Maximo Cobos

A Modified SRP-PHAT Functional for Robust Real-Time Sound Source Localization With Scalable Spatial Sampling

The Steered Response Power - Phase Transform (SRP-PHAT) algorithm has been shown to be one of the most robust sound source localization approaches operating in noisy and reverberant environments. However, its practical implementation is usually based on a costly fine grid-search procedure, making the computational cost of the method a real issue. In this letter, we introduce an effective strategy that extends the conventional SRP-PHAT functional with the aim of considering the volume surrounding the discrete locations of the spatial grid. As a result, the modified functional performs a full exploration of the sampled space rather than computing the SRP at discrete spatial positions, increasing its robustness and allowing for a coarser spatial grid. To this end, the Generalized Cross-Correlation (GCC) function corresponding to each microphone pair must be properly accumulated according to the defined microphone setup. Experiments carried out under different acoustic conditions confirm the validity of the proposed approach.

Robust acoustic source localization based on modal beamforming and time-frequency processing using circular microphone arrays.

Uniform circular array processing has been shown to be a very useful tool for broadband acoustic source localization over 360°. Specifically, beamforming methods based on circular harmonics have attracted a lot of research attention in the last several years, as modal array signal processing is a very active research topic. On the other hand, due to the sparsity properties of speech, source localization methods in the time-frequency (T-F) domain have also demonstrated their capability to locate several simultaneous sources with high accuracy. In this paper, a localization framework based on circular harmonics beamforming and T-F processing that provides accurate localization performance under very adverse acoustic conditions is presented. Modal processing and sparsity-based localization are jointly addressed to estimate the direction-of-arrival of multiple concurrent speech sources. Experiments in real and simulated environments with different microphone setups are discussed, showing the validity of the proposed approach and comparing its performance with other state-of-the-art methods.

Low-Cost Alternatives for Urban Noise Nuisance Monitoring Using Wireless Sensor Networks

Noise pollution caused by vehicular traffic is a common problem in urban environments that has been shown to affect peoples health and childrens cognition. In the last decade, several studies have been conducted to assess this noise, by measuring the equivalent noise pressure level (called Leq) to acquire an accurate sound map using wireless networks with acoustic sensors. However, even with similar values of Leq, people can feel the noise differently according to its frequency characteristics. Thus, indexes, which can express peoples feelings by subjective measures, are required. In this paper, we analyze the suitability of using the psychoacoustic metrics given by the Zwickers model, instead of just only considering Leq. The goal is to evaluate the hardware limitations of a low-cost wireless acoustic sensor network that is used to measure the annoyance, using two types of commercial and off-the-shelf sensor nodes, Tmote-Invent nodes and Raspberry Pi platforms. Moreover, to calculate the parameters using these platforms, different simplifications to the Zwickers model based on the specific features of road traffic noise are proposed. To validate the different alternatives, the aforementioned nodes are tested in a traffic congested area of Valencia City in a vertical and horizontal network deployment. Based on the results, it is observed that the Raspberry Pi platforms are a feasible low-cost alternative to increase the spatial-temporal resolution, whereas Tmote-Invent nodes do not confirm their suitability due to their limited memory and calibration issues.

A sparsity-based approach to 3D binaural sound synthesis using time-frequency array processing

Localization of sounds in physical space plays a very important role in multiple audio-related disciplines, such as music, telecommunications, and audiovisual productions. Binaural recording is the most commonly used method to provide an immersive sound experience by means of headphone reproduction. However, it requires a very specific recording setup using high-fidelity microphones mounted in a dummy head. In this paper, we present a novel processing framework for binaural sound recording and reproduction that avoids the use of dummy heads, which is specially suitable for immersive teleconferencing applications. The method is based on a time-frequency analysis of the spatial properties of the sound picked up by a simple tetrahedral microphone array, assuming source sparseness. The experiments carried out using simulations and a real-time prototype confirm the validity of the proposed approach.

Two-microphone multi-speaker localization based on a Laplacian Mixture Model

A method for multiple speaker localization using a two-microphone array is presented. Based on the disjointness property of speech signals in the time-frequency (TF) domain, the proposed approach models the distribution of Direction-Of-Arrival (DOA) estimates in the TF plane using a Laplacian Mixture Model (LMM). Reliable DOA estimates are pre-selected using a coherence-based criterion before fitting the model by means of the Expectation-Maximization (EM) algorithm. Results using public available data show the capability of the method to successfully detect the DOA of several sources in real environments.

A steered response power iterative method for high-accuracy acoustic source localization

Source localization using the steered response power (SRP) usually requires a costly grid-search procedure. To address this issue, a modified SRP algorithm was recently introduced, providing improved robustness when using coarser spatial grids. In this letter, an iterative method based on the modified SRP is presented. A coarse spatial grid is initially evaluated with the modified SRP, selecting the point with the highest accumulated value. Then, its corresponding volume is iteratively decomposed by using a finer spatial grid. Experiments have shown that this method provides almost the same accuracy as the fine-grid search with a substantial reduction of functional evaluations.

Stereo audio source separation based on time--frequency masking and multilevel thresholding

Source separation and up-mixing in real commercial music recordings is a challenging problem. In the last few years, some algorithms have provided interesting results, but the problem remains unsolved. In this paper we describe a method for separating the sources present in a two channel mixture based on the panning coefficients used in the stereo mixdown. The sources are separated by estimating time-frequency masks using the multilevel extension of the Otsu thresholding algorithm used in image segmentation. A refinement step is also carried out for extraction and reassignment of inter-source residuals. Examples of application and performance evaluation are also discussed.

Cumulative-sum-based localization of sound events in low-cost wireless acoustic sensor networks

Wireless acoustic sensor networks (WASNs) are known for their potential applications in multiple areas, such as audio-based surveillance, binaural hearing aids or advanced acoustic monitoring. The knowledge of the spatial position of a source of interest is usually a requirement for many of these applications. Therefore, source localization is an important problem to be addressed in WASNs. Unfortunately, most localization algorithms need costly signal processing stages that prevent them from being implemented in low-cost sensor networks, requiring additional modules for signal acquisition and processing. This paper presents a low-complexity method for acoustic event detection and localization considering a change detection statistical framework. Two possible implementation approaches based on the efficient cumulative sum (CUSUM) algorithm are presented and discussed. Results from simulations and a real deployment show that the proposed techniques can be easily implemented in low-cost sensor networks, providing good localization accuracy and making good use of the available node resources.

Computer-based detection and classification of flaws in citrus fruits

In this paper, a system for quality control in citrus fruits is presented. In current citrus manufacturing industries, calliper and color are successfully used for the automatic classification of fruits using vision systems. However, the detection of flaws in the citrus surface is carried out by means of human inspection. In this work, a computer vision system capable of detecting defects in the citrus peel and also classifying the type of flaw is presented. First, a review of citrus illnesses has been carried out in order to build a database of digitalized oranges classified by the kind of fault, which is used as a training set. The segmentation of faulty zones is performed by applying the Sobel gradient to the image. Afterwards, color and texture features of the flaw are extracted considering different color spaces, some of them related to high order statistics. Several techniques have been employed for classification purposes: Euler distance to a prototype, to the nearest neighbor and k-nearest neighbors. Additionally, a three layer neural network has been tested and compared, obtaining promising results.

Two-microphone separation of speech mixtures based on interclass variance maximization.

Sparse methods for speech separation have become a discussed issue in acoustic signal processing. These sparse methods provide a powerful approach to the separation of several signals in the underdetermined case, i.e., when there are more sources than sensors. In this paper, a two-microphone separation method is presented. The proposed algorithm is based on grouping time-frequency points with similar direction-of-arrival (DOA) using a multi-level thresholding approach. The thresholds are calculated via the maximization of the interclass variance between DOA estimates and allow to identify angular sections, wherein the speakers are located with a strong likelihood. These sections define a set of time-frequency masks that are able to separate several sound sources in realistic scenarios and with little computational cost. Several experiments carried out under different mixing situations are discussed, showing the validity of the proposed approach.