Echo Path Research Articles

Stereophonic acoustic echo cancellation based on semi-blind source separation

The non-uniqueness problem embedded in stereophonic acoustic echo cancellation causes performance degradation. Usually, an additional decorrelation preprocessor is required, which is inevitably detrimental to the speech quality. In this paper, we propose a stereophonic acoustic echo cancellation method based on semi-blind source separation. By optimizing a dedicated demixing matrix rooted from stereophonic acoustic echo cancellation, the proposed method is insensitive to the inter-channel correlation between the far-end signals, resulting in better convergence behavior without any decorrelation preprocessing. Experimental results demonstrate the efficacy and robustness of the proposed method in the presence of inter-channel correlation, echo path changes, and double-talk.

DsPIC Implementation and Performances Evaluation of Adaptive Filters for System Identification

This paper aims to implement and evaluate performances of adaptive filters applied to linear systems identification on a dsPIC platform. To do so, we considered two different applications: the identification of DC motor transfer function and the acoustic echo cancellation. For the DC motor case, real-time identification is not necessarily required. However, for the acoustic echo cancellation case, real-time identification is crucial, making execution time a key parameter in this scenario. We have implemented three adaptive algorithms: LMS, NLMS, and RLS. The first two are known for their computational simplicity, while RLS is recognized for its convergence speed. Performances evaluation is primarily based on accuracy and computation speed. A comparison is carried out in both cases to determine which algorithm is more suitable for each application. The results obtained showed that RLS is better suited for the DC motor case, while NLMS is more suitable for acoustic echo cancellation, particularly when the impulse response of the echo path is long.

Low-Complexity Data-Reuse RLS Algorithm for Stereophonic Acoustic Echo Cancellation

Stereophonic audio devices employ two loudspeakers and two microphones in order to create a bidirectional sound effect. In this context, the stereophonic acoustic echo cancellation (SAEC) setup requires the estimation of four echo paths, each one corresponding to a loudspeaker-to-microphone pair. The widely linear (WL) model was proposed in recent literature in order to simplify the handling of the SAEC mathematical model. Moreover, low complexity recursive least- squares (RLS) adaptive algorithms were developed within the WL framework and successfully tested for SAEC scenarios. This paper proposes to apply a data-reuse (DR) approach for the combination between the RLS algorithm and the dichotomous coordinate descent (DCD) iterative method. The resulting WL-DR-RLS-DCD algorithm inherits the convergence properties of the RLS family and requires an amount of mathematical operations attractive for practical implementations. Simulation results show that the DR approach improves the tracking capabilities of the RLS-DCD algorithm, with an acceptable surplus in terms of computational workload.

Coefficients-Switched Normalized Least-Mean- Squares Adaption in Echo Canceler of Sparse-Echo-Path

The Normalized Least-Mean-Squares (NLMS) algorithm commonly used in echo cancelers suffers from a number of limitations due to the existing sparsity in the echo path. Although some sparsity-aware algorithms have been proposed, the high computational complexity greatly increases the their difficulty of implementation. To reduce the complexity and improve the performance in echo cancelers, we introduce a Coefficients-Switched (CS) processing into the NLMS algorithm and propose the CS-NLMS algorithm and the Mean-square-deviation (MSD) analysis to evaluate its effectiveness. Simulations of the MSD were performed with different system parameters. The performance of the proposed algorithm is verified by comparison with the simulation results of the previously proposed algorithms. The superiority of the proposed algorithm in terms of computational complexity is also discussed.

Neural Multi-Channel and Multi-Microphone Acoustic Echo Cancellation

Deep learning is introduced in multi-channel (MC) and multi-microphone (MM) acoustic echo cancellation (AEC) without decorrelation to the loudspeaker signals and achieves remarkable performance. In this paper, we propose a complex spectral mapping framework with inplace convolution and frequency-wise temporal modeling for MCAEC problem, which efficiently models the echo paths and spatial information. The proposed method is a multi-input and multi-output (MIMO) scheme, which filters out echoes from all microphone signals simultaneously, so the computational cost is greatly reduced. In addition, a cross-domain loss function with a multi-task learning strategy is designed for better generalization capability. Experiments are conducted on various unmatched scenarios and results show that the proposed method significantly outperforms previous methods. Moreover, a lightweight version of the proposed model with 0.29 million trainable parameters also shows good performance, which is essential for resource-limited and real-time applications.

Joint Online Estimation of Early and Late Residual Echo PSD for Residual Echo Suppression

In hands-free telephony and other distant-talking applications, an acoustic echo cancellation system is typically required, where a short adaptive filter is often used in practice to achieve fast convergence at low computational cost. This may result in late residual echo (LRE) remaining due to under-modeling of the echo path and early residual echo (ERE) due to filter misalignment. Both residual echo components can be suppressed using a postfilter in the subband domain, which requires accurate estimates of the power spectral density (PSD) of the ERE and LRE components. State-of-the-art methods estimate the ERE and LRE PSDs independently of each other, where the ERE PSD is estimated by simply multiplying the loudspeaker PSD with a frequency-dependent scalar and the LRE PSD is estimated using a recursive estimator based on frequency-dependent reverberation scaling and decay parameters. In this paper, we propose to extend the ERE PSD estimator from a scalar to a moving average filter on the loudspeaker PSD. In addition, we propose a signal-based method to jointly estimate all model parameters for the ERE and LRE PSD estimators in online mode, and derive two gradient-descent-based algorithms to simultaneously update the model parameters by minimizing the mean squared log error. The proposed method is compared with state-of-the-art methods in terms of estimation accuracy of the model parameters as well as the residual echo PSDs. Simulation results using both artificially generated as well as measured impulse responses show that the proposed method outperforms state-of-the-art methods for all considered scenarios.

Dynamic Neural Network-based Solutions for Acoustic Echo Suppression

Adaptive filters are widely used as core algorithms in current acoustic echo cancellation (AEC) systems. Echo path estimation is carried out by using different adaptive strategies in the adaptive filter. When nonlinear echo occurs, the performance deteriorates and seriously affects the call quality of both terminals. Taking advantage of the excellent non-linear fitting ability of neural networks in this paper, dynamic neural networks with self-updating parameters work continuously during the inference stage. The algorithm was computed and evaluated using publicly available echo audio data, showing that the dynamic neural network performs approximately as well as the optimal algorithm in the linear echo environment, and outperforms existing algorithms in the non-linear echo environment.

Exploiting Ultra-Wideband Channel Impulse Responses for Device-Free Localization.

In radio-frequency (RF)-based device-free localization (DFL), the number of sensors acting as RF transmitters and receivers is crucial for accuracy and system costs. Two promising approaches for DFL have been identified in the past: radio tomographic imaging (RTI) and multi-static radar (MSR). RTI in its basic version requires many sensors for high accuracy, which increases the cost. In this paper, we show how RTI benefits from multipath propagation. By evaluating the direct and echo paths, we increase the coverage of the target area, and by utilizing UWB signals, the RTI system is less susceptible to multipath propagation. MSR maps reflections that occur within the target area to reflectors such as persons or other objects. MSR does not require that the person is located near a signal path. Both suggested methods exploit ultra-wideband (UWB) channel impulse response (CIR) measurements. CIR measurements and the modeling of multipath effects either increase the accuracy or reduce the required number of sensors for localization with RTI. We created a test setup and measure UWB CIRs at different positions with a commercially available off-the-shelf UWB radio chip, the Decawave DW1000. We compare the localization results of RTI, multipath-assisted (MA)-RTI, and MSR and investigate a combined approach. We show that RTI is improved by the analysis of multipath propagation; furthermore, MA-RTI results in a better performance compared to MSR: with 50% of all cases, the localization error is better than 0.82 m and in 80% of all cases 1.34 m. The combined approach results in the best localization result with 0.64 m in 50% of all cases.

Using 3D Ray Tracing Technology to Study the Disturbance Effect of Rocket Plume on Ionosphere

In this paper, the initial neutral atmospheric parameters, background ionospheric parameters and geomagnetic field parameters of the ionosphere are obtained by NRLMSISE-00 model, IRI-2016 model and IGRF-13 model, respectively. Considering the neutral gas diffusion process, ion chemical reaction and plasma diffusion process, a three-dimensional dynamic model of chemical substances released by rocket plume disturbing the ionosphere is constructed. The influence of the disturbance on the echo path of high frequency radio waves with different incident frequencies is simulated by using three-dimensional digital ray-tracing technology. Using this model, the process of ionospheric disturbance caused by the main chemical substances H2 and H2O in the rocket plume under three different release conditions: fixed-point release at 300 km, vertical path at 250–350 km and parabolic path at 250–350 km, and the influence of the ionospheric cavity on the radio wave propagation of high frequency radio waves at different frequencies are simulated. The main purpose of the article is to focus on the effect of the cavity generated by the rocket exhaust on the propagation of radio waves. It mainly studies the perturbation effect on the ionosphere under different release conditions, considers the neutral gas diffusion process, ion chemical reaction and plasma diffusion process, and establishes the three-dimensional dynamics of the ionospheric electron density and the spatiotemporal distribution of the plume plasma learning model. Finally, the three-dimensional ray-tracing algorithm is used to simulate the propagation path of the radio wave through the disturbance area. We considered three different release conditions, including fixed-point release, vertical path and parabolic path. The ionospheric disturbances produced by these different releases are compared and analyzed, and their effects on the propagation path of radio waves are studied.

Selective Partial Update Adaptive Filtering Algorithms for Block-Sparse System Identification

The block-sparse normalized least mean square (BS-NLMS) algorithm which takes advantage of sparsity, successfully shows fast convergence in adaptive block-sparse system identification, adaptive control, and other industrial informatics applications. It is also attractive in acoustic processing where long impulse response, highly correlated and sparse echo path are encountered. However, the major drawback of BS-NLMS is largely computational complexity. This paper proposes a novel selective partial-update block-sparse normalized least mean square (SPU-BS-NLMS) algorithm. Compared with conventional BS-NLMS for block-sparse system identification, the proposed elective partial-update block-sparse NLMS algorithm takes partial-update blocks scheme which is determined by the smallest squared Euclidean-norm at each iteration instead of entire block coefficients to save computations. Computational complexity analysis is conducted to help researchers select appropriate parameters for practical realizations and applications. Computer simulations on acoustic echo cancellation are conducted to verify the results and the effectiveness of the proposed algorithm.

Non-Line-of-Sight Moving Target Detection Method Based on Noise Suppression

At the present time, most of existing security systems only detect and track targets in line-of-sight (LOS). However, in practice, the locations of targets are often out of the line of sight. This article focuses on the non-line-of-sight (NLOS) moving target detection with low-power transmission signals by reflection. There are two key problems, the weak target echo signal and the multipath effect. In terms of the issues, this paper constructs the echo signal model of the NLOS target. On the basis of the echo model, the detection method of NLOS moving target based on millimeter-wave radar comes up, which is of great theoretical value and important practical significance for indoor security. This paper innovatively applies the polynomial fitting method to suppress static noise and range gating method to suppress noise from other range gates. Then, the location and velocity of the target are estimated by two-dimensional fast Fourier transform (FFT) and the multiple signal classification (MUSIC) method. Furthermore, in order to verify the accuracy of the NLOS target echo signal model proposed in this paper, we respectively simulated two important parts of the signal in the model, the target echo signal and the direct backscattered signal of the intermediate interface, both of which are multipath signals. We counted the echo path length distribution in these two parts, and applied the NLOS target detection method to process them respectively. In addition, we also simulated the NLOS target echo signal and obtained actual data in the actual scene, and processed both the simulated data and the actual data. Comparing the results of target detection with and without denoising methods, the effectiveness of the two denoising methods proposed in this paper is verified.

Time-Variance Aware Dynamic Kernel Generation for Real-Time Acoustic Echo Cancellation

Time-variant factors including dynamic delay and varying echo path often occur in real-world acoustic echo cancellation (AEC) applications. Current end-to-end deep neural network (DNN) based methods usually model the time-variant components implicitly and can hardly handle the unpredictable time-variance in real-time AEC. To explicitly capture the time-variant components, we propose a dynamic kernel generation (DKG) module that can be introduced as a learnable plug-in to a DNN-based end-to-end pipeline. Specifically, the DKG module generates a convolutional kernel regarding to each input audio frame, so that the DNN model is able to dynamically adjust its weights according to the input signal during inference. Experimental results verify that DKG module improves the AEC performance of the model under time-variant scenarios, especially in the double-talk cases.

Robust Proportionate Normalized Least Mean M-Estimate Algorithm for Block-Sparse System Identification

In practical applications, the impulse responses (IRs) of some network echo paths are blocksparse (BS), while the traditional proportionate and zero attraction algorithms do not consider the prior sparsity of the BS system, so they do not perform well in the block-sparse system identification (BSSI). In addition, most of the current BS filtering algorithms are based on the assumption of Gaussian noise, so the performance will deteriorate seriously in the background of impulse noise. To overcome the shortcoming, we use the mixed <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$l_{2,1} $ </tex-math></inline-formula> norm of the filter weight vector to fully tap the sparsity of the BS system, and combine the anti impulse noise characteristic of the M-estimate function to design and derive the BS proportionate normalized least mean M-estimate (BSPNLMM) algorithm from the perspective of basis pursuit (BP), which well realizes the BSSI in the presence of impulse noise. Then, we analyze the mean performance of the BSPNLMM algorithm in detail and give the stable step size bound. Finally, the superiority of the proposed BSPNLMM algorithm is verified by numerical simulations

Nonlinear stereophonic acoustic echo cancellation using sub-filter based adaptive algorithm

The nonlinear echo path resulting from the unwarranted acoustic echo and nonlinear distortion will degrade the echo cancellation performance of a stereophonic acoustic echo cancellation system. Techniques proposed in the literature rely on the assumption of a linear echo path. In this paper, a nonlinear stereophonic acoustic echo cancellation framework is proposed to address the problem of nonlinear distortion in stereophonic acoustic cancellation systems with a better trade-off between convergence and steady-state performance. The proposed approach uses the functional link-based adaptive filtering to deal with the nonlinear distortion in combination with the sub-filter approach to enhance the convergence performance. In addition to that, the convergence analysis of the proposed algorithm is presented in this paper. Simulations carried out using echo return loss enhancement and magnitude squared coherence as performance metrics using speech and colored noise input demonstrate the proposed framework's ability to improve the echo cancellation performance in the presence of distortion.

High-performance and ultra-compact spike-based architecture for real-time acoustic echo cancellation

In recent years, the implementation of advanced adaptive filters in embedded devices for acoustic echo cancellation has increased because most of them are used in portable devices, especially in Internet of Things systems, in which high-performance and low-area digital hardware implementations are required. In this work, we present a high-performance and ultra-compact spike-based hardware architecture to efficiently compute adaptive filters to be used as an acoustic echo canceller (AEC). To achieve this architecture, we address two factors. (1) We propose a new data-selective adaptive filter along with subband decomposition least mean square (LMS) method to reduce the computational cost by minimizing the number of operations required to efficiently update the filter coefficients. The proposed method requires approximately 40% fewer updates when compared with conventional subband adaptive filters. As a consequence, the spike-based hardware architecture simulates the proposed adaptive filter at high processing speeds. (2) We propose a compact and high-precision neural multiplier since acoustic echo cancellers require a large number of high-precision multiplications to efficiently identify the echo path. The proposed neural multiplier expends up to 15 times fewer synapses when compared with existing approaches, which represents a significant improvement in terms of area. In addition, this improvement avoids routing problems by implementing large-scale synapse connectivity in advanced FPGAs. Herein, we carry out exhaustive testing by simulating several acoustic echo scenarios. The results demonstrate that the features of the model and the hardware implementation techniques potentially allow easy integration into portable devices for use in real-world acoustic echo cancellation scenarios.

A package for the automated classification of images containing supernova light echoes

Context.The so-called light echoes of supernovae – the apparent motion of outburst-illuminated interstellar dust – can be detected in astronomical difference images; however, light echoes are extremely rare which makes manual detection an arduous task. Surveys for centuries-old supernova light echoes can involve hundreds of pointings of wide-field imagers wherein the subimages from each CCD amplifier require examination.Aims.We introduce ALED, a Python package that implements (i) a capsule network trained to automatically identify images with a high probability of containing at least one supernova light echo and (ii) routing path visualization to localize light echoes and/or light echo-like features in the identified images.Methods.We compared the performance of the capsule network implemented in ALED (ALED-m) to several capsule and convolutional neural networks of different architectures. We also applied ALED to a large catalogue of astronomical difference images and manually inspected candidate light echo images for human verification.Results.ALED-m was found to achieve 90% classification accuracy on the test set and to precisely localize the identified light echoes via routing path visualization. From a set of 13 000+ astronomical difference images, ALED identified a set of light echoes that had been overlooked in manual classification.

Deep learning-based stereophonic acoustic echo suppression without decorrelation.

Traditional stereophonic acoustic echo cancellation algorithms need to estimate acoustic echo paths from stereo loudspeakers to a microphone, which often suffers from the nonuniqueness problem caused by a high correlation between the two far-end signals of these stereo loudspeakers. Many decorrelation methods have already been proposed to mitigate this problem. However, these methods may reduce the audio quality and/or stereophonic spatial perception. This paper proposes to use a convolutional recurrent network (CRN) to suppress the stereophonic echo components by estimating a nonlinear gain, which is then multiplied by the complex spectrum of the microphone signal to obtain the estimated near-end speech without a decorrelation procedure. The CRN includes an encoder-decoder module and two-layer gated recurrent network module, which can take advantage of the feature extraction capability of the convolutional neural networks and temporal modeling capability of recurrent neural networks simultaneously. The magnitude spectra of the two far-end signals are used as input features directly without any decorrelation preprocessing and, thus, both the audio quality and stereophonic spatial perception can be maintained. The experimental results in both the simulated and real acoustic environments show that the proposed algorithm outperforms traditional algorithms such as the normalized least-mean square and Wiener algorithms, especially in situations of low signal-to-echo ratio and high reverberation time RT60.

Combined functional link adaptive filter for nonlinear acoustic echo cancellation

The usage of low-quality components in communicating devices may introduce nonlinearity in the audio signals. The degree of nonlinearity may vary over time due to the nonstationary nature of audio signals. In the presence of nonlinear audio signals, the performance of an acoustic echo cancellation (AEC) scheme is impacted by two major limitations. The first one is related to the modeling accuracy, and the second one is related to the adaptation performance. The issue of modeling accuracy exists because the available nonlinear AEC (NAEC) schemes consider the nonlinearity as time-invariant and employ a fixed architecture to model the echo path. The adaptation performance issue occurs due to the conflicting requirements of fast convergence and low steady-state error. There is a need for an architecture that not only models the nonlinearity efficiently but also provides better performance in terms of convergence. In this paper, three new architectures convex combination functional link adaptive filter (CC-FLAF), improved CC-FLAF (ICC-FLAF), and variable step size CC-FLAF (VSSCC-FLAF) are proposed to overcome the above limitations. The proposed architectures are based on the combination of adaptive filters and variable step size approach. Computer simulations demonstrate the effectiveness of proposed architectures, in improving the nonlinear modeling performance for NAEC applications.

A Collaborative Spline Adaptive Filter for Nonlinear Echo Cancellation

Acoustic echo causes the quality of communication to degenerate and results in loss of clarity. Though conventional linear adaptive filters have been applied successfully to eliminate linear acoustic echo, they can barely deal with the problem of nonlinear acoustic echo. In this paper, a collaborative spline adaptive filter is presented to eliminate nonlinear acoustic echo. The filter is composed of a linear adaptive filter in the upper branch and a nonlinear spline adaptive filter in the lower branch. The nonlinear one is a Hammerstein system consisting of a spline interpolation function and a subsequent linear adaptive filter. The two branches are collaboratively combined and a mixing parameter is adopted, which can be updated to adjust the proportion of the lower branch output signal. Experimental results show that the presented method can achieve a good performance in nonlinear echo cancelation regardless of whether the nonlinear degree of the echo path varies with time or not.

Optimum step-size control for a variable step-size stereo acoustic echo canceller in the frequency domain

The frequency-domain adaptive filter (FDAF) is a competitive candidate for stereo acoustic echo cancellation (SAEC) because of its good convergence and computational efficiency. However, there is a conflict between the convergence rate and the steady-state misalignment when using a constant step size. In this paper, a variable step-size approach is proposed for the stereo echo canceller in the frequency domain. The proposed method uses a first-order Markov model to describe the time-varying echo paths between the loudspeakers and the microphone. Following a statistical convergence analysis of the two-channel FDAF algorithm, the state recursion of the system distance in each frequency bin is given. The optimal step size is then derived by minimizing the system distance. Several practical problems, including noise power spectral density estimation and echo paths change detection, are discussed in details. Moreover, a close link between the proposed optimum step-size control method and the multichannel state–space frequency-domain adaptive filter is established. We show that the proposed variable step-size approach can also be incorporated into the block extended least-mean-square algorithm to fully exploit their benefits. Finally, computer simulations demonstrate that the proposed algorithm exhibits an excellent convergence performance and is very robust to the double talk.