This paper proposes new algorithms for adaptive FIR filters. The proposed algorithms provide both fast convergence and small final misadjustment with an adaptive step size even under an interference to the error. The basic algorithm pays special attention to the interference which contaminates the error. To enhance robustness to the interference, it imposes a special limit on the increment/decrement of the step-size. The limit itself is also varied according to the step-size. The basic algorithm is extended for application to nonstationary signals. Simulation results with white signals show that the final misadjustment is reduced by up to 22 dB under severe observation noise at a negligible expense of the convergence speed. An echo canceler simulation with a real speech signal exhibits its potential for a nonstationary signal.

This paper proposes a fast convergence algorithm for adaptive FIR filters with sparse taps. Coefficient values and positions are simultaneously controlled. The proposed algorithm consists of two stages： flat-delay estimation and tapposition control with a constraint. The flat-delay estimation is carried out by estimating the significant dispersive region of the impulse response. The constrained tap-position control is achieved by imposing a limit on the new-tap-position search. Simulation results show that the proposed algorithm reduces the convergence speed by up to 85% over the conventional algorithms for a white signal input. For a colored signal, it also converges in 40% of the convergence time by the conventional algorithms. The proposed algorithm is applicable to adaptive FIR filters which are to model a path with long flat delay, such as echo cancelers for satellite-link communications.

A noise suppression algorithm with high speech quality based on weighted noise estimation and MMSE STSA is proposed. The proposed algorithm continuously updates the estimated noise by weighted noisy speech in accordance with an estimated SNR. The spectral gain is modified with the estimated SNR so that it can better utilize the improvement in noise estimation. With a better noise estimate, a more correct SNR is obtained resulting in the enhanced speech with low distortion. Subjective evaluation results show that five-grade mean opinion scores of the new algorithm with and without a speech codec are improved by as much as 0.35 and 0.40 respectively, compared with either the original MMSE STSA or the EVRC noise suppression algorithm.

This paper proposes a fast timing recovery method with a decision feedback equalizer for baudrate sampling. The proposed method features two special techniques. The first one is for coarse estimation of the sampling phase. Internal signals of the oversampled analog-to-digital converter at different phases are directly taken out for parallel evaluation. The second technique provides fine tuning with a phase-modification stepsize which is adaptively controlled by the residual intersymbol interference. Simulation results by a full-duplex digital transmission system with a multilevel line code show superiority of the proposed method. The coarse timing estimation and the fine tuning reduce 75% and 40% of the time required by the conventional method,respectively. The overall saving in timing recovery is almost 60% over the conventional method. The proposed method could easily be extended to other applications with a decision feedback equalizer.

This paper proposes a new subband adaptive filtering algorithm for adaptive FIR filters. The number of taps for each subband filter is adaptively controlled based on a sum of the absolute coefficients or the coefficient power in conjunction with the subband signal power. Keeping the total number of taps constant, redundant taps are redistributed to subbands where the number of taps is insufficient. Simulation results with a white signal show that the number of taps in each subband approaches an optimum as each subband filter converges. For a colored signal, tap assignment by the new algorithm is as stable as for a white signal.

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 new adaptation-mode control (AMC) for a robust adaptive microphone array with an adaptive blocking matrix (RAMA-ABM). The proposed AMC is based on cross correlations of two microphone signals and uses a state machine for controlling the adaptation to avoid target-signal cancellation. Evaluation with sound data obtained in different acoustic environments demonstrates that the noise reduction by the proposed AMC is 3 dB better than that by the AMC based on the SNR estimate. Subjective listening tests show that the quality of the output signal by the proposed AMC is comparable to or even better than those by the conventional AMCs.

A wind-noise suppressor with SNR based wind-noise detection and speech-wind discrimination is proposed. Wind-noise detection is performed in each frame and frequency based on the power ratio of the noisy speech and an estimated stationary noise. The detection result is modified by speech presence likelihood representing spectral smoothness to eliminate speech components. To suppress wind noise with little speech distortion, spectral gains are made smaller in the frame and the frequency where wind-noise is detected. Subjective evaluation results show that the 5-grade MOS for the proposed wind-noise suppressor reaches 3.4 and is 0.56 higher than that by a conventional noise suppressor with a statistically significant difference.

This paper proposes gain relaxation in signal enhancement designed for speech recognition. Gain relaxation 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. The softer enhancement is a solution to overlooked performance degradation in signal enhancement combined with speech recognition which is encountered in commercial products with an unaware small local noise source. 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.

A new stereo echo canceler with input slides and counter-lateralization is proposed. Convergence of filter coefficients to the correct echo paths is obtained by pre-processing which delays the input signal periodically by one sample in one of the two channels. The time difference between the two stereo components of the input signals causes a shift of the sound image. This shift is compensated for by presenting the delayed component of the stereo signals to a loudspeaker at a higher intensity, and the other component at a lower intensity. Correct echo-path identification is analytically shown in a more general form than in the preceding literatures. A subjective listening test shows that this method is more effective for vocal musics. The processed signals are scored 0.45 lower than the original input signals, using the ITU-R five-grade impairment scale.

This paper proposes a scalable multiecho cancellation system based on multiple autonomic and distributed echo canceler units. The proposed system does not have any common control section. Distributed control sections are equipped with in multiple echo cancelers operating autonomically. Necessary information is transferred from one unit to the next one. When the number of echoes to be canceled is changed, the necessary number of echo canceler units, each of which may be realized on a single chip, are simply plugged in or unplugged. The proposed system also provides fast convergence thanks to the novel coefficient location algorithm which consists of flat-delay estimation and constrained tap-position control. The input signal is evaluated at each tap to determine when to terminate flat-delay estimation. The number of exchanged taps is selected larger in flat-delay estimation than in constrained tap-position control. The convergence time with a colored-signal input is reduced by approximately 50% over STWQ, and 80% over full-tap NLMS algorithm. With a real speech input, the proposed system cancels the echo by about 20 dB. Tap-positions have been shown to be controlled correctly.

This paper proposes an adaptive noise canceller with low signal-distortion in the presence of crosstalk. The proposed noise canceller has two pairs of cross-coupled adaptive filters, each of which consists of the main filter and a sub filter. The signal-to-noise ratios (SNRs) of the primary and the reference signals are estimated by the sub filters. To reduce signal distortion at the output of the adaptive noise canceller, the step sizes for coefficient adaptation in the main filters are controlled according to the estimated SNRs. Computer simulation results show that the proposed noise canceller reduces signal distortion in the output signal by up to 15 dB compared to the conventional noise canceller.

This paper proposes a low complexity noise suppressor with hybrid filterbanks and adaptive time-frequency tiling. An analysis hybrid filterbank provides efficient transformation by further decomposing low-frequency bins after a coarse transformation with a short frame size. A synthesis hybrid filterbank also reduces computational complexity in a similar fashion to the analysis hybrid filterbank. Adaptive time-frequency tiling reduces the number of spectral gain calculations. It adaptively generates tiling information in the time-frequency plane based on the signal characteristics. The average number of instructions on a typical DSP chip has been reduced by 30% to 7.5 MIPS in case of mono signals sampled at 44.1 kHz. A Subjective test result shows that the sound quality of the proposed method is comparable to that of the conventional one.

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 filters (CCAFs). The CCAFs, whose common input signal is the output of a fixed beamformer, minimize leakage of the target signal into the interference path of the GSC. Each coefficient of the CCAFs is constrained to avoid mistracking. In the multiple-input canceller, leaky adaptive filters 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, when designed to allow about 20of look-direction error, can suppress interference by more than 17 dB.

This paper proposes a directional noise suppressor with a specified constant beamwidth for directional interferences and diffuse noise. A directional gain is calculated based on interchannel phase difference and combined with a spectral gain commonly used in single-channel noise suppressors. The beamwidth can be specified as passband edges of the directional gain. In order to implement frequency-independent constant beamwidth, frequency-proportionate directional gains are defined for different frequencies as a constraint. 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 24.4dB and 0.3 over a conventional noise suppressor. In a speech recognition scenario, the proposed directional noise suppressor outperforms both the conventional nondirectional noise suppressor and the conventional directional noise suppressor based on phase based T/F filtering with a negligible degradation in the word error rate for clean speech.

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 (MFs) and subfilters (SFs) connected in parallel thereto. To reduce signal-distortion in the output, the stepsizes for coefficient adaptation in the MFs are controlled according to estimated signal-to-noise ratios (SNRs) of the input signals. This SNR estimation is carried out using SF output signals. The stepsizes in the SFs are determined based on the ratio of the primary and the reference input signals to cope with a wider range of SNRs. 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 recognition rate as in the noise-free condition is achieved.

This paper proposes a modified normalized LMS algorithm based on a long-term average of the reference input signal power. The reference input signal power for normalization is estimated by using two leaky integrators with a short and a long time constants. Computer simulation results compare the performance of the proposed algorithm with some previosuly proposed adaptive-step algorithms. The proposed algorithm converges faster than the conventional adaptive-step algorithms. Almost 30dB of the ERLE (Echo Return Loss Enhancement), which is comparable to the conventional algorithms, is achieved in noisy environments.

An automatic tap assignment method in sub-band adaptive filter is proposed in this letter. The number of taps of the adaptive filter in each band is controlled by the mean-squared error. The numbers of taps increase in the bands which have large errors, while they decrease in the bands having small errors, until residual errors in all the bands become the same. In this way, the number of taps in a band is roughly proportional to the length of the impulse response of the unknown system in this band. The convergence rate and the residual error are improved, in comparison with existing uniform tap assignment. Effectiveness of the proposed method has been confirmed through computer simulation.

This paper proposes near-field sound-source localization based on crosscorrelation of a signed binary code. The signed binary code eliminates multibit signal processing for simpler implementation. Explicit formulae with near-field assumption are derived for a two microphone scenario and extended to a three microphone case with front-rear discrimination. Adaptive threshold for enabling and disabling source localization is developed for robustness in noisy environment. The proposed sound-source localization algorithm is implemented on a fixed-point DSP. Evaluation results in a robot scenario demonstrate that near-field assumption and front-rear discrimination provides almost 40% improvement in DOA estimation. A correct detection rate of 85% is obtained by a robot in a home environment.