Echo Suppression Research Articles

Multichannel Dynamic Sound Rendering and Echo Suppression in a Room Using Wave Field Synthesis

With multichannel sound rendering systems, an immersive sound can be experienced by a broader audience. Rendering of a stationary source is widely studied in the literature. However, a dynamic sound, where the source’s position is continuously changing, provides improved source localization and a listener experience more immersed in the rendered sound. This paper investigates the multichannel sound rendering of a dynamic sound source using a wave field synthesis method in a reverberant shoe-box room model. Finite difference time domain is employed to visualize the reproduced sound field in the time domain. To address the challenge of room echo-induced degradation in sound quality, a method is presented in which the reflected sound is compensated by rendering the primary source’s images from the auxiliary actuators, enhancing SNR by approximately 2–3.5 dB.

On Efficient Echo Suppression With Phaseless Data in the Joint Frequency and Spatial Domain for Antenna Pattern Measurement in a Nonanechoic Chamber

On Efficient Echo Suppression With Phaseless Data in the Joint Frequency and Spatial Domain for Antenna Pattern Measurement in a Nonanechoic Chamber

Double Branches and Stages Neural Network for Joint Acoustic Echo and Noise Suppression

Double Branches and Stages Neural Network for Joint Acoustic Echo and Noise Suppression

Single-Frequency Phaseless Data-Based Echo Suppression for Antenna Pattern Measurement in a Nonideal Chamber

Single-Frequency Phaseless Data-Based Echo Suppression for Antenna Pattern Measurement in a Nonideal Chamber

Full-duplex acoustic interaction system for cognitive experiments with cetaceans

Abstract Cetaceans show high cognitive abilities and strong social bonds. Their primary sensory modality to communicate and sense the environment is acoustics. Research on their echolocation and social vocalizations typically uses visual and tactile systems adapted from research on primates or birds. Such research would benefit from a purely acoustic communication system to better match their natural capabilities. We argue that a full duplex system, in which signals can flow in both directions simultaneously is essential for communication research. We designed and implemented a full duplex system to acoustically interact with cetaceans in the wild, featuring digital echo-suppression. We pilot tested the system in Arctic Norway and achieved an echo suppression of 18 dB. We discuss the limiting factors and how to improve the echo suppression further. The system enabled vocal interaction with the underwater acoustic scene by allowing experimenters to listen while producing sounds. We describe our motivations, then present our pilot deployment and give examples of initial explorative attempts to vocally interact with wild orcas and humpback whales.

Polarised full-waveform warning LIDAR with dust backscattering suppression

LIDAR, when used under adverse weather conditions such as fog and dust, suffers from attenuation and backscattering. In particular, backscattering creates spurious echo peaks in the received waveforms at incorrected ranges and leads to false alarms. As a solution, this paper proposes polarised full-waveform warning LIDAR for dust backscattering echo suppression, intended for applications under dusty environments. To this end, first, a theoretical model of the polarised LIDAR is developed based on the LIDAR equation and depolarisation characteristics of dust backscattering and target reflection. Thereafter, the depolarisation characteristics are experimentally evaluated under an artificially generated dusty environment; the results reveal that the target reflection features a strong depolarisation characteristic, whereas dust backscattering has a weak one. When the LIDAR emits linearly polarised light, a significant orthogonal-polarised component exists owing to the target reflection, and dust backscattering induces a mainly parallel-polarised component. On interposing an orthogonal-polarisation analyser in the receiving optical path, the parallel-polarised component from the dust backscattering and target reflection is blocked, whereas the orthogonal-polarised component is allowed to pass; this helps improve the contrast of the target reflection to dust backscattering. Finally, based on the theoretical model of the polarised LIDAR and experimental results, an orthogonal-polarised LIDAR prototype is designed and fabricated. Using this prototype, target detection experiments in a dusty environment are performed; the corresponding experimental results indicate that the orthogonal-polarised LIDAR effectively suppresses dust backscattering echoes and that the contrast of the target reflection to dust backscattering is improved by 4.78–7.12 dB. Utilization of the polarised full-waveform warning LIDAR can significantly reduce the false alarm rate and improves the LIDAR system performance when used under dusty environment conditions, in applications such as smart mining and autonomous driving.

Method for Digital Cancellation of System Interference in a Full-Duplex Power Line Communication System

Every year the number of IoT devices is growing, which gives a boost to development of such technology as Power Line Communication (PLC). The term PLC refers to the use of a power supply circuit to transmit data. Full-duplex technology can be applied to increase the data transfer rate in a PLC. The difficulty in using a full duplex is the need to suppress the transmitter’s own signal (interference signal). Analog cancellation and digital cancellation are used for this purpose. Common algorithms for digital cancellation are adaptive algorithms. In this paper adaptive algorithms such as Particle Swarm Optimization (PSO) and Recursive Least Squares (RLS) are considered and a modification of the classical RLS is proposed. The proposed algorithm uses precalculated weight coefficients together with periodic reinitialization of the filter. This avoids the influence of significant changes in the input signal on the further process of filter adaptation and provides the highest level of echo suppression. The proposed algorithm provides an improvement of parameters such as Error Vector Magnitude (EVM) and Signal-to-Interference Ratio (SIR) in comparison with classical RLS and PSO, according to the results of the study.

Ultrashort Echo Time (UTE) MRI porosity index (PI) and suppression ratio (SR) correlate with the cortical bone microstructural and mechanical properties: Ex vivo study.

Ultrashort echo time (UTE) MRI can image and consequently enable quantitative assessment of cortical bone. UTE-MRI-based evaluation of bone is largely underutilized due to the high cost and time demands of MRI in general. The signal ratio in dual-echo UTE imaging, known as porosity index (PI), as well as the signal ratio between UTE and inversion recovery UTE (IR-UTE) imaging, known as the suppression ratio (SR), are two rapid UTE-based bone evaluation techniques (∼ 5mins scan time each), which can potentially reduce the time demand and cost in future clinical studies. This study aimed to investigate the correlations of PI and SR measures with cortical bone microstructural and mechanical properties. Cortical bone strips (n=135) from tibial and femoral midshafts of 37 donors (61±24years old) were scanned using a dual-echo 3D Cones UTE sequence and a 3D Cones IR-UTE sequence for PI and SR calculations, respectively. Average bone mineral density, porosity, and pore size were measured using microcomputed tomography (μCT). Bone mechanical properties were measured using 4-point bending tests. The μCT measures showed significant correlations with PI (moderate to strong, R=0.68-0.71) and SR (moderate, R=0.58-0.68). Young's modulus, yield stress, and ultimate stress demonstrated significant moderate correlations with PI and SR (R=0.52-0.62) while significant strong correlations with μCT measures (R>0.7). PI and SR can potentially serve as fast and noninvasive (non-ionizing radiation) biomarkers for evaluating cortical bone in various bone diseases.

Acoustic Echo Cancellation with the Normalized Sign-Error Least Mean Squares Algorithm and Deep Residual Echo Suppression

This paper presents an echo suppression system that combines a linear acoustic echo canceller (AEC) with a deep complex convolutional recurrent network (DCCRN) for residual echo suppression. The filter taps of the AEC are adjusted in subbands by using the normalized sign-error least mean squares (NSLMS) algorithm. The NSLMS is compared with the commonly-used normalized least mean squares (NLMS), and the combination of each with the proposed deep residual echo suppression model is studied. The utilization of a pre-trained deep-learning speech denoising model as an alternative to a residual echo suppressor (RES) is also studied. The results showed that the performance of the NSLMS is superior to that of the NLMS in all settings. With the NSLMS output, the proposed RES achieved better performance than the larger pre-trained speech denoiser model. More notably, the denoiser performed considerably better on the NSLMS output than on the NLMS output, and the performance gap was greater than the respective gap when employing the RES, indicating that the residual echo in the NSLMS output was more akin to noise than speech. Therefore, when little data is available to train an RES, a pre-trained speech denoiser is a viable alternative when employing the NSLMS for the preceding linear AEC.

Advances in magnetic resonance imaging for the assessment of paediatric focal epilepsy: a narrative review.

Epilepsy affects approximately 50 million people worldwide, with 30-40% of patients not responding to medication, necessitating alternative therapies such as surgical intervention. However, the accurate localization of epileptogenic lesions, particularly in pediatric magnetic resonance imaging (MRI)-negative drug-resistant epilepsy, remains a challenge. This paper reviews advanced neuroimaging techniques aimed at improving the detection of such lesions to enhance surgical outcomes. A comprehensive literature search was conducted using PubMed, focusing on advanced MRI sequences, focal epilepsy, and the integration of artificial intelligence (AI) in the diagnostic process. New MRI sequences, including magnetization prepared 2 rapid gradient echo (MP2RAGE), edge-enhancing gradient echo (EDGE), and fluid and white matter suppression (FLAWS), have demonstrated enhanced capabilities in detecting subtle epileptogenic lesions. Quantitative MRI techniques, notably magnetic resonance fingerprinting (MRF), alongside innovative post-processing methods, are emphasized for their effectiveness in delineating cortical malformations, whether used alone or in combination with ultra-high field MRI systems. Furthermore, the integration of AI in radiology is progressing, providing significant support in accurately localizing lesions, and potentially optimizing pre-surgical planning. While advanced neuroimaging and AI offer significant improvements in the diagnostic process for epilepsy, some challenges remain. These include long acquisition times, the need for extensive data analysis, and a lack of large, standardized datasets for AI validation. However, the future holds promise as research continues to integrate these technologies into clinical practice. These efforts will improve the clinical applicability and effectiveness of these advanced techniques in epilepsy management, paving the way for more accurate diagnoses and better patient outcomes.

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.

Double-Talk Detection-Aided Residual Echo Suppression via Spectrogram Masking and Refinement

Acoustic echo in full-duplex telecommunication systems is a common problem that may cause desired-speech quality degradation during double-talk periods. This problem is especially challenging in low signal-to-echo ratio (SER) scenarios, such as hands-free conversations over mobile phones when the loudspeaker volume is high. This paper proposes a two-stage deep-learning approach to residual echo suppression focused on the low SER scenario. The first stage consists of a speech spectrogram masking model integrated with a double-talk detector (DTD). The second stage consists of a spectrogram refinement model optimized for speech quality by minimizing a perceptual evaluation of speech quality (PESQ) related loss function. The proposed integration of DTD with the masking model outperforms several other configurations based on previous studies. We conduct an ablation study that shows the contribution of each part of the proposed system. We evaluate the proposed system in several SERs and demonstrate its efficiency in the challenging setting of a very low SER. Finally, the proposed approach outperforms competing methods in several residual echo suppression metrics. We conclude that the proposed system is well-suited for the task of low SER residual echo suppression.

Метод подавления акустического эха на основе рекуррентной нейронной сети и алгоритма кластеризации

The article solves the problem of acoustic echo suppression based on a neural network that evaluates an ideal binary mask IBM using features extracted from a mixture of near-end and far-end signals. The novelty of the proposed method lies in the use of the clustering algorithm in addition to the bidirectional recurrent neural network BLSTM. To evaluate the use of the EM, Mean-Shift, k-Means clustering algorithms, the models have been trained and tested on the TIMIT database. For each model, the ERLE, PESQ, STOI metrics have been calculated to characterize its quality. The use of the EM and Mean-Shift clustering algorithms appeared to be inefficient compared to the BLSTM algorithm at a signal-to-echo ratio of 10 dB. With a signal-to-echo ratio of 6 dB, BLSTM+Mean-Shift resulted in a marginal improvement in the PESQ metric compared to the BLSTM algorithm. The results of the experiments show the effectiveness of the proposed BLSTM model when using a network with the K-Means algorithm, compared to using a pure BLSTM for echo cancellation in double-talk scenarios. With a signal-to-echo ratio of 10 dB, the STOI metric, which characterizes speech intelligibility, has improved by 7%, and the PESQ metric, which characterizes the quality of speech restoration, by 18.8%.

Ultrashort Echo Time (UTE) MRI Porosity Index (PI) and Suppression Ratio (SR) Correlate with the Cortical Bone Microstructural and Mechanical Properties: Ex Vivo Study

Ultrashort Echo Time (UTE) MRI Porosity Index (PI) and Suppression Ratio (SR) Correlate with the Cortical Bone Microstructural and Mechanical Properties: Ex Vivo Study

Neural Cascade Architecture for Multi-Channel Acoustic Echo Suppression

Traditional acoustic echo cancellation (AEC) works by identifying an acoustic impulse response using adaptive algorithms. This paper proposes a neural cascade architecture for joint acoustic echo and noise suppression to address both single-channel and multi-channel AEC (MCAEC) problems. The proposed cascade architecture consists of two modules. A convolutional recurrent network (CRN) is employed in the first module for complex spectral mapping. Its output is fed as an additional input to the second module, where a long short-term memory network (LSTM) is utilized for magnitude mask estimation. The entire architecture is trained in an end-to-end manner with the two modules optimized jointly using a single loss function. The final output is generated using the enhanced phase and magnitude obtained from the first and the second module, respectively. The cascade architecture enables the proposed method to obtain robust magnitude estimation as well as phase enhancement. The proposed method is investigated under different AEC setups. We find that the deep learning based approach avoids the no-uniqueness problem in traditional MCAEC. For MCAEC setups with multiple microphones, combining deep MCAEC with supervised beamforming further improves the system performance. Evaluation results show that the proposed approach effectively suppresses acoustic echo and noise while preserving speech quality, and consistently outperforms related methods under different setups.

Experimental Demonstration of Nadir Echo Removal in SAR Using Waveform Diversity and Dual-Focus Postprocessing

Synthetic aperture radar (SAR) provides high-resolution images for remote-sensing applications regardless of sunlight and weather conditions. The pulsed operation of SAR may lead to an occurrence of nadir echoes in SAR images that significantly affect the image quality in case the pulse repetition frequency (PRF), which is not properly constrained within the SAR system design. As an alternative, pulse-to-pulse variation of the transmitted waveform and dual-focus postprocessing can be exploited to remove the nadir echo and alleviate the PRF constraints (also in ScanSAR operation). This work provides a demonstration of the latter concept through an experimental acquisition of the TerraSAR-X satellite. The experiment is designed by selecting the scene and the acquisition parameters in order to have the nadir echo appearing in the SAR image. The waveform variation is achieved by alternating up- and down-chirps on transmit. The analysis of the results shows the effectiveness of dual-focus postprocessing for nadir echo suppression.

Processing Techniques for Nadir Echo Suppression in Staggered Synthetic Aperture Radar

Synthetic aperture radar (SAR) is a class of high-resolution imaging radar particularly suitable for satellite remote sensing with diverse applications, such as biomass and ice monitoring, generation of digital elevation models, and measuring of subsidence. Staggered SAR is a novel mode of operation under consideration for the next-generation SAR missions such as Tandem-L and NASA-ISRO SAR. It uses digital beamforming and a continuous variation of the pulse repetition interval (PRI) to achieve high azimuth resolution over a much wider, continuous swath than is traditionally possible. This PRI variation renders infeasible use of existing methods to avoid nadir echoes, which might impair the quality of staggered SAR images. This work proposes processing techniques that mitigate the impact of nadir echoes in staggered SAR through localization and thresholding-and-blanking of these echoes in range-compressed data and recovery of part of the underlying useful signal through interpolation. The performance of these processing techniques is evaluated through simulations using real TerraSAR-X data. The proposed technique can be implemented as an optional stage in the processing chain of future staggered SAR missions and leads to improved image quality at a reasonable additional computational cost.

Nonlinear residual echo suppression based on dual-stream DPRNN

The acoustic echo cannot be entirely removed by linear adaptive filters due to the nonlinear relationship between the echo and the far-end signal. Usually, a post-processing module is required to further suppress the echo. In this paper, we propose a residual echo suppression method based on the modification of dual-path recurrent neural network (DPRNN) to improve the quality of speech communication. Both the residual signal and the auxiliary signal, the far-end signal or the output of the adaptive filter, obtained from the linear acoustic echo cancelation are adopted to form a dual-stream for the DPRNN. We validate the efficacy of the proposed method in the notoriously difficult double-talk situations and discuss the impact of different auxiliary signals on performance. We also compare the performance of the time domain and the time-frequency domain processing. Furthermore, we propose an efficient and applicable way to deploy our method to off-the-shelf loudspeakers by fine-tuning the pre-trained model with little recorded-echo data.

A Waveform-Encoded SAR Implementation Using a Limited Number of Cyclically Shifted Chirps

Synthetic aperture radar (SAR) provides high-resolution images of the Earth’s surface irrespective of sunlight and weather conditions. In conventional spaceborne SAR, nadir echoes caused by the pulsed operation of SAR may significantly affect the SAR image quality. Therefore, the pulse repetition frequency (PRF) is constrained within the SAR system design to avoid the appearance of nadir echoes in the SAR image. As an alternative, the waveform-encoded SAR concept using a pulse-to-pulse variation of the transmitted waveform and dual-focus postprocessing can be exploited for nadir echo removal and to alleviate the PRF constraints. In particular, cyclically shifted chirps have been proposed as a possible waveform variation scheme. However, a large number of distinct waveforms is required to enable the simple implementation of the concept. This work proposes a technique based on the Eulerian circuit for generating a waveform sequence starting from a reduced number of distinct cyclically shifted chirps that can be effectively exploited for waveform-encoded SAR. The nadir echo suppression performance of the proposed scheme is analyzed through simulations using real TerraSAR-X data and a realistic nadir echo model that shows how the number of distinct waveforms and therefore the system complexity can be reduced without significant performance loss. These developments reduce the calibration burden and make the concept viable for implementation in future SAR systems.