Many algorithms for speech dereverberation and noise reduction rely on an estimate of the coherent-to-diffuse power ratio (CDR). Such systems typically operate in very diverse acoustic conditions, and CDR estimators relying on very weak model assumptions about the acoustic sound field of the desired speech and interfering noise are hence desirable. A CDR estimator whose design is based on this premise is devised in this contribution. The proposed non-iterative CDR estimator exploits the effective rank of the spatial covariance matrix of the recorded input signals by assuming it to be lower for a coherent sound field than for a diffuse sound field. In addition to this weak assumption, related methods usually require information about, e.g., the array geometry, direction-of-arrival (DOA) of the desired source or rely on a coherence model for the desired signal or background noise, which is not required for the proposed method. Despite the use of little a priori information about the acoustic sound field, the new estimator achieves a significantly higher estimation accuracy for the CDR in comparison to related state-of-the-art approaches which use explicit coherence models.

This contribution presents a novel approach for coherence-based signal enhancement. An estimator for the coherent-to-diffuse ratio (CDR) is devised, which exploits the concept of generalized magnitude coherence and thus, unlike common state-of-the-art schemes, can simultaneously take advantage of more than two microphones. Moreover, the speech enhancement by CDR-based spectral weighting is not performed as a post-filtering step, but by enhancing the most appropriate microphone signal. This signal is implicitly determined as part of the CDR estimation such that the presented technique does not depend on an estimation of the direction-of-arrival (DOA) or similar side-information about the desired source.The application of the new approach to binaural hearings aids shows that it achieves a consistently better speech enhancement performance than comparable state-of-the-art approaches.

Autonomous systems (ASs) operating in real-world environments are exposed to a plurality and diversity of sounds that carry a wealth of information for perception in cognitive dynamic systems. While the importance of the acoustic modality for humans as “ASs” is obvious, it is investigated to what extent current technical ASs operating in scenarios filled with airborne sound exploit their potential for supporting self-awareness. As a first step, the state of the art of relevant generic techniques for acoustic scene analysis (ASA) is reviewed, i.e., source localization and the various facets of signal enhancement, including spatial filtering, source separation, noise suppression, dereverberation, and echo cancellation. Then, a comprehensive overview of current techniques for ego-noise suppression, as a specific additional challenge for ASs, is presented. Not only generic methods for robust source localization and signal extraction but also specific models and estimation methods for ego-noise based on various learning techniques are discussed. Finally, active sensing is considered with its unique potential for ASA and, thus, for supporting self-awareness of ASs. Therefore, recent techniques for binaural listening exploiting head motion, for active localization and exploration, and for active signal enhancement are presented, with humanoïd robots as typical platforms. Underlining the multimodal nature of self-awareness, links to other modalities and nonacoustic reference information are pointed out where appropriate.

The ability to localize and track acoustic events is a fundamental prerequisite for equipping machines with the ability to be aware of and engage with humans in their surrounding environment. However, in realistic scenarios, audio signals are adversely affected by reverberation, noise, interference, and periods of speech inactivity. In dynamic scenarios, where the sources and microphone platforms may be moving, the signals are additionally affected by variations in the source-sensor geometries. In practice, approaches to sound source localization and tracking are often impeded by missing estimates of active sources, estimation errors, as well as false estimates. The aim of the LOCAlization and TrAcking (LOCATA) Challenge is an open-access framework for the objective evaluation and benchmarking of broad classes of algorithms for sound source localization and tracking. This article provides a review of relevant localization and tracking algorithms and, within the context of the existing literature, a detailed evaluation and dissemination of the LOCATA submissions. The evaluation highlights achievements in the field, open challenges, and identifies potential future directions.

The use of autonomous systems (ASs), such as humanoid robots, drones or self-driving vehicles, has expanded significantly in recent years. For such systems, acoustic scene analysis can provide useful information about the environment and supports the AS to react appropriately. However, compared to most other application areas, analysis and enhancement of acoustic signals captured by ASs is not only complicated by external sources of signal degradation but also by very specific challenges like internal and self-created ego-noise. This paper first gives an overview of a typical acoustic scenario an AS is exposed to. Then, we consider the specific problem of ego-noise suppression and propose to use motor data to predict the characteristic time-varying harmonic structure of ego-noise. This knowledge is then incorporated into a multichannel dictionary-based algorithm. The resulting two-stage ego-noise reduction scheme is evaluated for ego-noise of a humanoid robot and outperforms a comparable method that uses no motor data but a a larger dictionary.

A novel numerical design approach for allpass transformed DFT filter-banks with subsampling is proposed. Such frequency warped filter-banks possess a non-uniform time-frequency resolution which is of interest, e.g., for speech and audio processing. The coefficients of the FIR synthesis filters are determined by a constrained least-squares error (CLSE) design. Thereby, the stopband energy of the synthesis filters is minimized with the constraint for perfect reconstruction (PR). This design by quadratic programming accounts also for polyphase network (PPN) filter-banks where the prototype filter degree exceeds the number of subbands to facilitate an enhanced frequency selectivity. In contrast to known designs for allpass transformed analysis-synthesis filter-banks (AS FBs), perfect signal reconstruction can be achieved effectively with synthesis filters featuring a pronounced bandpass characteristic.

An improved approach for the estimation of the frequency dependent reverberation time (RT) by means of allpass transformed filter-banks is presented. It is shown that by means of these warped filter-banks, a much more accurate RT estimation at lower frequencies can be obtained than by octave filter-banks, which are commonly used for the estimation of the frequency dependent RT. Furthermore, allpass transformed filter-banks can achieve a much better approximation of the non-uniform frequency resolution of the human auditory system than octave filter-banks. A uniform or non-uniform (auditory) frequency resolution can thereby be simply adjusted by a single allpass coefficient. The RT estimation can be done with an allpass transformed DFT or DCT filter-bank. The warped DCT filter-bank is of special interest as it provides real-valued subband signals. This facilitates the use of a maximum-likelihood (ML) estimator for either a non-blind estimation of the frequency dependent RT from a room impulse response or a blind estimation from a reverberant speech signal.

Allpass transformed filter banks provide a nonuniform frequency resolution and can be used in mobile speech processing systems, e.g., cellular phones or digital hearing aids. The nominal design of such an allpass transformed analysis-synthesis filter bank (AS FB) with near perfect reconstruction (NPR) is achieved by numerical optimization of finite-impulse response (FIR) equalizers in each subchannel. The underlying nominal optimization problem is an equality constrained leasts-quares problem. In a robust design, we take into account coefficient uncertainty in a possible implementation of such a filter bank. We will describe this uncertainty by the choice of two simple set-based worst-case uncertainty models, namely a norm bound error model and a coefficient bound error model. When including these error models, both robust designs can be recast as second-order cone programs (SOCP) and solved efficiently by standard numerical optimization methods. Furthermore, we will provide design examples to show that both robust designs maintain a good overall performance with respect to NPR while offering less sensitivity to quantization errors.

This paper describes the speech and audio codec that has been submitted to ITU-T by Huawei and ETRI as a candidate for the upcoming super-wideband and stereo extensions of Rec. G.729.1 and G.718. The core codec in the current implementation is G.729.1 and the encoded frequency range is increased from 7 kHz to 14 kHz. Therefore, the maximum bit rate is raised from 32 kbit/s to 64 kbit/s by adding five bitstream layers. A comprehensive overview of the codec is presented with a focus on the mono coding components. The results of the listening tests that have been conducted during the ITU-T qualification phase are summarized. The proposed codec passes all quality requirements for mono input signals.

A new design for a recursive DFT analysis-synthesis filter-bank (AS FB) with critical subsampling and near-perfect reconstruction is proposed. The analysis filter-bank consists of allpass polyphase filters to achieve a high frequency selectivity with a low algorithmic complexity and low signal delay. The condition for perfect reconstruction (PR) leads to either unstable or anti-causal synthesis filters. This problem is solved by stable allpass polyphase synthesis filters which are designed by analytical closed-form expressions. The first design can achieve arbitrarily small aliasing, amplitude and phase distortions in dependence of the tolerable signal delay and algorithmic complexity. The second design avoids aliasing and amplitude distortions and minimizes the phase distortions at the expense of an increased system delay. The proposed IIR filter-bank possesses also a significantly lower algorithmic complexity than comparable FIR filter-banks.

This paper proposes a new speech enhancement algorithm for the suppression of background noise and late reverberation without a priori knowledge. A generalized spectral weighting rule based on estimations for the spectral variances of the late reverberant speech and background noise is used for speech enhancement. This allows to suppress speech distortions due to late room reflections without knowledge about the complete room impulse response. In contrast to existing methods, all needed quantities are estimated entire blindly from the reverberant and noisy speech signal. The new algorithm has also a low signal delay which is important, e.g., for speech enhancement in mobile communication devices or hearing aids.

A new design approach for an allpass transformed analysis-synthesis filter-bank (AS FB) with subsampling is proposed, which can be used for adaptive subband processing with non-uniform time-frequency resolution. Signal reconstruction is performed by an FIR synthesis filter-bank whose coefficients are determined by two different linear least-squares (LS) error designs. The presented unconstrained LS optimization leads to perfect reconstruction (PR), but an insufficient bandpass characteristic for the synthesis subband filters. This problem is solved by a second, equality constrained least-squares (CLS) error design, which achieves near-perfect reconstruction (NPR). In the process, the linear distortions are minimized with the constraint for a linear, time-invariant (LTI) overall transfer function. The aliasing-free reconstruction error is significantly lower than for known designs of allpass transformed DFT filter-banks with NPR. Moreover, the new filter-bank design allows for a low and adjustable signal delay.

A novel design for a two-channel IIR quadrature-mirror filter (QMF) bank with near-perfect reconstruction (NPR) is presented. The analysis filter-bank is given by an efficient polyphase network (PPN) implementation based on allpass filters. The arising phase distortions are almost compensated by stable allpass filters, designed via analytical closed-form expressions. In a first design, the remaining aliasing, amplitude and phase distortions become arbitrarily small in dependence of the tolerable system delay and algorithmic complexity, respectively. In a second design, aliasing and amplitude distortions are completely canceled and phase distortions are minimized at the expense of an additional signal delay. The proposed QMF banks have a lower algorithmic complexity than comparable designs.

An efficient realization of a low delay filter-bank, termed as generalized filter-bank equalizer (FBE), will be proposed for noise reduction with low signal delay. The FBE is equivalent to a time-domain filter with coefficients adapted in the frequencydomain. This filter-bank structure ensures perfect signal reconstruction for a variety of spectral transforms with less restrictions than for an analysis-synthesis filter-bank (AS FB). A non-uniform frequency resolution can be achieved by an allpass transformation. In this case, the FBE has not only a lower signal delay than the AS FB, but also a lower algorithmic complexity for most parameter configurations. Another advantage of the FBE is the lower number of required delay elements (memory) compared to the AS FB. The noise reduction achieved by means of the AS FB and the FBE is approximately equal.

This paper describes an eigendecomposition based approach to estimate the direction of arrival (DOA) of speech or noise sources in a reverberant room by using a microphone array. Unitary ESPRIT, a DOA estimation algorithm for narrowband signals, is applied to these wideband signals, which results in DOA estimates for each frequency. A single estimate is then obtained by means of a histogram. The estimated DOAs of the speech source are used to enhance a noisy speech signal by means of a superdirective beamformer. Experiments in a noisy reverberant room showed a higher estimation accuracy, and thus a better speech enhancement performance, at low SNRs compared to other DOA estimation techniques.