in the low-frequency band. Specifically, the high-frequency band has over separation when the separation to be suitable for the low-frequency band is performed, and the separation of the low-frequency band is insufficient when the separation to be suitable for the high-frequency band is performed. Then, in this paper, we propose the method that has higher separation precision. The proposed method is as follows: 1) divide the mixed signals into the signals of the high-frequency band and the signals of the low-frequency band, 2) separate each of the divided signals, 3) reconstruct the regenerated source signals from each of the separated signals. We evaluate the proposed method by computer simulation.

There is the over-sampling method which is one of the methods for realizing the blind identification. In the conventional method, the unknown system is usually assumed by FIR system and the estimation filter is also used by FIR filter. However, if a target system has long impulse length in many real environment, we may have to assume IIR system for the unknown system. If we use FIR filters to estimate such systems, the identification performance is lower. Therefore, it is necessary to research the blind identification method for the unknown system as IIR system. Then, we propose the blind identification based on over-sampling method using orthogonal filter for IIR system and we evaluate the identification performance by computer simulations.

A novel switched-capacitor algorithmic digital-to-analog converter (D/A) is described. It is formed of one sample/hold and divider. Since no integrator is used, error caused by open loop gain of operational amplifier can easily be compensated. Required components are only two operational amplifiers, three unit capacitors and nine analog switches. Accuracy estimation shows that 12 bit resolution is possible on capacitor mismatch of 0.075%.

There is the over-sampling method which is one of the methods for realizing the blind identification. In the conventional method, the unknown system is assumed by FIR type and the estimation filter is used by FIR filter. However, we have the case which can not assume the unknown system to be FIR type. At the result, estimation performance of the conventional method is lower. Then, in this paper, as the first step, we assume AR type which is the simplest IIR type as the unknown system. And, we propose the blind identification method for AR model which is used as unknown system using oversampling method and we evaluate the identification performance by computer simulation.

The blind equalization is a method which estimates the inverse system of the channel by using the only received signals, and the transmitted signals are regenerated by using the inverse system and the received signals. In the conventional method, it is assumed that the input signals of the equalization system don't include noise. In the actual blind equalization, however, the equalization performance of the conventional methods are lower because of the equalization system's input signals include noise. To solve the problem, in the prior study, several persons in the members of our study propose the blind identification algorithm using the total least squares (TLS) based on over-sampling method to use blind equalization.Then, in this paper, we considerate the blind equalization method using the blind identification algorithm based on prior study with higher estimation performance when received signals include noise. As the result, the proposed method is higher equalization performance than the conventional method in the received signals including noise, by computer simulation.

In this report, we consider the blind identification using n-times over-sampling, where the noises are included to the received signals of the channel. In communication system, one of methods to realize identification by using only received signals is the method based on cyclostationary process, where its process is obtained by n-times over-sampling of the received signals. There is an algorithm using subspace method as one of the representative methods based on n-times over sampling methods. In the algorithm using subspace method, the received signals space are separated into signals subspace and noises subspace, and the unknown system is offline estimated by using these orthogonality, In this method, practical use is difficult because this is no suitable for online processing. It is because its many computational complexity. In order to solve the problem, we propose the adaptive method based on TLS. The proposed method is a few computational complexity because it is adaptive method. Therefore, we are able to realize online processing of the blind identification. As the result for performance confirmation of the proposed method by computer simulation, it is cleared that the proposed method has higher estimation precision and faster convergence rate. It is thought that estimation precision is higher because we use TLS, And it is thought that convergence rate is faster because we use updating rule such as RLS and variable step gain. From presentation above, the proposed method is the algorithm that has practical application.

The over-sampling method is used for one of the blind channel estimation method. Applying TLS algorithm to this method, the estimation performance is high performance even if it is added noise to received signals as input signals into the estimation filter. However, if the impulse response of channel is long, we need large taps of FIR filter as the estimation filter. There by, if the taps of estimation filter are large taps, the convergence rate is slower. In addition, if the taps of estimation filter are small taps, the estimation performance is lower.

Recently, the realization of transformation method from whisper voice to voice sounds is expected. We easily think of the method using PARCOR analysis when we consider this method. This method is as follows: 1) The vocal tract characteristic is estimated by using the PARCOR analysis. 2) The voice sounds can be generated from the whisper voice by convoluting a periodic pitch with the estimated vocal tract characteristic. However, in 1), this method needs a lot of computational complexity to estimate the vocal tract characteristic. In this paper, we propose the estimation method of vocal tract characteristic with a little computational complexity using the blind estimation based on over sampling instead of PARCOR analysis method.

There is the over-sampling method which is one of the methods for realizing the blind estimation. In the conventional methods, The algorithms are proposed for the estimation system's input signals not including noise. In the actual blind estimation, however, the estimation precisions of the conventional methods are lower because of the estimation system's input signals include noise. To solve the problem, in the prior study, several persons in the members of our study propose the blind estimation algorithm using the total least squares (TLS) based on over-sampling method. In the prior study, it is used that the method which updates the filter coefficients by using the constant step gain and a simple gradient method. As the result, the estimation precision of the prior study is higher than those of the conventional methods. The prior study, however, has a problem that its convergence rate is slower. In this paper, and then, for developing the faster algorithm than the prior study, we considerate the algorithm which updates the filter coefficients by using the variable step gain and the updating method of the filter coefficients such as the recursive least squares (RLS), where it is the algorithm using TLS based on over-sampling method. As the result, it is clear that the proposed method has a faster convergence rate than the prior study, where these estimation precisions are about equal too, by computer simulation.

The over-sampling method is used for one of the blind channel estimation method. The estimation performance is high performance if it is not added noise to received signals as input signals into the estimation filter. But if it is added noise to input signals, the estimation performance is lower. Because, generally, LMS algorithm etc. are used for this method, and these algorithms are not effective for the input signals including the noise. Therefore, in this paper, we apply TLS algorithm to the blind channel estimation method based on over-sampling because TLS algorithm is effective for the input signals including the noise. The proposed algorithm works very well compared with the conventional method using LMS algorithm by computer simulation.

In this paper, we propose a method for calibrating relative position and orientation of multiple unsynchronized cameras from general moving points. If the sampling times of multiple cameras are different from each other, there is no corresponding point in the images of moving points and we cannot calibrate these cameras. In this paper we analyze geometric relationships of multiple cameras in the frequency space, and show that they enable us to calibrate multiple unsynchronized cameras accurately even if exact corresponding points do not exist in images. The proposed method does not require any interpolation of moving points in images.

In a transmission system, the conventional adaptive blind equalizers are implemented a direct estimation for the parameters of the equalizers. But both convergence rate of these algorithms and reliability of the regenerated signals degrade down.

Some of authors proposed the blind equalization algorithm for communication systems. In the algorithm, we realize higher reliability of the recovered signals and faster convergence rate of the algorithm, because the parameter included in Sato cost function is adjusted by using both outputs of the equalizer and the decision device. However, we cannot apply this method to communication systems except binary code systems. Then, in this paper, we extend the previously proposed method to the method applied for multilevel code systems. And we confirm the performance evaluation of the extended method by the computer simulation.

In a teleconference system, a cause of deterioration for sound quality is echo signals. A convectional echo canceller had to stop updating parameters during the double talk. So, the reliability of transmitted signals deteriorate during the double talk. Therefore, blind signal separation (BSS) method is the method that is suitable for echo cancellers, because the method doesn't need to stop updating parameters during double talk. However, it is difficult to directly apply, because BSS method has two faults. So we solve these faults by using system structure of the echo canceller. As the result, we can apply BSS method to echo canceller. And, it has a higher reliability of transmitted signals.

On blind equalization for long impulse response, such as indoor sound, conventional blind equalizer using the FIR filter has problem which is slower convergence rate of the algorithm. To solve this problem, we previously proposed the blind equalizer using the orthogonal filter. But, in the paper, we confirmed the validity of the proposed algorithm through only computer simulation. Then, in this paper, we analyze the validity of the orthogonal filter in detail.

Some of authors proposed the blind equalization algorithm for communication systems. The algorithm has higher reliability of the recovered signals and faster convergence rate of the algorithm. In this method, we control a parameter included Sato cost function using both outputs of equalizer and decision device. However, this method can apply only binary code system. Then, in this paper, we extend previously proposed method to the method for multilevel code system. First, we present the process of extension. Next, we confirm the performance evaluation of the extended method by the computer simulation

A convectional stereo echo canceller had to stop updating parameters during the double talk. So, it has a problem for reliability of transmitted signals. In this paper, we propose a stereo echo canceller using blind signal separation method (BSS). In our proposed system, we remove echo signals by separating echo signals and speakers signals. So, it is not to need double talk processing. As a result, it has a high reliability of transmitted signals.

In a conventional blind source separation, we assume linear channel for mixed process. But, in many cases, the channels for biomedical signals are nonlinear channel. So we cannot use the conventional method. Then, we present non-liner model using Volterra series, and we propose a new blind source separation method for non-liner model.

In this paper, we consider the blind equalization for long impulse response length, such as indoor sound. We use FIR or IIR filter for a conventional blind equalizer. But, these equalizers have each problems. A problem of the conventional blind equalizer using FIR filter is lower convergence rate, and a problem of blind equalizer using IIR filter is to have to control the stability of the system. Then, we propose a new blind equalizer algorithm using the orthogonal filter,and we get rid of these problems. Finally, we confirm the validity of the proposed algorithm through computer simulation.

A novel buffer-based switched-capacitor (SC) integrator integrable by a method of reducing the capacitance ratio is presented. By this method, a high-Q SC filter can be made by a realizable capacitance ratio in a CMOS process. The proposed integrator can also be operated over a wide frequency range because it uses a unity gain buffer (UGB). >