Steered Response Power With Phase Transform Research Articles

Semi-Supervised Source Localization in Reverberant Environments With Deep Generative Modeling

Localization in reverberant environments remains an open challenge. Recently, supervised learning approaches have demonstrated very promising results in addressing reverberation. However, even with large data volumes, the number of labels available for supervised learning in such environments is usually small. We propose to address this issue with a semi-supervised learning (SSL) approach, based on deep generative modeling. Our chosen deep generative model, the variational autoencoder (VAE), is trained to generate the phase of relative transfer functions (RTFs) between microphones. In parallel, a direction of arrival (DOA) classifier network based on RTF-phase is also trained. The joint generative and discriminative model, deemed VAE-SSL, is trained using labeled and unlabeled RTF-phase sequences. In learning to generate and classify the sequences, the VAE-SSL extracts the physical causes of the RTF-phase (i.e., source location) from distracting signal characteristics such as noise and speech activity. This facilitates effective end-to-end operation of the VAE-SSL, which requires minimal preprocessing of RTF-phase. VAE-SSL is compared with two signal processing-based approaches, steered response power with phase transform (SRP-PHAT) and MUltiple SIgnal Classification (MUSIC), as well as fully supervised CNNs. The approaches are compared using data from two real acoustic environments - one of which was recently obtained at Technical University of Denmark specifically for our study. We find that VAE-SSL can outperform the conventional approaches and the CNN in label-limited scenarios. Further, the trained VAE-SSL system can generate new RTF-phase samples which capture the physics of the acoustic environment. Thus, the generative modeling in VAE-SSL provides a means of interpreting the learned representations. To the best of our knowledge, this paper presents the first approach to modeling the physics of acoustic propagation using deep generative modeling.

Iterative Modified SRP-PHAT with Adaptive Search Space for Acoustic Source Localization

The Steered Response Power with Phase Transform (SRP-PHAT) is one of the leading methods for sound source localization because of its proven robustness in a noisy and reverberant environment. However, for the higher accuracy, the maximum power location needs to be searched over a dense grid, making the computational cost a real issue for large-scale microphone arrays. Modified SRP-PHAT introduced an effective strategy that computes the SRP functional at each discrete location, also considering the volume surrounding it. Subsequently, it explores the entire search space and achieves improved robustness over a coarse spatial grid. To accomplish higher spatial resolution with minimal computations, the Modified SRP-PHAT was further extended by employing an iterative grid decomposition search strategy. However, in challenging environments, the performance of the method deteriorates considerably with the iterative approach. In this paper, we suggest furtherance to iterative modified SRP-PHAT by employing adaptive search space in each iteration. The performance of the proposed method is extensively tested and compared to the state-of-the-art techniques on a simulated dataset and SMARD database to validate its reliability. The results exhibited that the proposed method achieves higher accuracy with only a marginal increase in computation expense than that of iterative modified SRP-PHAT.

SRP-PHAR Combined Velocity Scanning for Locating the Shallow Underground Acoustic Source

Shallow underground acoustic source localization is a component of near-field source localization, which is involved in numerous application fields. The positioning accuracy is mainly limited by the accuracy of time of arrival (TOA)/time difference of arrival (TDOA) extraction and velocity extraction from noisy data. The steered response power with phase transform (SRP-PHAT) is one of the most robustness and high-precision acoustic source localization approaches, which avoids extracting the TDOA in advance. But SRP-PHAT is constrained for only using under known velocity. Furthermore, it is barely possible for shallow underground sound source localization to easily obtain high-quality velocity models. This paper proposes an improved SRP-PHAT with unknown velocity (SRP-PHAT-UNVEL), which avoids extracting the TDOA and velocity in advance. SRP-PHAT-UNVEL matches the calculated time delay curve with the actual time delay curve by scanning of the candidate spatial position and the candidate velocity simultaneously, so as to maximize the output energy to fulfill the positioning. However, SRP-PHAT-UNVEL has larger computational complexity as it proceeds with the optimization of space and velocity. Since the spatial position and velocity affect the shape and the curvature of the calculated delay curve respectively, these are two relatively independent processes. Therefore, the simultaneous optimization of space and velocity can be replaced by a two-stage optimization to improve the efficiency and accuracy of SRP-PHAT-UNVEL. Spatial optimization is equivalent to the optimization of SRP-PHAT, and the spatial optimization of the bat algorithm has faster convergence rate and higher location precision than traditional methods. Velocity optimization can be achieved by the common linear search technique since the function of velocity and energy is an ideal convex function. Simulation experiment results show that the proposed method is insensitive to noise, which can achieve high accuracy of the acoustic source position and velocity simultaneously. With grouping the measured data, the proposed method can further improve the robustness and accuracy by fusing the grouping location results with principal component analysis.

Direction of Arrival Estimation with Microphone Arrays Using SRP-PHAT and Neural Networks

The Steered Response Power with phase transform (SRP-PHAT) is one of the most employed techniques for Direction of Arrival (DOA) estimation with microphone arrays, but its computational complexity grows when the search space increases. To solve this issue, we propose the use of Neural Networks (NN) to obtain the DOA from low-resolution SRP-PHAT power maps.

On the Use of a GPU-Accelerated Mobile Device Processor for Sound Source Localization

The growing interest to incorporate new features into mobile devices has increased the number of signal processing applications running over processors designed for mobile computing. A challenging signal processing field is acoustic source localization, which is attractive for applications such as automatic camera steering systems, human-machine interfaces, video gaming or audio surveillance. In this context, the emergence of systems-on-chip (SoC) that contain a small graphics accelerator (or GPU), contributes a notable increment of the computational capacity while partially retaining the appealing low-power consumption of embedded systems. This is the case, for example, of the Samsung Exynos 5422 SoC that includes a Mali-T628 MP6 GPU. This work evaluates an OpenCL-based implementation of a method for sound source localization, namely, the Steered-Response Power with Phase Transform (SRP-PHAT) algorithm, on GPUs of this type. The results show that the proposed implementation, given the audio samples, is able to perform audio localization in real time with high-resolution spatial grids using up to 12 microphones.

On the performance of multi-GPU-based expert systems for acoustic localization involving massive microphone arrays

Sound source localization is an important topic in expert systems involving microphone arrays, such as automatic camera steering systems, human–machine interaction, video gaming or audio surveillance. The Steered Response Power with Phase Transform (SRP-PHAT) algorithm is a well-known approach for sound source localization due to its robust performance in noisy and reverberant environments. This algorithm analyzes the sound power captured by an acoustic beamformer on a defined spatial grid, estimating the source location as the point that maximizes the output power. Since localization accuracy can be improved by using high-resolution spatial grids and a high number of microphones, accurate acoustic localization systems require high computational power. Graphics Processing Units (GPUs) are highly parallel programmable co-processors that provide massive computation when the needed operations are properly parallelized. Emerging GPUs offer multiple parallelism levels; however, properly managing their computational resources becomes a very challenging task. In fact, management issues become even more difficult when multiple GPUs are involved, adding one more level of parallelism. In this paper, the performance of an acoustic source localization system using distributed microphones is analyzed over a massive multichannel processing framework in a multi-GPU system. The paper evaluates and points out the influence that the number of microphones and the available computational resources have in the overall system performance. Several acoustic environments are considered to show the impact that noise and reverberation have in the localization accuracy and how the use of massive microphone systems combined with parallelized GPU algorithms can help to mitigate substantially adverse acoustic effects. In this context, the proposed implementation is able to work in real time with high-resolution spatial grids and using up to 48 microphones. These results confirm the advantages of suitable GPU architectures in the development of real-time massive acoustic signal processing systems.

Optimized Acoustic Localization with SRP-PHAT for Monitoring in Distributed Sensor Networks

Acoustic localization by means of sensor arrays has a variety of applications, from conference telephony to environment monitoring. Many of these tasks are appealing for implementation on embedded systems, however large dataflows and computational complexity of multi-channel signal processing impede the development of such systems. This paper proposes a method of acoustic localization targeted for distributed systems, such as Wireless Sensor Networks (WSN). The method builds on an optimized localization algorithm of Steered Response Power with Phase Transform (SRP-PHAT) and simplifies it further by reducing the initial search region, in which the sound source is contained. The sensor array is partitioned into sub-blocks, which may be implemented as independent nodes of WSN. For the region reduction two approaches are handled. One is based on Direction of Arrival estimation and the other - on multilateration. Both approaches are tested on real signals for speaker localization and industrial machinery monitoring applications. Experiment results indicate the method’s potency in both these tasks.