The problem of training with a small set of positive samples is known as few-shot learning (FSL). It is widely known that traditional deep learning algorithms usually show very good performance when trained with large datasets. However, in many applications, it is not possible to obtain such a high number of samples. This paper deals with the application of FSL to the detection of specific and intentional acoustic events given by different types of sound alarms, such as door bells or fire alarms, using a limited number of samples. These sounds typically occur in domestic environments where many events corresponding to a wide variety of sound classes take place. Therefore, the detection of such alarms in a practical scenario can be considered an open-set recognition (OSR) problem. To address the lack of a dedicated public dataset for audio FSL, researchers usually make modifications on other available datasets. This paper is aimed at providing the audio recognition community with a carefully annotated dataset1 for FSL in an OSR context comprised of 1360 clips from 34 classes divided into pattern sounds and unwanted sounds. To facilitate and promote research on this area, results with state-of-the-art baseline systems based on transfer learning are also presented.

Sound source separation has become a topic of intensive research in the last years. The research effort has been specially relevant for the underdetermined case, where a considerable number of sparse methods working in the time-frequency (T-F) domain have appeared. In this context, although binary masking seems to be a preferred choice for source demixing, the estimated masks differ substantially from the ideal ones. This paper proposes a maximum a posteriori (MAP) framework for binary mask estimation. To this end, class-conditional source probabilities according to the observed mixing parameters are modeled via ratios of dependent Cauchy distributions while source priors are iteratively calculated from the observed histograms. Moreover, spatially smoothed posteriors in the T-F domain are proposed to avoid noisy estimates, showing that the estimated masks are closer to the ideal ones in terms of objective performance measures.

A real time speaker localization and detection system for videoconferencing environments is presented. In this system, a recently proposed modified Steered Response Power - Phase Transform (SRP-PHAT) algorithm has been used as the core processing scheme. The new SRP-PHAT functional has been shown to provide robust localization performance in indoor environments without the need for having a very fine spatial grid, thus reducing the computational cost required in a practical implementation. Moreover, it has been demonstrated that the statistical distribution of location estimates when a speaker is active can be successfully used to discriminate between speech and non-speech frames by using a criterion of peakedness. As a result, talking participants can be detected and located with significant accuracy following a common processing framework.

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.

A refinement technique based on time-frequency masking is proposed to improve source isolation in blind audio source separation algorithms. The refinement technique uses an energy-normalized source-to-interference ratio in order to identify and eliminate interfering energy from the extracted sources. Some examples using this refinement method with different separation algorithms are discussed. The results show that source isolation can be significantly enhanced with negligible degradation of the separated sources.

Block-based physical modeling is one of the most promising methods for digital sound synthesis. This technique proposes to model and discretize each instrument element separately and then, to implement their interaction. In this paper, the use of the linear bicharacteristic scheme (LBS) or upwind leapfrog is proposed for digital sound synthesis of string instruments. It provides an efficient and accurate alternative stencil to the classical leapfrog scheme of the finite difference time domain (FDTD) method. Moreover, the conversion of dependent wave equation variables into characteristic variables makes this method suitable to interact with wave digital filter models and others paradigms. This technique is extensively presented and finally justified with some examples

A direct method that uses a random optimization process is proposed for the design of parametric IIR (infinite impulse response) filters for audio equalization. The method uses the loudspeaker response as starting point in order to obtain the desired electro-acoustical target. The algorithm uses a bank of parametric second order filters, known as peak filters in the audio field, to equalize and approximate the filtered response to the desired one. When the recursive process is finished, the parameters that define the filters (frequency, gain and Q) are obtained in correction order of importance; first the ones that provide deep correction in the response. This characteristic allows the implementation of scalable systems with different degrees of complexity and correction.

The condition number of the matrix of electro-acoustic head-related transfer functions (HRTF) in a two-channel sound reproduction system has been used as a measure of robustness of the Atal-Schroeder (1962) crosstalk canceler. A comparative study has been made using results produced by computer simulations and HRTFs measured in an anechoic chamber by means of a dummy head. It has been found that acoustic scattering by the head has a very important and beneficial influence on robustness, specially for large loudspeaker separations. For narrow loudspeaker separations of less than about 40 degrees it is found that crosstalk cancellation exhibits a large variation of alternating very low and very high robustness. Also, simulations and measurements have been made of the natural channel separation under the same conditions. Scattering by the head is seen to provide a good level of natural channel separation at high frequencies and large loudspeaker angles. At low frequencies or small loudspeaker angles natural channel separation is poor.

The reproduction of three-dimensional (3-D) sound (either from binaural recordings or virtual source synthesis) through loudspeakers, also known as transauralization, adds difficulties to the traditional method using headphones: the cross-talk effect between speakers and the listener's ears must be taken into account. The common way of removing this distortion consists of placing a filter bank before reproduction through the loudspeakers, but these crosstalk cancellers are notoriously nonrobust to slight movements of the head. Recent theoretical works studied the influence of loudspeaker positions relating to robustness of the system. This paper presents a series of experiments that validate the theory and add new ideas for the design of crosstalk cancellers.

The use of Kalman filtering techniques for multichannel inverse filter design in sound reproduction systems is presented. The method exploits the nonsymmetric Toeplitz-block structure of the actual multichannel system matrix that is used to specify the electroacoustic transmission paths. The result is a matrix of causal finite impulse response filters which best approximate a given time domain response in the least squares sense. The method presents low memory and computational requirements and its time updating properties make it suitable for real time applications in nonstationary environments. Under certain conditions the method carries out the exact multichannel system inverse as formulated by the multiple-input/output inverse filtering theorem. Some practical examples are also presented which illustrate the properties of the algorithm.

A time domain recursive algorithm for multichannel inverse filter design in sound reproduction systems is presented. The method exploits the non-symmetric Toeplitz-block structure of the actual multichannel system matrix that is used to specify the electroacoustic transmission paths. The result is a matrix of causal finite impulse response filters which best approximate a given time domain response in the least squares sense. The method presents low memory and computational requirements and its time updating properties make it suitable for real time applications in non-stationary environments. A cross talk cancellation experiment is presented which illustrates the properties of the algorithm.

This paper addresses the problem of computation of inverse filters in transauralization and multichannel sound reproduction. A huge set of block Toeplitz equations must be usually solved in order to design good filters under the least squares error criterion. A practical and efficient method of main matrix construction and the solution of the set of equations are thoroughly described. An optimal implementation of the Levinson algorithm in two steps that dramatically reduces the computational cost is also presented. The algorithm uses the generalized Levinson recursion for block matrices as a first step and the Gohberg-Semencul formula as the second step. This allows to save about 75% of operations in a typical transauralization problem. A study of the performance of the algorithm in terms of computational cost compared to other algorithms is also carried out.