This paper investigates the impact of input feature frequency resolution and hyperparameter choices on deep neural network based binaural speech enhancement (SE) for hearing aids. We analyze how time-frequency resolution parameters, such as window length and hop ratio, influence speech intelligibility and computational efficiency. Performance metrics, including Modified Binaural Short-Time Objective Intelligibility (MBSTOI) and Signal-to-Noise Ratio improvement (SNRi) are used to evaluate model efficacy. Our findings show that higher frequency resolutions improve intelligibility but demand greater computational resources, while optimized window and hop settings strike a balance between performance and complexity. Additionally, we use Pareto curves to analyze trade-offs between model complexity and performance, offering guidelines for designing efficient audio pipelines for resource-constrained devices, paving the way for practical binaural SE deployment in hearing aids.

This paper describes how to encode into a spherical-harmonic representation of a 3-D sound field using a cylindrical microphone array. The standard way to represent a sound field is to use higher order Ambisonics specified over a spherical region. If the sound field is instead measured using a cylindrical array, it is more naturally expressed in terms of cylindrical harmonics. In this paper, the transforms for conversion between cylindrical and spherical-harmonic representations are derived. Simulations then compare the performance of cylindrical and spherical array processing in estimating the spherical-harmonic coefficients of a sound field.

Impulse response shaping (IRS) is a prefiltering technique for modifying the characteristics of a linear channel to achieve desirable characteristics. Acoustic impulse response shaping is used to apply partial equalization to reduce the effects of reverberation on audio signals propagating inside a room and is thus used for listening room compensation. To be capable of responding rapidly to time varying room acoustics, a shaping algorithm must be computationally fast; this is difficult to achieve since the filters to be designed may have thousands of coefficients. This paper derives variants of the Dual Augmented Lagrangian Method, presents their application to the design of the shaping filters, and validates their efficacy through experimental studies. In addition, the formulations have been modified to design shaping filters that are robust to change in microphone positions.

Spatial audio may be performed in rooms using active-compensated sound field reproduction (AC-SFR), to cancel the reverberant reflections. Reproduction of a sound field may be performed using precalculated loudspeaker filters. However the robustness can be poor due to the temperature-induced perturbation of the room impulse responses (RIR). To compensate for reverberation, filters with significant non-causal components are designed. It is known that the requirements on the loudspeaker filters for room equalization may be reduced by applying the method of impulse response shaping (IRS). In this paper, we show that actively compensating paths with a long propagation distance are more heavily influenced by temperature than short paths. We hence impose a shaping penalty on the non-causal components of designed loudspeaker filters, to reduce the error caused by a change in temperature. We obtain AC-SFR designs that are more robust to changes in temperature.

Acoustic impulse response shaping is a prefiltering technique to reduce the perceptible effects of room reverberation. The room impulse responses to be shaped must first be measured, and these measurements can contain errors. Furthermore, room responses vary with changes in temperature and humidity and also with changes in measurement position. This letter presents a method for enhancing the robustness of the shaping procedure so that the effects of these errors and changes are minimized. The method uses a stochastic model of the channel variations to explicitly limit the probability of large deviations from the desired performance. The method is evaluated on realistic channel perturbations and the resulting shaped responses are shown to comply with the robustness specification.

Impulse response shaping is a technique for partly equalizing impulse responses. In acoustics, it can be used for the reproduction of audio signals mitigated by distortions in a room. The most significant phenomenon among the distortions is reverberation, a straightforward characterization of which is the room impulse response. Room responses can be characterized but could contain measurement errors or noise. In addition, room responses vary with changes in atmospheric conditions such as temperature and humidity and also due to change in positions inside a room. The design of a shaping filter robust to at least some of these variations is likely to be very useful, which is considered in this work. The method uses a computationally efficient approach based on Basis Pursuit DeNoising (BPDN).

This paper proposes an efficient parameterization of the room transfer function (RTF). Typically, the RTF rapidly varies with varying source and receiver positions, hence requires an impractical number of point to point measurements to characterize a given room. Therefore, we derive a novel RTF parameterization that is robust to both receiver and source variations with the following salient features: 1) The parameterization is given in terms of a modal expansion of 3D basis functions. 2) The aforementioned modal expansion can be truncated at a finite number of modes given that the source and receiver locations are from two sizeable spatial regions, which are arbitrarily distributed. 3) The parameter weights/coefficients are independent of the source/receiver positions. Therefore, a finite set of coefficients is shown to be capable of accurately calculating the RTF between any two arbitrary points from a pre-defined spatial region where the source(s) lie and a pre-defined spatial region where the receiver(s) lie. A practical method to measure the RTF coefficients is also provided, which only requires a single microphone unit and a single loudspeaker unit, given that the room characteristics remain stationary over time. The accuracy of the above parameterization is verified using appropriate simulation examples.

An impulse response shortening approach is used to perform acoustic crosstalk cancellation. Crosstalk canceler filters are traditionally designed using least squares, with an approach that equalizes all room reverberation. However, depending upon end application, some reverberation may be permissible in the delivered signals. This idea is used to create more efficient crosstalk cancellation filters. The filter design is formulated as a minimax problem solvable with linear programming methods. Penalty functions on crosstalk levels and detrimental reverberation are introduced, which allow control of the reverberant tails and crosstalk levels. Shorter crosstalk cancellation filters are designed, by leaving in early echoes and/or allowing a slower decay of the late reverberant tail.

The implementation of 3D sound reproduction is well founded theoretically, but the requirements for the number of loudspeakers and array geometry make it impractical using conventional technology. In practice, a 2D array of loudspeakers is commonly used which restricts the reproduction of sources to those in the horizontal plane and requires a restricted form for the free-field loudspeaker driving signals based on the sectorial spherical harmonics. However, reflections can impair reproduction quality when using the free-field solution. Since first-order loudspeakers are commercially available which increase the direct to reflected sound ratio, we extend the sectorial solution for the loudspeaker excitation signals to the first-order case. We also investigate the reproduction accuracy for both zeroth and first-order loudspeakers using numerical simulations.

Audio rendering problems are not always well-posed. An approach is devised for solving ill-posed sound field reproduction problems using regularization, where the Tikhonov parameter is chosen by upper bounding the summed square of the loudspeaker weights. The method ensures that the sound in the room remains at reasonable levels.

A recent approach to surround sound is to perform exact control of the sound field over a region of space. Here, the driving signals for an array of loudspeakers are chosen to create a desired sound field over an extended area. An interesting subtopic is multi-zone surround sound, where two or more listeners can experience totally independent sound fields. However, multi-zone surround sound is a challenge because implementation can be very non-robust. We formulate multi-zone sound reproduction as a convex optimization problem, where the sound energy leakage into other listener zones is limited to fixed levels, and a constraint is placed on the loudspeaker weights to improve the robustness. An interior point algorithm is de vised for computing the loudspeaker weights, and its performance is compared with least squares approaches of multi-zone reproduction in typical two-zone cases.

The reproduction of a sound field over a spatial region can require large output powers from the loudspeakers. In such a case, the sound field reproduction can be non-robust or cause loudspeaker failure. One method of limiting loudspeaker power is to regularize the solution. We instead assume each loudspeaker is capable of delivering a certain peak power, and find a solution minimizing the error between the created and desired field subject to this power constraint. We apply a power constraint at a single frequency and also constrain total output power summed across many frequencies. Rapidly converging interior point methods are derived and shown to possess similar robustness to the regularized solution.

We devise a method for sound field reproduction (SFR) around a solid object in a reverberant room. Until now, work have focussed on reproducing sound in an empty listening space and, for the most part, in non-reverberant environments. However, in a reverberant, room as soon as a listener steps into the space he alters his acoustic environment, generating a sound component which is body-scattered and successively reverberated throughout the room. Building on the model of the sound field around a solid sphere in free space, we extend to reproduction around a human head in a reverberant room. In doing so, we show the relationship between the pressure matching and mode matching approaches of SFR.

In this paper we consider the spatial degrees of freedom in the context of a multi-antenna wireless communication system. We investigate how the degrees of freedom depend on important system parameters such as the spatial extent of a region containing the antennas and, more importantly, the angular correlation of multipath. These results naturally augment known results which show how the degrees of freedom are affected by multipath from a restricted range of angles. We clarify the distinction between the spatial degrees of freedom with respect to an orthonormal basis and the concept of richness of multipath which is related to the Karhenen Loeve expansions for a random multipath field.

This paper investigates the correlations between sensor signals in multipath environments created by correlated scatterers. We derive a closed form expression for the correlation in fields created by arbitrary scatterer correlations and scatterer powers, using Fourier techniques, and propose reasonable function forms for scatterer correlation. Simulations show notable differences from the uncorrelated scatterer case

In this paper, we use a novel MIMO channel model to characterize the dependence of ergodic capacity and diversity order on the joint statistics of the angular power density. The scattering environment of a MIMO channel is characterized by a double directional angular power distribution, describing the power transferred in each direction from transmitter aperture to receiver aperture. Angular power, which is typically separable Kronecker-modelled, is here generalized to include joint distribution properties using well-known bivariate probability density functions. We show that the joint properties of the power density, namely the shape and the orientation of power distribution contours, have significant impact on capacity and diversity of non-line-of-sight (NLOS) channels

In this paper, we develop a channel simulator to generate the channel gains to an arbitrary array of receiver antennas, for a general class of non-line-of-sight channels. The channel scattering environment is defined by the angular power distribution as seen by the receiver. We derive the second order statistics of the channel gains in terms of the parameters of the angular power distribution. As an illustration of the channel simulator, we compare the performance of different direction-of-arrival techniques

This paper presents an efficiently parametrised model for second-order-statistics dominated MIMO channels. Recently, new MIMO models have been developed to emulate the statistics of real measurements: (1) analytic models which parametrise the statistics of the channel gains, and (2) geometric models which interpret the channel as separate multi-paths. Unfortunately analytic models are tied to the measurement array geometry, while geometric models significantly increase model complexity. We present a new stochastic framework, based on a modal decomposition of the MIMO channel, which allows channel models for arbitrary array geometries from a single set of measured data. Such a framework yields simple MIMO models that efficiently parametrise the channel, with adjustable accuracy. Results show that the new models match the capacity of real and simulated data as well as similar models

In this paper, we present a novel method of soundfield reproduction (SFR) for reverberant acoustic environments. Using an efficient parametrization of the acoustic transfer function (ATF) over a region of space, we devise a method for accurate SFR over the whole of the reproduction region. This method is based on a practical method of determining the ATF between each loudspeaker and the reproduction region.

We investigate the performance of acoustic equalization in reverberant environments. We first highlight an efficient general representation of a sound field using spherical harmonics. We then use this representation to develop a concise closed-form expression for robustness of equalization to sensor movement. This expression is used (i) to characterize equalization performance for a general class of non-isotropic sound fields and (ii) to quantify the improvements to equalizer robustness that can be obtained by using a directional microphone. This approach does not use any of the assumptions of statistical acoustics, but instead exploits the inherent properties of a sound field as described by the wave equation.

This paper examines the improvement that can be attained with perfect knowledge of the sound source directivity pattern and orientation in beamformer designs in the problem of speech acquisition. Data-independent beamformers are derived through formulation of a constrained optimization problem with a unity-gain constraint. Using computer simulation, these beamforming schemes are compared to the delay and sum (DS) beamformer and the best single sensor in a reverberant room environment. Criteria used to measure performance are: (1) the direct to reverberant ratio, to assess extent of reverberation suppression; and (2) an objective measure of speech intelligibility called the speech transmission index (STI). For human-speaker source directivity, simulation results show that modest improvements to performance are obtainable.