This paper reviews pioneering works in microphone array processing and multichannel speech enhancement, highlighting historical achievements, technological evolution, commercialization aspects, and key challenges. It provides valuable insights into the progression and future direction of these areas. The paper examines foundational developments in microphone array design and optimization, showcasing innovations that improved sound acquisition and enhanced speech intelligibility in noisy and reverberant environments. It then introduces recent advancements and cutting-edge research in the field, particularly the integration of deep learning techniques such as all-neural beamformers. The paper also explores critical applications, discussing their evolution and current state-of-the-art technologies that significantly impact user experience. Finally, the paper outlines future research directions, identifying challenges and potential solutions that could drive further innovation in these fields. By providing a comprehensive overview and forward-looking perspective, this paper aims to inspire ongoing research and contribute to the sustained growth and development of microphone arrays and multichannel speech enhancement.

This paper proposes a new technical writing method based on a three-point analysis and a slide-first principle. The three-point analysis clarifies the most essential components of a paper, namely, a value, a trick, and a problem. The value is the outcome or the achievement of the research, which is something first in the world. The trick is the novelty such as a means, a system, or an ingredient, which enables the value. The problem is what has been solved by the research and opposite to the value. The three-point analysis helps the author avoid a commonly encountered problem that the outcome of the research does not solve the problem stated in the paper. Slide preparation with a large minimum font size develops a solid logic in the paper with the value, the trick, and the problem. Contents of each slide is clearly instructed assuming ten total pages so that unexperienced authors would not get lost. Due to simple expressions, the logical structure is more visible and easily modified by simply changing the order of the slides. 93% of those who learned the proposed method rated their satisfaction as excellent or good, demonstrating that it is useful for a wide range of authors.

This paper proposes an adaptive noise canceller algorithm with an SNR-based stepsize and data-dependent averaging. A first and a second SNR with different time constants for averaging are used one after the other to control the stepsize in adaptation. The time constants reflect the accuracy of the SNR estimates such that lower accuracy is offset with a larger time constant. The first SNR estimate with lower accuracy guarantees the initial coefficient growth whereas the second SNR estimate provides more accurate adaptation control. Switchover from the first to the second SNR estimate takes place when the coefficient growth is saturated. Evaluations with clean speech and noise recorded at a busy station demonstrate that the coefficient error by the proposed algorithm is as much as 6dB smaller than that by conventional algorithms with a constant averaging. It is a counterintuitive result that averaging a first estimate does not always make it better.

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.

This paper proposes a robust adaptive noise canceller algorithm with SNR-based stepsize control and noise-path gain compensation. The stepsize for coefficient adaptation is controlled with an estimated SNR for low distortion of the output signal and a small residual noise. A second SNR estimate, which is the output over an adjusted reference input, initially controls the stepsize to promote co-efficient growth, followed by a more accurate first SNR estimate defined as the output over the noise replica. The power gap between the reference input and the noise replica is compensated for by a factor estimated during an initial target-signal absence. Switchover from the second to the first SNR estimate takes place when the coefficient growth is saturated to guarantee robustness to different noise-path gains. Evaluations with clean speech and noise recorded at a busy station demonstrate that conventional algorithms exhibit initial in-crease in the coefficient error and never reach the switchover status at a high SNR whereas the proposed algorithm achieves as much as 8dB smaller coefficient error than that without gain compensation.

This paper proposes a fast convergence algorithm for adaptive noise cancellers with novel SNR (signal-to-noise ratio) estimation and stepsize control. The stepsize for coefficient adaptation is controlled with an estimated SNR for low distortion of the output signal and a small residual noise. A first SNR estimate, defined as a ratio of the noise-canceller output to the reference input, enables quick coefficient growth in the initial adaptation stage. A second SNR estimate, which replaces the reference input in the denominator of the first SNR estimates with the noise replica, achieves low distortion in the output and small residual noise simultaneously. Switchover from the first to the second SNR estimate takes place when the latter becomes comparable to the former. Evaluations with clean speech and noise recorded at a busy station demonstrate that the time until SNR-estimate switchover is reduced by 91% at SNR=0dB compared to the conventional algorithm, leading to fast evolution of adaptive filter coefficients.

This paper presents a noise-robust hearable device with an adaptive noise canceller. The hearable device has an inner-ear and an outer-ear microphone to collect ear-canal speech signal and the outer-ear environmental signal, respectively. The environmental signal drives an adaptive filter to generate a noise replica which is subtracted from the inner-ear microphone signal to cancel the noise. Coefficients of the adaptive filter are updated by an NLMS (normalized least mean square) algorithm with an SNR-controlled stepsize for small speech distortion. The adaptive noise canceller is realized on a DSP (digital signal processor) chip with 39 McPS for 64 taps. Evaluation results in speech recognition demonstrates that a word recognition rate of 89% is achieved when the noise level is 80 dBA with 38% improvement over the inner-ear signal without adaptive noise cancellation.

This paper proposes an adaptive noise canceller with SNR estimate switchover for stepsize control. The SNR estimate is switched from a first estimate to a second one when the latter takes a smaller value than the former. The first estimate is a power ratio of the primary and the reference input signals, which is newly introduced to enable initial coefficient adaptation. The second estimate is a power ratio of the error signal and the adaptive filter output, which is used in a conventional stepsize control. Switchover from the first estimate to the second guarantees initial coefficient adaptation by the former and provides accurate SNR estimation by the latter. Evaluations with speech-speech inputs and speech-noise inputs demonstrate that use of the second SNR followed by the first SNR successfully makes the coefficient vector norm approach the true value which represents convergence.

This paper proposes diagonal microphone placement for the landscape/portrait interchangeable mode of a personal computer. Two microphones are diagonally located at a screen corner to provide useful directivity for target-signal enhancement in both landscape and portrait modes. The diagonal microphone placement guarantees a nonzero microphone spacing in either the landscape and the portrait mode along the ground surface leading to successful interference attenuation. Evaluations in a speech recognition scenario demonstrate that the diagonal microphone placement is effective in both the landscape and the portrait position with comparable command recognition rates of 80 to 99%.

This paper proposes automatic gain control with integrated signal enhancement for specified target and background-noise levels. The input signal is separated into a target signal and background noise by signal enhancement. The processing gains for the target and the noise are calculated from their estimated magnitudes and desirable output levels. After independent amplification/attenuation of the target signal and the noise, they are summed up to construct the output signal with their specified levels. Because of independent gain control, a more suitable signal for user needs can be obtained. Evaluation results demonstrate that the proposed gain control provides as much as 6 dB better gain than a commercial voice recorder with comparable signal quality.

This paper proposes gain relaxation in signal enhancement designed for speech recognition with an unaware local noise source. An attention is drawn to a new performance degradation problem in signal enhancement combined with automatic speech recognition (ASR), which is encountered in real products with an unaware noise source. Gain relaxation, as a solution, selectively applies softer enhancement of a target signal to eliminate potential degradation in speech recognition caused by small undesirable distortion in the target signal components. Evaluation of directional interference suppression with signals recorded by a commercial PC (personal computer) demonstrates that signal enhancement over the input is achieved without sacrificing the performance for clean speech.

this paper proposes a directional noise suppressor with a specified constant beamwidth. A directional gain is calculated based on interchannel phase difference and combined with a spectral gain commonly used in single-channelnoise suppressors (NSs). The beamwidth can be specified as passband edges of the directional gain. In order to implement frequency-independent constant beamwidth, frequency-proportionate band-edge phase differences are determined for the passband. Evaluation with signals recorded by a commercial PC demonstrates good agreement between the theoretical and the measured directivity. The signal-to-noise ratio improvement and the PESQ score for the enhanced signal are improved by 24.4 dB and 0.3 over a conventional NS.

This paper proposes an auto-focusing noise suppressor for cellphone movies based on phase randomization and power compensation. The input signal is analyzed in the frequency domain to detect and preserve important spectral components including peaks. All other components are suppressed to the background signal level that is estimated during absence of the important components. Residual spikes by auto-focusing noise are suppressed by phase randomization, which is not possible without phase manipulation. Power reduction by phase randomization is compensated for by an analytically obtained factor. Subjective evaluation results demonstrate that the proposed auto-focusing noise suppressor achieves a score of 1.0 in the 7-grade comparison MOS (CMOS) compared to the one without compensation with a statistically significant difference.

This paper proposes a new generalized sidelobe canceller with a compact array of microphones suitable for mobile terminals. The output of the fixed beamformer (FBF) is further processed by a newly introduced decorrelation unit which has an auxiliary input signal to improve poor interference suppression of FBF in low frequencies. The output of the decorrelation unit is used as the reference signal for the adaptive blocking matrix and the input for the multi-input canceller. Because low and high frequency components are cancelled or suppressed by the decorrelation unit and FBF, better output-signal quality is obtained. Output signal comparison confirms approximately 12 dB higher interference cancellation by the new beamformer.

This paper proposes an auto-focusing noise suppressor (AF-NS) for cellphone movies. For relatively small clicks compared to the target signal, it consists of magnitude suppression and complementary phase modification. The input signal is analyzed in the frequency domain to detect and preserve important spectral components including peaks and their vicinities. All other components are suppressed to the environmental signal level that is estimated during absence of the important components. Residual spikes by auto-focusing noise are successfully suppressed by phase randomization. Subjective evaluation results demonstrate that the proposed AF-NS achieves scores of 1.6 and 1.8 in the 7-grade modified CMOS with statistically significant differences compared to the input noisy signal and an AF-NS with no phase randomization.

This paper proposes a new paradigm with phase randomization for single-channel signal enhancement. In contrast to literatures which pursue better target signal quality, the new method tries to minimize artifacts in the residual noise. Applications of signal enhancement are revisited to highlight today's examples where environmental signal is often considered as a part of target and SNR may take a negative value. A signal example demonstrates that conventional signal enhancement with magnitude-only modification is insufficient from both objective and subjective points of view. A new framework with phase randomization as well as a specific algorithm is developed. Enhanced signals show that phase randomization is an integral component for sufficient enhancement. A subjective evaluation result demonstrates that the new paradigm with phase randomization is superior to the magnitude-only enhancement with statistically significant differences.

This paper proposes a noise canceller with adaptive delay compensation for a distant noise source. A delay element is introduced in the noise path to offset the delay caused by the distance from the distant noise source to the reference microphone. To cope with a variety of the distance, it is controlled based on the significance of adaptive-filter coefficients. The delay is increased when the coefficients in the head are more significant than those in the tail, and vice versa. This adaptive control makes the adaptive filter with a limited number of coefficients cover the most significant part of the impulse response of the noise path, thereby achieving the best possible noise cancellation. Simulation results demonstrate superior identification capability, that leads to noise cancellation as good as that with the true amount of delay.

This paper proposes a stereo echo canceller with simultaneous input-sliding and sliding-period control. Periodic delays with different phases in two channels provide an additional condition for echo cancellation, leading to faster convergence. It also guarantees symmetrical shifts of a sound image around its original position for better subjective quality of the far-end signal. Sliding-period control with more slides for less interchannel correlation of the input signals achieves less noticeable sliding artifacts and faster convergence simultaneously. Subjective evaluation results demonstrate that simultaneous sliding achieved comparable far-end signal (FES) quality to that of the original FES. It is also shown that sliding-period control tops up 14% improvement in convergence time without trading off the subjective quality of the FES.

This paper presents a single-chip speech dialogue module and its evaluation on a personal robot. This module is implemented on an application processor that was developed primarily for mobile phones to provide a compact size, low power-consumption, and low cost. It performs speech recognition with preprocessing functions such as direction-of-arrival (DOA) estimation, noise cancellation, beamforming with an array of microphones, and echo cancellation. Text-to-speech (TTS) conversion is also equipped with. Evaluation results obtained on a new personal robot, PaPeRo-mini, which is a scale-down version of PaPeRo, demonstrate an 85% correct rate in DOA estimation, and as much as 54% and 30% higher speech recognition rates in noisy environments and during robot utterances, respectively. These results are shown to be comparable to those obtained by PaPeRo.

This paper proposes a robust NLMS algorithm with a novel noise modeling based on stationary/nonstationary noise decomposition. The ambient noise including the near-end signal is modeled as a weighted sum of the stationary and the nonstationary components. These components are independently estimated with an appropriate time constant for better accuracy. The estimates are weighted by the stationary/ nonstationary likelihood before summation. The integrated noise estimate controls the coefficient adaptation stepsize such that it is an upward convex function of the reference input with a noise offset for robustness. Evaluations in a fullband and a subband echo cancellation scenarios show that ERLE has been improved by as much as 40 dB over the algorithm with a conventional noise model in both single- and double-talk sections with no double-talk detection.

This paper presents implementation of a multipurpose noise suppressor based on a novel scalable framework for a wide range of applications. Spectral-gain calculation incorporates adaptive control of a spectral-gain floor for an appropriate balance between residual noise and distortion in the enhanced speech. Use of a common spectral gain among channels calculated from a down-mix signal helps reduce computations on gain calculation compared to independent spectral-gain calculation. Subjective evaluation results demonstrate that the difference in the 5-grade mean opinion score with and without the multipurpose noise suppressor in the downlink is improved by as much as 0.5. When it is used as an uplink noise suppressor, the enhanced-speech quality is statistically better than or equal to that with a conventional noise suppressor. Sampling rate vs. MIPS for monaural and stereo implementations, clock rate vs. current, frame size/delay vs. MIPS based on TMS320V5510 are also presented.

This paper proposes a new acoustic echo suppressor based on a frequency-domain model of highly nonlinear residual echo. The proposed echo suppressor controls the gain in each frequency bin based on the new model where the highly nonlinear residual echo is approximated as a product of a regression coefficient and the echo replica in the frequency domain. To reduce annoying modulation by the error of the model, a flooring operation of estimated near-end signal level is introduced to the gain control. Simulation results with speech data recorded by a hands-free cellphone show that the proposed echo suppressor reduces the highly nonlinear residual echo to an almost inaudible level

This paper proposes a low-complexity noise suppressor with nonuniform subbands and a frequency-domain highpass filter. A highpass filter to suppress undesirable low frequency components in the input signal is replaced with a simple DC offset canceler in the time-domain combined with a frequency-domain weighting function. Frequency bins after Fourier transform are nonuniformly grouped to reduce the computations for calculating the spectral gain. A new decomposition pattern is developed for higher quality of the enhanced speech. The number of operations on typical DSP chips is 3.2 MIPS that is approximately 50% off the conventional load. Subjective evaluation results demonstrates that the low-complexity noise suppressor achieves comparable quality of the enhanced speech to that of the conventional noise suppressor

This paper presents a noise-robust double-talk detection algorithm based on normalized cross-correlation and a noise offset. The noise offset alleviates undesirable influence by the background noise existing in the microphone signal. It is estimated from the echo-cancelled signal when its autocorrelation is low and its power is smaller than the echo-replica power. A detection threshold of the new normalized cross-correlation is adaptively controlled based on the echo-to-NES (near-end speech) ratio. Simulation results demonstrate superior performance of the new algorithm.

This paper proposes a generalized proportionate variable step-size algorithm based on affine projection. It controls the step size for each tap individually based on the gradient approximated by the difference between the current coefficient and an averaged filter coefficient with delay. This step-size control is specifically effective for tracking fast changing acoustic environments. It is shown that the P-NLMS family and the ES algorithm are special cases of the proposed algorithm in terms of step-size control. Simulation results in the context of echo cancellation in artificial and real environments demonstrate that this step-size control combined with the affine projection algorithm (APA) reduces the echo by up to 5 dB compared to the standard APA. The best tracking performance, without a priori knowledge on the acoustic environment, among the same family of algorithms is obtained.

This paper proposes an adaptive noise canceller (ANC) with low signal-distortion for human-robot communication. The proposed ANC has two sets of adaptive filters for noise and crosstalk; namely, main filters (MF) and subfilters (SF) connected in parallel thereto. To reduce signal-distortion in the output, the stepsizes for coefficient adaptation in the MF are controlled according to estimated signal-to-noise ratios (SNR) of the input signals. This SNR estimation is carried out using SF output signals. The stepsizes in the SF are determined based on the ratio of the primary and the reference input signals to cope with a wider range of SNR. This ratio is used as a rough estimate of the input signal SNR at the primary input. Computer simulation results using TV sound and human voice recorded in a carpeted room show that the proposed ANC reduces both residual noise and signal-distortion by as much as 20 dB compared to the conventional ANC. Evaluation in speech recognition with this ANC reveals that with a realistic TV sound level, as good a recognition rate as in the noise-free condition is achieved.

The paper presents a family of 3GPP-standard noise suppressors and the evaluation results. The family consists of a high-quality version and a low-complexity version. These noise suppressors are based on the MMSE STSA (minimum mean square error short time spectral amplitude) algorithm originally proposed by Y. Ephraim and D. Malah (see IEEE Trans. Acoust., Speech, Sig. Processing, vol.ASSP-32, no.6, p.1109-21, 1984). To meet the 3GPP requirements with better speech quality, weighted noise estimation, synthesis windowing, and pseudo noise injection are incorporated. Weighted noise estimation enables continuous noise estimation even in the speech period by using weighted noisy speech. The weight is controlled such that a higher estimated SNR gives a smaller weight. A synthesis window function is applied between inverse transform and overlap-add processing for smooth transition at frame boundaries. Pseudo noise injection, which is not available in the low-complexity version, modifies the spectral gain based on its nonlinearity. The whole family satisfies all the 3GPP requirements. Results of a full set of evaluations specified by 3GPP are presented for the high-quality version.

A noise suppression algorithm with synthesis windowing and pseudo noise injection is proposed. A synthesis window function is applied between inverse transform and overlap-add processing for smooth transition at frame boundaries. The pseudo noise injection modifies the spectral gain based on its nonlinearity. It is selectively disabled for stationary noise so that possible distortion is minimized. Evaluations including subjective assessment show that annoyable peaks of the background noise are suppressed by as much as 15 dB, resulting in a better MOS than that by the EVRC noise suppressor.

This paper proposes an efficient multichannel line echo canceler algorithm with interchannel distribution of computations. A limited number of coefficient adaptations are distributed among channels depending on the degree of convergence and the input signal power to achieve up to 50% reduction of total computations. Coefficient adaptations are more frequently performed in the channels where convergence is behind others and there is a sufficient input power. As the index to the degree of convergence, an averaged sum of squared coefficients is used. Its gradient weighted with the input signal power is evaluated for interchannel distribution of coefficient adaptations. Simulation results with white Gaussian signals and speech signals demonstrate good convergence. The efficiency of coefficient adaptations is improved by 30% over the conventional algorithm. The computational savings can be used to accommodate more channels on the same chip, or to cover a longer echo-path with additional delays by codecs and/or ATM cell assembly/disassembly. It is also promising for analog interface in IAD and SOHO routers for VoIP/VoDSL applications.

This paper presents an adaptive microphone array using two auxiliary fixed beamformers for good sound quality. One auxiliary fixed beamformer is introduced in the target signal path to avoid suppression of high-frequency components in the total output. The other auxiliary fixed beamformer is used for adaptation-mode control to eliminate the hysteresis in the relationship between the signal direction and sensitivity. Both auxiliary fixed beamformers bring about good sound quality, which improve intelligibility in speech communications and speech recognition rate. The proposed microphone array is implemented on a DSP system, which demonstrates flat frequency response and less hysteresis in its directivity pattern.

A robust adaptive microphone array (RAMA) using a new adaptation-mode control method (AMC) is proposed, and its evaluation results by hardware are presented. The adaptation of the RAMA is controlled based on an SNR (signal-to-noise) estimate using the output powers of the fixed beamformer and the adaptive blocking matrix. The RAMA is implemented on a multi-DSP real time signal-processing system with a C-compiler. Simulation results with real acoustic data show that the AMC based on the SNR estimate causes less breathing noise than the conventional AMC and that it obtains 1.0-point higher score on a 5-point mean opinion score scale. Evaluation through a real time signal-processing system demonstrates that noise reduction achieved by the RAMA is over 12 dB even in reverberant environments.

A new stereo echo canceler with pre-processing for correct echo-path identification is proposed. The pre-processing is accomplished by a two-tap time-varying filter which delays the input signal periodically by one sample in one of the two channels. Aliasing components and audible clicks by pre-processing are made inaudible by selecting appropriate parameters for the filter. Simulations with the NLMS algorithm and a white Gaussian signal confirm the correct echo-path identification. For speech signals, the convergence speed of the proposed echo canceler is more than three times faster than that of an echo canceler with nonlinear transformations. Results of a subjective listening test demonstrate that quality of the pre-processed signals is 4.38 using the CCIR five-grade impairment scale. This is acceptable for general teleconference applications.

This paper presents a new robust adaptive microphone array (AMA) and its evaluation in an echoic environment. The proposed AMA is a generalized sidelobe canceller equipped with a variable blocking matrix using coefficient-constrained adaptive filters, and a multiple-input canceller using norm-constrained adaptive filters (NCAFs). Because the NCAFs have selective nonlinearity in the relationship between coefficient norm and coefficient error, the proposed AMA has better spatial selectivity than the conventional AMA. Evaluation with real acoustic data captured in a room of 0.3-second reverberation time shows that the noise is suppressed by 19 dB. In subjective evaluation, the proposed AMA obtains 3.8 on a 5-point mean opinion score scale.

This paper proposes a tonal component coding algorithm for a codec that employs a transform followed by Huffman coding, such as MPEG-2 Audio NBC (non-backward compatible). After the input audio signal is mapped onto a frequency domain, the proposed algorithm withdraws local maximum components that degrade coding efficiency. By this withdrawal, the flatness of the spectrum increases and the efficiency in Huffman coding is improved. The withdrawn components are encoded separately as side information. When the frequency resolution of the time/frequency mapping is high, this algorithm works more effectively since local maximum samples appear more frequently with such a mapping. Simulation results show that this algorithm achieves as much as 11% bit reduction per frame and improves the coding efficiency in 41% of all the audio frames.

This paper presents the implementation and performance evaluation of sparse-tap adaptive FIR filters with tap-position control. For the reduction of computational costs, a small number of filter coefficients are assigned to significant regions in the impulse response. A floating-point digital signal processor (DSP) ADSP-21020 is used for the implementation. For one second of end-to-end delay and 32 milliseconds of total significant-region length, the number of computations is reduced to almost one-tenth compared with a full-tap FIR adaptive filter using the same DSP. Experimental results using a telephone network simulator show that the implemented echo canceller reduces the echo by almost 20 dB for both white-noise and real speech signals. Tracking performance for phase rolls and flat-delay changes is also examined.

This paper proposes a new robust adaptive beamformer applicable to microphone arrays. The proposed beamformer is a generalized sidelobe canceller (GSC) with a variable blocking matrix using coefficient-constrained adaptive digital filters (CCADFs). The CCADFs minimize leakage of target signal into the interference path of the GSC. Each coefficient of the CCADFs is constrained to avoid mistracking. The input signal to all the CCADFs is the output of a fixed beamformer. In a multiple-input canceller, leaky ADFs are used to decrease undesirable target-signal cancellation. The proposed beamformer can allow large look-direction error with almost no degradation in interference-reduction performance and can be implemented with a small number of microphones. The maximum allowable look-direction error can be specified by the user. Simulation results show that the proposed beamformer designed to allow about 20 degrees of look-direction error can suppress interference by more than 17 dB.

This paper proposes a new noise-robust adaptive FIR filtering algorithm with an adaptive step-size which takes non-stationarity of speech signals into account. The proposed algorithm controls the step-size based on the reference input signal power and the estimated noise power. Implementation of an acoustic echo canceller based on this algorithm using digital signal processors is also described. Computer simulation results using a real speech signal show that the proposed algorithm improves the ERLE (echo return loss enhancement) by more than 10 dB compared with conventional noise-robust adaptive-step algorithms. The implemented echo canceller achieves 25 dB of the ERLE, which agrees with the computer simulation results.

This paper proposes a new subband adaptive filtering algorithm with intersubband tap assignment. The number of taps for each subband adaptive filter is adaptively controlled based on a sum of squared coefficients to the impulse-response tail of the unknown system. The sum is weighted by the input reference signal power to reflect its power imbalance. For rapid convergence of intersubband tap assignment, a time-varying number is newly introduced as the tap-redistribution step-size. This variable is controlled based on repeated redistribution of the maximum number of taps to the same subband. For colored signals, the proposed algorithm reduces the convergence time for intersubband tap-assignment by more than 60% compared with a fixed tap-redistribution step-size. It is insensitive to parameter-selection and has good capability to track a path change. It is also shown that the proposed algorithm is effective for real speech signals.

This paper presents a stereo echo canceller implemented by digital signal processors (DSPs) and its performance evaluation. The implemented echo canceller, which takes the cross-correlation between stereo signals into account, has a single adaptive filter per channel. This structure reduces the number of adaptive filters by half compared with the conventional stereo echo canceller which requires two adaptive filters per channel. Analytical results with real impulse responses clarify the conditions in which this echo canceller achieves better performance with a smaller number of taps than the conventional one. Applicable arrangements of microphones and a talker at the far-end are shown. Evaluation by the implemented hardware supports the analytical results. Satisfactory performance is obtained for both white-noise and real-speech inputs.>

This paper proposes a fast convergence algorithm for sparse-tap adaptive FIR filters for an unknown number of multiple echoes. Coefficient values and tap-positions are simultaneously controlled. A constrained region for new-tap positions is selected from equisize subgroups of all possible tap-positions and it hops from one subgroup to another to cover multiple echoes. The hopping order and the staying time for each subgroup are adaptively determined based on the absolute coefficient values. Simulation results with colored signals show that the proposed algorithm saves more than 80% in the convergence time over the full-tap NLMS and 50% over the STWQ. Tracking capability of the proposed algorithm exhibits its superior characteristics.>

A Hi-Fi audio codec with an improved adaptive transform coding (ATC) algorithm is presented using digital signal processors (DSPs). An audio signal with a 20 kHz bandwidth sampled at 48 kHz is coded at a rate of 128 kb/s. The algorithm utilizes adaptive block size selection, which is effective for preecho suppression. A modified discrete cosine transform (MDCT) with a simple window set is employed to reduce block boundary noise without decreasing the performance of transform coding. In addition, a fast MDCT calculation algorithm, based on a fast Fourier transform, is adopted. Weighted bit allocation is employed to quantize the transformed coefficients. The codec was realized by a multiprocessor system composed of newly developed DSP boards. Subjective tests with the codec show that the coding quality is comparable to that of compact disc signals.>

A fast convergence algorithm for adaptive FIR (finite impulse response) filters with coarsely located taps is proposed. The algorithm has two-stage adaptation and tap-position control with a constraint on the region. The two-stage adaptation includes estimation of the flat delay of the impulse response and coefficient adaptation with independent tap-position control. The constrained tap-position control is achieved by imposing a limit on new-tap-position search. Assuming an echo canceler application, performance in single-talk was evaluated by means of a white Gaussian signal. The algorithm reduces the convergence speed by up to 80% over both STWQ (scrub taps waiting in a queue) and the full-tap NLMS (normalized least mean square) and always provides faster convergence speed than the full-tap NLMS. The proposed algorithm is applicable to adaptive FIR filters to model a path with long flat delay, such as echo cancelers for satellite communications.>

A coding technique is presented for high-quality audio signals based on adaptive transform coding (ATC). Adaptive block size selection by the proposed algorithm ensures an appropriate block size resulting in improved SNR (signal-to-noise ratio) for a wide variety of source signals. A feedback approach, based on SNR, and a feedforward approach, based on interblock differences in input time-domain samples, to adaptive block size assignment are proposed and evaluated. Computer simulation results show that average segmental SNR by the feedback approach is improved by as much as 4.8 dB over the conventional fixed-block-size ATC. The feedforward approach is realized with much-simplified hardware; nevertheless, its SNR degradation from that by the feedback approach is 1.6 dB, even in the worst case. Both approaches are successful in pre-echo suppression to a satisfactory level. Time-domain aliasing cancellation has the potential to increase the superiority of the new algorithm.>

The authors present a fast convergence algorithm for adaptive FIR (finite impulse response) filters. The algorithm controls the step size on the basis of the moving average of the misadjustment level for a faster convergence. Several operations in addition to the LMS operations are used to estimate the misadjustment and average it to provide a stable step size. The algorithm is applied to adaptive noise cancellation to observe the performance. Computer simulation results for noise cancellation show that the convergence characteristics of the algorithm are improved by as much as 90% over those of the lease-mean-squares (LMS) algorithm and even those of the VS (variable step) algorithm for a multitone signal. For a real speech signal, improvement of the proposed algorithm is degraded, but still more than 50% of the LMS convergence time is saved in the simulated case.>

This paper proposes a new convergence method for adaptive digital filters applied to echo cancellation and decision feedback equalization. The proposed method, using Disturbing Signal Canceler (DSC), saves adaptive filters from falling into a deadlock. DSC takes advantage of a line code property, wherein the number of symbol waveforms is limited and each waveform has both polarities. Since DSC is realized simply with an adder and a delay element, it is suitable for LSI implementation. Theoretical analyses were carried out to show the convergence characteristics and the convergence condition. Computer simulations prove that both an echo canceler and a decision feedback equalizer based on the proposed method, employing DSC, possess satisfactory performance.

This paper proposes an ISDN subscriber loop transmission system based on an echo cancellation technique which fully supports the CCITT recommended basic interface for ISDN, i.e.,B + B + Dover metallic cables. System design considerations of cable loss, crosstalk, impulsive noise, and system requirements such as power feeding on a caLl-by-call basis are discussed, and a transmission system is proposed. The proposed system is characterized by the use of a new line code, named "modified biphase code." The use of this code with fixed line loss equalization and coherent detection achieves circuit simplification. A special adaptation algorithm is incorporated with the echo canceller that significantly reduces convergence time. Experimental results show satisfactory performance with respect to the residual echo and the bit error rate. Since the system requires only a simple nonadaptive circuit for analog portion, it is suitable for LSI implementation.