Search

Adaptive close-talking differential microphone arrays (ACT-MAs) inherently suppress farfield noise while emphasizing desired nearfield signals. This paper discusses the applicability of ACT-MAs for noise reduction in mobile phones. In order to utilize the advantages of ACTMAs, we need to improve the robustness to microphone mismatch and improve parameter estimation accuracy. In this paper we propose a method to improve the robustness of the ACTMA algorithm by taking microphone gain mismatch into account in the detection of background noise and mobile phone user activity, performing online microphone gain calibration, steering the null of the ACTMA to the rear of the mobile phone, and performing parameter estimation only when mobile phone user activity is detected. Thus, the robust ACTMA is applicable for performing noise reduction in mobile phones. Experiments with recorded data demonstrate the effectiveness of this method.

Standard approaches for the cancellation of acoustic echoes in telecommunication systems assume that the echo path to be identified can be modeled by a linear system. However, in applications such as echo cancellation for mobile communication terminals, non-negligible nonlinear distortions are introduced by loudspeakers and their amplifiers, resulting in a significant degradation of the performance of purely linear approaches. In this contribution we consider so-called power filters in order to cope with these kinds of nonlinear echo paths. By combining time-domain orthogonalization methods with a DFT-domain implementation of power filters we derive corresponding quasi-complete orthogonalized versions. As the statistical properties of speech input are non-stationary, the orthogonalization must follow this time-variance, too. Furthermore, we introduce a step-size control for the adaptation of the equivalent orthogonalized structure. Experiments with real hardware show that, with the proposed nonlinear approach, an increase in echo attenuation can be achieved, if the loudspeaker system introduces saturation-type nonlinearities in the echo path.

Finite-length Volterra filters are known to be able to model a wide range of real world nonlinear systems. In this paper, we present an algorithm that allows for different memory lengths of the linear and quadratic Volterra kernel while preserving the advantages of fast convolution techniques in the frequency domain for a second-order Volterra filter. This is achieved by extending partitioned block methods to second-order Volterra filters. To obtain corresponding adaptive realizations of the proposed approach, we present generalizations of known frequency-domain algorithms for Volterra filters and partitioned block frequency-domain adaptive filters for linear systems, respectively. To exploit the advantages of adaptive frequency-domain algorithms with respect to convergence speed, we provide a frequency bin-wise normalization of the step-size parameters. To evaluate the performance of the proposed approach, simulation results are given for the application to nonlinear acoustic echo cancellation. The results confirm the improved convergence of a second-order partitioned block frequency-domain adaptive Volterra filter (PBFDAVF) compared with time-domain adaptation of the kernel coefficients.

Acoustic echo cancellation has traditionally employed basically all variants known from deterministic adaptive filter design, such as least mean-square (LMS), recursive least-squares (RLS), and frequency-domain adaptive filters (FDAF). More recently, a stochastic adaptive filter design based on the concept of acoustic state-space modeling of the echo path has been introduced to accommodate for an ever sought unification of adaptive filtering and adaptation control. The corresponding Kalman filter theory has been formulated for single-channel, multi-channel, and nonlinear echo cancellation problems. This paper closes an important gap by formulating the state-space model and the corresponding adaptive algorithm for the partitioned-block filtering structure which is especially relevant in practice. This structure allows for the use of significantly longer filter lengths in comparison to previous work, and for the flexible design and implementation of acoustic echo cancellers for widely differing acoustic conditions.

Modern systems for acoustic man-machine communication, such as teleconferencing equipment or speech dialog systems, will have to provide hands-free voice interfaces with full-duplex ability in order to allow for most natural interaction with users. Acoustic echo control is the enabling technology for suppressing the unavoidable feedback between acoustic reproduction and recording, i.e., the echo, such that an enhanced microphone signal is transmitted to a remote talker or supplied to a speech terminal. In this chapter, we first recall the established signal processing strategies for acoustic echo cancellation, mostly based on the family of deterministic least-squares adaptive filters. We then proceed with our main theme, which consists of the recent developments towards a more statistical framework for acoustic echo control. Our approach is guided by the motivation of taking various uncertainties regarding the echo generation process into account, while maintaining backwards compatibility for the inclusion of established technology. With unified notation, we present different optimum filtering solutions which were tailored for robustness against the specific uncertainties related to echo path variability, nonlinearity, and clock drift in a system. Furthermore, we provide a generalization of our concepts to systems with multiple reproduction channels, where algorithmic complexity and correlation of loudspeaker signals represents additional challenges.

The recording and reproduction of spatial audio gains increasing attention since multichannel sound systems have been established in modern audio applications. Directional Audio Coding (DirAC) represents an efficient description of spatial sound in terms of a few audio downmix signals and parametric side information in frequency sub-bands, namely the direction-of-arrival and diffuseness of the sound. Traditionally, the directional parameters are derived based on the active sound intensity vector, which can be efficiently estimated via two or threedimensional microphone grids. However, due to form factor constraints, linear microphone arrays are often more suitable in practice and thus, alternative parameter estimation techniques are required. In this contribution, a framework for parametric spatial audio processing based on linear microphone arrays is presented allowing for an efficient recording, enhancement, and reproduction of spatial sound. Different application examples are discussed including parametric directional filtering and de-reverberation.

The European research project “Together anywhere, together anytime” (TA2) explores how technology can make communication and engagement easier among groups of people separated in space and time, e.g. families living in different cities. It combines high quality communication with shared group activities using distributed multimedia applications and broadband technology in the living room. This paper provides an overview of the TA2-project with a focus on highdefinition audio communication, which has the goal of providing an experience “as if speaking to someone in the same room”. The most important aspects for making audio communication a natural experience are high audio bandwidth, robust transmission with low delay, hands-free operation, and spatial audio (surround sound) reproduction. These requirements are addressed by the Audio Communication Engine (ACE), which includes audio coding, acoustic processing and IP transport in a single software module. Based on concept demonstrators developed in the TA2- project, results from user experience tests are presented which indicate that more than 90% of users rate the audio quality as natural when using high-definition audio, compared to less than 10% when standard audio quality is used.

Directional Audio Coding (DirAC) represents an efficient scheme to analyze and reproduce spatial sound; the coded stream consists of a single-channel audio signal and few parameters. Estimation of these parameters ideally relies on figure-of-eight microphones and one omnidirectional microphone. The figure-of-eight directivity can be efficiently approximated by first-order differential microphone arrays. However, due to spatial sampling, there are considerable deviations from the required directivity at high frequencies. These deviations lead to incorrect spatial parameter estimates; especially, the instantaneous direction-of-arrival (DOA) becomes biased and - beyond a certain aliasing frequency - ambiguous. Parameters like the DOA drive subsequent processing units, e. g., for spatial filtering. In this paper we propose a novel strategy to compensate bias and ambiguity not directly w. r. t. the DOA but the output parameters of a spatial filtering processing unit. Simulation results confirm the benefits of the novel method, which does not involve additional computational load at run-time.

Directional Audio Coding (DirAC) is an efficient technique to capture and reproduce spatial sound. The analysis step outputs a mono DirAC stream, comprising an omnidirectional microphone pressure signal and side information, i.e., direction of arrival and diffuseness of the sound field expressed in time-frequency domain. This contribution proposes a method to merge two or more mono DirAC streams for a joint playback at the reproduction side. This problem arises in applications such as immersive spatial audio teleconferencing. With respect to a trivial direct merging, the proposed method is more efficient as it does not require the synthesis step. From this follows the benefit that the loudspeaker setup at the reproduction side does not have to be known in advance. Simulations and informal listening tests confirm the absence of any artifacts and that the proposed method is practically indistinguishable from the ideal merging.

Multichannel sound systems become more and more established in modern audio applications. Consequently, the recording and the reproduction of spatial audio gains increasing attention. Directional Audio Coding (DirAC) represents an efficient approach to analyze spatial sound and to reproduce it using arbitrary loudspeaker configurations. In DirAC, the direction-of-arrival and the diffuseness of sound within frequency subbands is used to encode the spatial properties of the observed sound field. The estimation of these parameters is based on an energetic sound field analysis using three- dimensional microphone arrays. In practice, however, physical design constraints make three-dimensional microphone configurations often not acceptable. In this paper, we consider a new approach to microphone array processing that allows for an estimation of both direction-of-arrival of sound and diffuseness based on planar microphone configurations. The performance of the proposed method is evaluated via simulations and real measured data.

Efficient combinations of coding and manipulation of audio signals in the spectral domain are often desirable in communication systems. The modified discrete cosine transform (MDCT) represents a popular spectral transform in audio coding as it leads to compact signal representations. However, as the MDCT corresponds to a critically sampled filter bank, it is in general not appropriate to directly apply it to filtering tasks. In this paper we present a method to compensate for aliasing terms that arise from such direct MDCT domain filtering. The discussion is thereby based on a rigorous matrix representation of critically sampled filter banks which also leads to corresponding efficient realizations. As an application showcase, noise reduction for MDCT based speech coding is considered in simulations.

Loudspeakers and amplifiers of mobile communication receivers may cause significant nonlinear distortion in the acoustic echo path, resulting in a limitation of the performance of linear echo cancelers. In this contribution, we present a nonlinear acoustic echo suppressor in order to increase the attenuation of the nonlinearly distorted residual echo. The proposed approach is based on a power filter model of the acoustic echo path which is applied to the estimation of the power spectral density of the nonlinear residual echo. These time-variant estimates are used to appropriately adjust the frequency-dependent gain values of the echo suppressor. The performance of the proposed approach is evaluated in realistic experimental set-ups.

In acoustic echo cancellation as, e.g., for mobile communication receivers, loudspeakers and their amplifiers cause significant nonlinear distortion in the echo path, resulting in a degradation of the performance of linear echo cancelers. In order to cope with this type of nonlinear echo path, we discuss an orthogonalized version of power filter that can be considered as a parallelized realization of the cascade of a memoryless polynomial followed by a linear filter. As, in the echo cancellation context, the statistics of the speech input are non-stationary and not known in advance, the orthogonalization follows the signal statistics. The performance of the resulting novel nonlinear structure is evaluated by experiments using real hardware.

The compensation of nonlinear signal distortions as caused by small, low-cost loudspeakers driven at high volume by second-order Volterra filters has been investigated in recent works. In the field of acoustic echo cancellation, such undesired signal components are removed by adaptive filtering. Since the performance of the usually applied LMS adaptation suffers from correlations of the input data, preceding decorrelation of the excitation signal is desirable in order to increase the speed of convergence. This contribution discusses an efficient configuration for incorporating a decorrelation filter into nonlinear adaptive filtering scenarios. Simulation results for both noise and speech indicate an increase in convergence speed by providing decorrelated signals as filter input.

Directional audio coding (DirAC) is a recent method for spatial audio processing, based on a perceptually motivated representation of spatial sound. Due to its efficiency, DirAC has already been proposed for spatial audio teleconferencing scenarios. Modern hands-free communication systems usually include beamforming techniques to improve speech intelligibility by suppressing diffuse background noise and interfering sources. In this paper, we propose a novel spatial filtering method which can be integrated into the DirAC spatial codec. It uses a spectral weighting of the recorded audio signal, where the design of the corresponding spatial filter transfer function is based on the DirAC parameters, i. e., direction-of-arrival and diffuseness of the sound field. Simulation results show that compared to a standard beamformer the novel technique offers significantly higher interference attenuation, while introducing similar distortion of the desired signal.