Research on soundscapes has shifted the focus of environmental acoustics from noise levels to the perception of sounds, incorporating contextual factors. Soundscape emotion recognition (SER) models perception using a set of features, with arousal and valence commonly regarded as sufficient descriptors of affect. In this work, we blend \emph{graph learning} techniques with a novel \emph{information criterion} to develop a feature selection framework for SER. Specifically, we estimate a sparse graph representation of feature relations using linear structural equation models (SEM) tailored to the widely used Emo-Soundscapes dataset. The resulting graph captures the relations between input features and the two emotional outputs. To determine the appropriate level of sparsity, we propose a novel \emph{generalized elbow detector}, which provides both a point estimate and an uncertainty interval. We conduct an extensive evaluation of our methods, including visualizations of the inferred relations. While several of our findings align with previous studies, the graph representation also reveals a strong connection between arousal and valence, challenging common SER assumptions.

An index of an effective number of variables (ENV) is introduced for model selection in nested models. This is the case, for instance, when we have to decide the order of a polynomial function or the number of bases in a nonlinear regression, choose the number of clusters in a clustering problem, or the number of features in a variable selection application (to name few examples). It is inspired by the idea of the maximum area under the curve (AUC). The interpretation of the ENV index is identical to the effective sample size (ESS) indices concerning a set of samples. The ENV index improves drawbacks of the elbow detectors described in the literature and introduces different confidence measures of the proposed solution. These novel measures can be also employed jointly with the use of different information criteria, such as the well-known AIC and BIC , or any other model selection procedures. Comparisons with classical and recent schemes are provided in different experiments involving real datasets. Related Matlab code is given.

We introduce a generalized information criterion that contains other well-known information criteria, such as Bayesian information Criterion (BIC) and Akaike information criterion (AIC), as special cases. Furthermore, the proposed spectral information criterion (SIC) is also more general than the other information criteria, e.g., since the knowledge of a likelihood function is not strictly required. SIC extracts geometric features of the error curve and, as a consequence, it can be considered an automatic elbow detector. SIC provides a subset of all possible models, with a cardinality that often is much smaller than the total number of possible models. The elements of this subset are “elbows” of the error curve. A practical rule for selecting a unique model within the sets of elbows is suggested as well. Theoretical invariance properties of SIC are analyzed. Moreover, we test SIC in ideal scenarios where provides always the optimal expected results. We also test SIC in several numerical experiments: some involving synthetic data, and two experiments involving real datasets. They are all real-world applications such as clustering, variable selection, or polynomial order selection, to name a few. The results show the benefits of the proposed scheme. Matlab code related to the experiments is also provided. Possible future research lines are finally discussed.

In the last decade, soundscapes have become one of the most active topics in Acoustics, providing a holistic approach to the acoustic environment, which involves human perception and context. Soundscapes-elicited emotions are central and substantially subtle and unnoticed (compared to speech or music). Currently, soundscape emotion recognition is a very active topic in the literature. We provide an exhaustive variable selection study (i.e., a selection of the soundscapes indicators) to a well-known dataset (emo-soundscapes). We consider linear soundscape emotion models for two soundscapes descriptors: arousal and valence. Several ranking schemes and procedures for selecting the number of variables are applied. We have also performed an alternating optimization scheme for obtaining the best sequences keeping fixed a certain number of features. Furthermore, we have designed a novel technique based on Gibbs sampling, which provides a more complete and clear view of the relevance of each variable. Finally, we have also compared our results with the analysis obtained by the classical methods based on p-values. As a result of our study, we suggest two simple and parsimonious linear models of only 7 and 16 variables (within the 122 possible features) for the two outputs (arousal and valence), respectively. The suggested linear models provide very good and competitive performance, with$R^{2}>0.86$and$R^{2}>0.63$(values obtained after a cross-validation procedure), respectively.

Effective operation of next-generation communication networks requires the deployment of a high number of base stations (BSs) capable of adapting dynamically their available resources to the changing environment. The resources include link layer variables (user-channel allocation and user-BS assignment) that, due to their binary nature, render the design challenging. This work proposes algorithms for user-BS allocation in cellular networks where users access orthogonally and close-by BSs use non-interfering channels. The user-BS allocation algorithms are designed jointly with the power, rate, and user-channel allocation, and take into account the dynamic environment. Three different algorithms are designed, each of them updates (adapts) the user-BS allocation at a different speed. We show that although the linear relaxation of all the binary variables is not optimal, a Benders' decomposition approach can be used to find the optimal solution. To accomplish this, we split the original problem so that the user-BS variables are isolated, relax the remaining binary variables, and solve the (sub-)problems iteratively.

3G small cells are becoming a promising solution for inexpensive 3G coverage in rural small villages, but the high cost of backhaul infrastructures remains an obstacle for the worldwide adoption of this kind of solutions. On the other hand, broadband wireless infrastructures may be found in many rural areas for communications in specific sectors such as health or education services, environmental monitoring, etc. These sectorial broadband communications infrastructures are often multi-hop wireless networks using long-distance WiFi links or WiMAX. We propose the possibility of 3G operators sharing those third parties' infrastructures. It would permit them to have a low-cost backhaul solution for their rural 3G cells in those areas where the expected demand of 3G services does not ensure enough revenues to justify the deployment of dedicated infrastructures. This is a win-win solution because bandwidth-renting may help to achieve financial stability for such sectorial networks, while operators can extend their coverage to regions where this would not be economically sustainable otherwise. This paper studies the technical feasibility of the proposed solution. Different multi-hop broadband rural networks are studied, and a general architecture is derived from them. Then, additional requirements are introduced in order to make such networks support the QoS requirements of 3G voice and data traffic. Three techniques are considered in order to manage the network in such a way that 3G traffic and other traffics may be transported together in the network without compromising the quality of the backhaul: DiffServ, plain MPLS and MPLS-TE. Measurements taken on a real testbed in a laboratory allow us to compare the performance achieved with these three techniques, and the advantages and drawbacks of each of them are presented.

In this paper, we propose the use of an ad-hoc wireless network formed by a fraction of the passing vehicles (sensor vehicles) to periodically recover their positions and speeds. A static roadside unit (RSU) gathers data from passing sensor vehicles. Finally, the speed/position information or space-time velocity (STV) field is then reconstructed in a data fusion center with simple interpolation techniques. We use widely accepted theoretical traffic models (i.e., car-following, multilane, and overtake-enabled models) to replicate the nonlinear characteristics of the STV field in representative situations (congested, free, and transitional traffic). To obtain realistic packet losses, we simulate the multihop ad-hoc wireless network with an IEEE 802.11p PHY layer. We conclude that: 1) for relevant configurations of both sensor vehicle and RSU densities, the wireless multihop channel performance does not critically affect the STV reconstruction error, 2) the system performance is marginally affected by transmission errors for realistic traffic conditions, 3) the STV field can be recovered with minimal mean absolute error for a very small fraction of sensor vehicles (FSV) ≈ 9%, and 4) for that FSV value, the probability that at least one sensor vehicle transits the spatiotemporal regions that contribute the most to reduce the STV reconstruction error sharply tends to 1. Thus, a random and sparse selection of wireless sensor vehicles, in realistic traffic conditions, is sufficient to get an accurate reconstruction of the STV field.

Self-organizing algorithms (SOAs) for wireless sensor networks (WSNs) usually seek to increase the lifetime, to minimize unnecessary transmissions or to maximize the transport capacity. The goal left out in the design of this type of algorithms is the capability of the WSN to ensure an accurate reconstruction of the sensed field while maintaining the self-organization. In this work, we formulate a general framework where the data from the resulting clusters ensures the well-posedness of the signal processing problem in the cluster. We develop the second-order data-coupled clustering (SODCC) algorithm and the distributed compressive-projections principal component analysis (D-CPPCA) algorithm, that use second-order statistics. The condition to form a cluster is that D-CPCCA does not fail to resolve the Principal Components in any given cluster. We show that SODCC is scalable and has similar or better message complexity than other well-known SOAs. We validate these results with extensive computer simulations using an actual LS-WSN. We also show that the performance of SODCC is, comparative to other state-of-the-art SOAs, better at any compression rate and needs no prior adjustment of any parameter. Finally, we show that SODCC compares well to other energy efficient clustering algorithms in terms of energy consumption while excelling in data reconstruction Average SNR.

The optimal configuration for a Large Scale Wireless Sensor Networks (LS-WSN) is the one that minimizes the sampling rate, the CPU time and the channel accesses (thus maximizing the network lifetime), with a controlled distortion in the recovered data. Initial deployments of LS-WSN are usually not able to adapt to changing environments and rarely take into account either the spatial or temporal nature of the sensed variables, both techniques that optimize the network operation. In this work we propose the use of Self-Organized Distributed Compressive Projection (SODCP) in order to let the nodes to form clusters in a distributed and data-driven way, exploiting the spatial correlation of the sensed data. We compare the performance of this innovative technique, using actual data from the LUCE LS-WSN, with two different baselines: Centralized Compressive Projection (CCP) and Distributed Compressive Projection (DCP). The former uses no clustering, whereas the latter makes use of an a priori clustering that favors proximity and balances the number of nodes in each cluster. We show that SODCP outperforms DCP (in terms of Signal-to-Noise vs. Compression Rate). We also show that the performance of SODCP converges to that of CCP for relatively high compression rates of 55%.

In this work we present a Wireless Sensor Network (WSN) system designed for the on-board determination of human gait entropy. The usage of nonlinear entropy-based metrics has proven to be a useful tool for analyzing the complexity of biological systems. The final goal of entropy calculation in this type of biological system is to identify possible causes of future injuries (in order to improve aging) and the early injury detection (ideal for elite athletes). Existing systems for human gait analysis are limited to traditional data gathering, e.g. continuous measurement and wireless transmission to a Data Fusion Center (DFC), due to the computational burden of entropy calculation. In addition, actual systems are likely to interfere the natural movement due to their cumbersome nature. The WSN presented here uses four sensor nodes, located in both ankles and hip sides, and are equipped with triaxial accelerometers. We propose the use of low-complexity algorithms in order to perform onboard entropy determination prior to wireless transmission. The proposed system can be used to reliably determine long-term human gait entropy.

This paper addresses the problem of finding an analytical expression for the end-to-end Average Bit Error Rate (ABER) in multihop Decode-and-Forward(DAF) routes within the context of wireless networks. We provide an analytical recursive expression for the most generic case of any number of hops and any single-hop ABER for every hop in the route. Then, we solve the recursive relationship in two scenarios to obtain simple expressions for the end-to-end ABER, namely: (a) The simplest case, where all the relay channels have identical statistical behaviour; (b) The most general case, where every relay channel has a different statistical behaviour. Along with the theoretical proofs, we test our results against simulations. We then use the previous results to obtain closed analytical expressions for the end-to-end ABER considering DAF relays over Nakagami-m fading channels and with various modulation schemes. We compare these results with the corresponding expressions for Amplify-and-Forward (AAF) and, after corroborating the theoretical results with simulations, we conclude that DAF strategy is more advantageous than the AAF over Nakagami-m fading channels as both the number of relays and m-index increase.

When a terminal is recruited to cooperate with other neighboring terminals, its channel state and carrier frequency offset (CFO) may be unknown to the destination. Under these circumstances, this paper considers the use of distributed double-differential (DD) modulation, which simplifies receiver implementation because it by-passes channel and CFO estimation. Two double-differential codecs are proposed transmitting: i) across orthogonal channels using time-division multiplexing, achieving rate and error performance similar to that of co-located multi-antenna DD systems; or ii) simultaneously, benefiting from the distinct CFOs across terminals and bypassing the need of ordering protocols. Both (i)-(ii) approaches are considered in adaptive- and selective-relaying cooperation protocols demonstrating that maximum spatial diversity is achievable. Simulations corroborate the theoretical error performance claims.