This paper proposes a linear microphone array parallel to the driving direction for in-car speech enhancement. In contrast to other linear microphone arrays in the car cabin reported in a literature or implemented as a commercial product, the array axis is arranged in parallel to the driving direction. Thanks to the 90°-rotated array axis with the constraints on the microphone position specific to the car environment, a mirror image of the directivity toward the talker with respect to the array axis is no longer projected in the direction of interference and redirected to a direction with no interference. As a result, the talker speech can be discriminated from the interference by directivity, leading to good interference reduction with little speech distortion. Simulation results with signals recorded in a car environment show that the proposed linear microphone array with the array axis parallel to the driving direction has a null in the direction of interference, which leads to 13% higher word accuracy than the conventional microphone array with the array axis perpendicular to the driving direction.

We propose a method for applying a sound source localization (SSL) model trained on simulated data in a real-world environment, with a domain transfer (DT) model for the SSL inside a product structure. The DT model transfers real data into pseudo-simulation data. The SSL model trained on the simulation data is then adapted to the real data using the DT model. As such, our method consists of an SSL model and a DT model. The SSL model predicts the position of a sound source inside the structure, whereas the DT model transforms the data. Because our simulation is not perfect, real data are extrapolated for use with the SSL model. However, the data transformed by the DT model are interpolated within the feature space. The outcome is that the performance of the SSL model in the real world is improved. In our study, the frequency spectra of accelerometers observed on the outer surface of the structure are the model input. The goal is to predict the position of the sound source. The SSL model is built using deep and convolutional neural networks, and the DT model is built using either an autoencoder, a deep convolutional autoencoder, or pix2pix. The two-dimensional distributions of the t-distributed Stochastic Neighbor Embedding indicate that using pix2pix as the DT model shows the best performance. Furthermore, our method's performance for SSL is improved by 57% for the classification problem and by 27% for the regression problem when compared to the case where no transformation is applied.

In public spaces, it is important to reduce ambient noise and surrounding conversation sounds for smooth conversations. Since partitions are often installed at bank counters and shared offices to separate the space from the surrounding area, active noise control can be applied to these partitions. In this paper, we propose a dual active noise control (ANC) that can reduce unwanted sound simultaneously in two spaces separated by the partition while using common sensors. In the proposed dual ANC, one sensor acts as a reference sensor in one system and as an error sensor in the other system. We theoretically clarify signal processing mechanisms to realize simultaneous noise reduction. Simulation results indicate that the proposed dual ANC can achieve more than 6 dB noise reduction simultaneously in the two spaces and the film-like speaker is more suitable for secondary sources to improve the noise reduction performance.

In this paper, we propose a method for controlling the directivity of parametric array loudspeakers (PAL) by optimizing the weights for the sideband signals. By setting the weights for the sideband signals in the single side-band (SSB) modulation to the optimum values, the directivity of the audible sound can be controlled to be ideal. The sum of squares of the difference between the directivity of the reproduced audible sound and the ideal directivity is used as an evaluation function for weight design, and the weights are updated to minimize the evaluation function by the quasi-Newton method. Through the numerical analysis of the sound pressure distribution of PAL beam control using the weights designed by the proposed method, it is demonstrated that the grating lobes are suppressed and the acoustic beam is formed to the desired direction compared with the case using the weights designed by the conventional Chebyshev function.

The virtual sensing technique allows the active noise control (ANC) system to work with error microphones that are placed far from the desired zone of quietness (ZoQ). Conventionally, a training stage is required to obtain the auxiliary filters with the temporary error microphones placed in the ZoQ. When the characteristics of the primary noise changes, the auxiliary filters have to be retrained. As a result, the conventional virtual sensing technique can only be used when the frequency band of the primary noise remains unchanged. In order to solve this limitation, this paper proposes a selective virtual sensing technique for the multi-channel feedforward ANC system. The selective virtual sensing technique obtains a bank of auxiliary filters in the subband structure. Based on the frequency-band-matching mechanism, a linear combination of the auxiliary filters is calculated and used in the real-time control stage. Experimental results show that the selective virtual sensing technique achieves better noise reduction performance than the conventional virtual sensing technique when the frequency band of the primary noise fluctuates.

The parametric array loudspeaker (PAL) modulates the audio signal on an ultrasonic carrier. When the modulated signal is transmitted in air, an audio beam is created based on the nonlinear acoustic principle. Each modulation method has advantages and disadvantages. The frequency modulation (FM) is favorable for its low cost and high volume, but the tradeoff is its complicated nonlinear distortion, which is difficult to be reduced. In this paper, the Volterra filter is adopted to model the nonlinearity of the FM-based PAL. A novel complex inverse system is devised to effectively reduce the nonlinear distortion. Three practical aspects are addressed. First, the computational complexity of the Volterra filter is reduced by the parallel cascade structure with almost no compromise to the model accuracy. Second, Volterra filters are identified at discrete input levels to treat the nonlinearity that keeps changing with the time-varying audio input. Third, when the input level is high, a separation approach is proposed to refine the identified Volterra filters, which eventually improves the performance of the proposed inverse system.

When estimating the correlation/spectral structure of a locally stationary process, one should choose the so-called estimation bandwidth, related to the effective width of the local analysis window. The choice should comply with the degree of signal nonstationarity. Too small bandwidth may result in an excessive estimation bias, while too large bandwidth may cause excessive estimation variance. The paper presents a novel method of adaptive bandwidth selection. The proposed approach is based on minimization of the cross-validatory performance measure for a local vector autoregressive signal model and, unlike the currently available methods, does not require assignment of any user-dependent decision thresholds.

Parametric array loudspeaker (PAL) modulates the audio input on an ultrasonic carrier and relies on airborne nonlinear acoustic effects to generate the audible sound output. The sound output is mainly confined in the beam of the ultrasonic carrier and thus shows a pronounced directivity. There are three parameters that together influence the output volume of a PAL. They are the input level, modulation index, and ultrasound level. In existing PALs, the volume knob is associated with the ultrasound level, while the modulation index is either fixed in the circuit or rarely adjustable by another knob. In this paper, an automatic gain control is proposed to improve the sound quality of the PAL by minimizing the modulation index, maintaining the output-to-input ratio, and ensuring the ultrasound level within the safety range. Simulation and measurement results validate that the proposed approach leads to a reduction in the average total harmonic distortion (THD) level by more than one third for all the tested modulation methods.

The ultrasound-to-ultrasound Volterra filter (U2VF), which was previously proposed to represent the nonlinear response of the parametric array loudspeaker (PAL), identifies the PAL as a nonlinear system that uses ultrasonic signals as its input. It has been proven that the U2VF is a more generic model as compared to the audio-to-audio Volterra filter (A2VF), when the modulation method is adaptive or the input is time varying. However, there is no explicit solution to a linearization system based on the U2VF. Therefore, this paper proposes a synthesis method to extract A2VFs from the U2VF based on the parallel cascade structure. The mature technique of building a linearization system based on the A2VF can hence be adapted to consort with the U2VF.

A theory that predicts the behaviors of the Filtered-X LMS algorithm was derived by using a statistical-mechanical method. In this paper, the theory is generalized to explain the system behaviors in the case of an actual primary path. In the theory, cross-correlations between the element of a primary path and that of an adaptive filter and autocorrelations of the elements of the adaptive filter are treated as macroscopic variables. Simultaneous differential equations that describe the dynamical behaviors of the macroscopic variables are obtained under conditions in which the tapped-delay line is sufficiently long. The equations are analytically solved to obtain the correlations and finally compute the mean-square error. In order to generalize the theory to the case of an actual primary path, the correlations of the elements of the primary path are absorbed. The generalized theory quantitatively predict the behaviors in the case of an actual primary path.

Volterra filters can be applied to a wide range of nonlinear systems, keeping only the low order kernels to yield a good approximation. The parametric array loudspeaker (PAL), as a weak nonlinear acoustic system, is an attractive directional sound reproduction device. Volterra filters have been adopted in the linearization system of the PAL that efficiently reduces the nonlinear distortion with no need of solving the nonlinear acoustic equation. In this paper, the ultrasound-to-ultrasound Volterra filter is proposed, being inspired by the nonlinear acoustic principle, to provide a better systematic representation of the PAL. Experiment results are presented to prove the effectiveness of the proposed approach, where the sparse NLMS algorithm is carried out in the identification.

In this paper, we apply a virtual sensing technique to a head-mounted ANC system we have already proposed. Adaptive feedback ANC system can reduce noise with periodicity or having narrow band components. However, since quiet zones are formed only at the locations of error microphones, an adequate noise reduction cannot be achieved at the locations where error microphones cannot be placed. In the head-mounted ANC system, the error microphones are placed near the opening of the ear canals. However, the intended locations at which the maximum noise reduction should be achieved are near eardrums. Hence, we try to introduce a virtual sensing technique to a head-mounted ANC system to achieve higher noise reduction. Experimental results demonstrate that the proposed system can achieve higher noise reduction than the conventional system. Moreover, a subjective assessment result demonstrates that the proposed system can also give higher noise reduction to human auditory.

We analyze the learning curves of the FXLMS algorithm using a statistical-mechanical method when the reference signal is not necessarily white. We treat the nonwhite reference signal by introducing the correlation function of the signal to the method proposed in our previous study. Cross-correlations between the element of a primary path and that of an adaptive filter and autocorrelations of the elements of the adaptive filter are treated as macroscopic variables. We obtain simultaneous differential equations that describe the dynamical behaviors of the macroscopic variables under the conditions in which the tapped-delay line is long. We analytically solve the equations to obtain the correlations and finally compute the meansquare error. The obtained theory quantitatively agrees with the results of computer simulations. The theory also gives the upper limit of the step size in the FXLMS algorithm.

A new adaptive comb filtering algorithm, capable of tracking the fundamental frequency and amplitudes of different frequency components of a nonstationary harmonic signal embedded in white measurement noise, is proposed. Frequency tracking characteristics of the new scheme are studied analytically, proving (under Gaussian assumptions and optimal tuning) its statistical efficiency for quasi-linear frequency changes. Laboratory tests show that the proposed algorithm can be successfully used for active control of MRI noise.

In this paper, the compensation ability of nonlinear distortions for loudspeaker systems is demonstrated using dynamic distortion measurement. Two linearization methods using a Volterra filter and a Mirror filter are compared. The conventional evaluation utilizes swept multi-sinusoidal waves. However, it is unsatisfactory because wideband signals such as those of music and voices are usually applied to loud speaker systems. Hence, the authours use dynamic distortion measurement employing a white noise. Experimental results show that the two linearization methods can effectively reduce nonlinear distortions for wideband signals.

Recently, magnetic resonance imaging (MRI) devices are used in many medical institutions on the grounds of safety and convenience. An open-configuration MR system is introduced at Shiga University of Medical Science in order to conduct microwave coagulation therapy by using near-real-time MR images. However, this system has a fatal defect. When MRI device works to take images, it also generates serious noises (MR noise). Hence, an operator and other medical staff (ex. nurses and anesthetists) suffer from MR noise and cannot communicate with each other during the operation. In this paper, we therefore propose a head-mounted ANC system in order to reduce the MR noise, and some experimental results demonstrate cancellation performance of the system implemented by digital signal processor (DSP).

In this paper, we propose an acoustic echo cancellation (AEC) using a sub-adaptive filter. In the AEC, the step-size parameter of the adaptive filter must be varied according to the situations where a double talk and an echo path change occur. The proposed AEC can appropriately control the step-size parameter even if the double talk and the echo path change simultaneously occur because the optimal step-size parameter can be obtained according to the output of the sub-adaptive filter and the echo path change detector is controlled through the double talk detector. Hence, the proposed AEC can realize superior convergence property to the conventional one. Simulation results demonstrate that the proposed AEC can achieve higher ERLE and faster convergence than the conventional one.

In this paper, we propose an updating algorithm considering the frequency characteristic of the secondary path in order to improve the cancellation performance of active noise control (ANC) system using the perturbation method. Since the ANC system using the perturbation method does not need the secondary path model, it has an advantage that can track the path changes. However, the conventional perturbation method has a problem that the cancellation performance deteriorates as the frequency of noise becomes high. The proposed method can improve the cancellation performance for the noise at the higher frequency to consider the damping of acoustic wave. The effectiveness of the proposed method is dmonstrated through experimental results.

In this paper, we present a novel multi-channel active noise control (ANC) system with freely movable error microphones. This ANC system uses a simultaneous perturbation algorithm and has an advantage that secondary path models (estimation of secondary paths) are not required unlike the conventional MEFX (multiple error filtered-X) based ANC. This system can consequently control noise stably because there are not modeling errors which cause system instability. The computational complexity is also very small. We demonstrate that the proposed multi-channel ANC system can operate stably under an environment where the error microphones always move.

In this paper, we propose a linearization (compensation of nonlinear distortion) method for loudspeaker systems using the MINT (multiple-input/output inverse-filtering theorem) and Volterra filters. In the proposed method, linear inverse filtering of a target loudspeaker system is realized by using the MINT so that exact linear inverse filtering can be realized. The linearization performance becomes consequently very high. On the other hand, since the conventional linearization method cannot realize exact linear inverse filtering, the performance deteriorates remarkably. Experimental results demonstrate that the proposed method has about 20dB higher performance than the conventional one.

This paper proposes a novel method for transmitting JPEG2000 images in MC-DS-CDMA taking into account the codestream structure. In the conventional MC-DS-CDMA that transmits the serial-to-parallel converted codestream by turns, the JPEG2000 image is corrupted by bit errors of the important parts of the codestream. To solve this problem, we determine an order of priority of the JPEG2000's codestream and then transmit each codestream with different weights in order to reduce bit errors in the important parts. Also, some divided images of different weights are transmitted, and on the receiver the original image can be restored by filtering the divided images based on the determined weights.

We propose a frequency-domain active noise control (ANC) system without a secondary path model. The proposed system is based on the frequency-domain simultaneous perturbation (FDSP) method with variable perturbation. In this system, the coefficients of the adaptive filter are updated only by error signals. The conventional ANC system using the filtered-x algorithm becomes unstable due to the error between the secondary path, from secondary source to error sensor, and its model. In contrast, the proposed ANC system has the advantage of not using the model. Furthermore, the variable perturbation brings fast convergence. Simulation results demonstrate the efficiency of the proposed ANC system compared with the conventional ANC systems.

We propose a frequency domain active noise control system using optimal step-size parameters at each frequency. The proposed ANC system can converge faster than the conventional ANC system using the Filtered-x LMS algorithm with the optimal step-size parameter. Moreover, the proposed system can converge by setting the step-size parameters at unstable frequencies to 0 in the case where the phase error of the secondary path model does not satisfy the stable condition, whereas the conventional ANC system cannot converge in this case. In this paper, the theoretical equation of the optimal step-size parameters is derived by using available information during system operation. Next, we present the structure of the ANC system using the optimal step-size parameters obtained from the theoretical equation. Moreover, a control technique determining unstable frequencies is introduced. Finally, simulation results demonstrate the efficiency of the proposed ANC system.

We propose two methods for reducing the computational complexity of Volterra filters. First, a method reducing the computational complexity of Volterra filters is proposed. This method can be realized by incorporating multirate signal processing into the Volterra filters. Hence, it is possible to operate the bandlimited Volterra filter at a low sampling rate and with a short system length. Second, we also propose a method to replace the conventional Volterra filter with one including many zero coefficients by using multirate signal processing. The conventional Volterra filter is bandlimited in order to avoid aliasing so that waste arithmetic is done. In contrast, the Volterra filter including many zero coefficients derived by the proposed method can eliminate such waste arithmetic. We demonstrate their effectiveness in application to loudspeaker systems whose nonlinear distortions are generally concentrated in the lower frequency band. Even though the processed frequency band is limited, the proposed method has about 0.03 times as many computational complexities as the conventional method.

The proposed system is based on the simultaneous perturbation optimization method with block process. Consequently, the coefficients of the adaptive filter in the proposed system are updated by only error signals. The conventional ANC system using the filtered-x algorithm becomes unstable due to the error between the secondary path and its model. On the other hand, the proposed ANC system has an advantageous property in not using the model. In this paper, we show the principle of the proposed ANC system and examine the efficiency of computer simulations.

Modeling a loudspeaker system with the Volterra series expansion is essential to eliminate the nonlinear distortion. We have proposed a method for measuring the Volterra kernel of a loudspeaker system by multi sinusoidal waves. This method, however, has the problems of not considering the phase property of the nonlinear element and the third-order distortion. Therefore, we propose a novel method measuring the second-order Volterra kernel by multi sinusoidal waves. In this method, the phase property of the nonlinear element and the third-order distortion are considered. Moreover, we off-line eliminate the nonlinear distortion of the loudspeaker system with the Volterra kernel measured by the novel method and show the method's effectiveness.