In speech processing applications, e.g., speech recognition, hearing aids (HAs), video conferencing, and human-computer interaction, speech enhancement or noise reduction is an essential front-end task, as the recorded speech signals are inevitably corrupted by interference, including coherent/incoherent noise and reverberation. Traditional noise reduction algorithms are mostly based on spatial filtering techniques using a microphone array. The performance of the noise reduction algorithms scales with the number of microphones that are involved in filtering, but a large-sized microphone array cannot be mounted in many realistic systems, e.g., HAs. In the last few decades, with a great development in micro-electro-mechanical systems, wireless devices are more and more commonly-used in our daily life, like the smartphone, laptop, wireless HA, and ipad. These devices have acoustic sensors equipped and a capability of wireless communication, leading to a wireless acoustic sensor network (WASN). The WASN can be organized in a centralized fashion where all the devices are only allowed to connect with a fusion center (FC), or in a decentralized way where the devices are connected with the close-by counterparts via wireless links. ThisWASN can resolve the disadvantages of the traditional microphone array systems, since thewireless devices can be placed anywhere in the vicinity and one device is able to make use of measurements from other external devices. More importantly, the acoustic scene can be sampled more comprehensively, resulting in a potential improvement in noise reduction performance.@en

Power usage is an important aspect of wireless acoustic sensor networks (WASNs) and reducing the amount of information that is to be transmitted is one effective way to save it. In previous contributions, we presented sensor selection as well as rate distribution methods to reduce the power usage of beamforming algorithms in WASNs. Taking only transmission power into account, it was shown that rate distribution is a generalization of sensor selection and that rate distribution is more efficient than sensor selection with respect to the power usage versus performance trade-off. However, this excludes the energy consumption that it takes to keep the WASN nodes activated. In this paper, we present a more detailed comparison between sensor selection and rate-allocation by taking also into account the power to keep sensors activated for centralized WASNs. The framework is formulated by minimizing the total power usage, while lower bounding the noise reduction performance. Numerical results show that whether rate distribution is more efficient than sensor selection depends on the actual power that is used to keep sensors activated.

@en

In this letter, we propose a decentralized framework for rate-distributed linearly constrained minimum variance (LCMV) beamforming in wireless acoustic sensor networks. To save the energy usage within the network, we propose to minimize the transmission cost and put a constraint on the noise reduction performance. Subsequently, we decentralize the obtained LCMV filter structure by exploiting an imposed block diagonal form of the noise correlation matrix. As a result, the beamformer weights are calculated in a decentralized fashion and each node can determine its quantization rate locally. Finally, numerical results validate the proposed method.

@en

Glucosinolates can be hydrolyzed by the enzyme commonly known as myrosinase (E.C. 3.2.1.147)to a variety of biological compounds. Myrosinase (MYR)has been immobilized through the flexible spacers of different length on cross-linked chitosan resin (CCR). Ethylene diamine, hexamethylenediamine and decamethylene-diamine have been separately used as spacers. The influence of the flexible spacer length on the immobilized MYR (IMYR)properties were evaluated. The optimum pH and V _max of IMYR linearly increase with the flexible spacer length, and the optimum temperature and K _m of IMYR show an opposite trend. The recyclability of IMYR was good, with 90% recovery of activity after 10 cycles and 80% recovery after 30 cycles. IMYR was highly stable under storage conditions, with 95% recovery of activity after one year storage at 4 °C. The IMYR with the longest flexible spacer, decamethylene-diamine, was used as a biocatalyst for sulforaphene production. The overall hydrolysis ratio of glucoraphenin was 93.25 ± 0.91% and the activity of DDMCCR-IMYR remained 95% after 10 days of continuous use.

@en

In this paper, we present an algorithm to estimate the relative acoustic transfer function (RTF) of a target source in wireless acoustic sensor networks (WASNs). Two well-known methods to estimate the RTF are the covariance subtraction (CS) method and the covariance whitening (CW) approach, the latter based on the generalized eigenvalue decomposition. Both methods depend on the use of the noisy correlation matrix, which, in practice, has to be estimated using limited and (in WASNs) quantized data. The bit rate and the fact that we use limited data records therefore directly affect the accuracy of the estimated RTFs. Therefore, we first theoretically analyze the estimation performance of the two approaches in terms of bit rate. Second, we propose a rate-distribution method by minimizing the power usage and constraining the expected estimation error for both RTF estimators. The optimal rate distributions are found by using convex optimization techniques. The model-based methods, however, are impractical due to the dependence on the true RTFs. We therefore further develop two greedy rate-distribution methods for both approaches. Finally, numerical simulations on synthetic data and real audio recordings show the superiority of the proposed approaches in power usage compared to uniform rate allocation. We find that in order to satisfy the same RTF estimation accuracy, the rate-distributed CW methods consume much less transmission energy than the CS-based methods.

@en

In wireless acoustic sensor networks (WASNs), sensors typically have a limited energy budget as they are often battery driven. Energy efficiency is therefore essential to the design of algorithms in WASNs. One way to reduce energy costs is to only select the sensors which are most informative, a problem known as sensor selection. In this way, only sensors that significantly contribute to the task at hand will be involved. In this work, we consider a more general approach, which is based on rate-distributed spatial filtering. Together with the distance over which transmission takes place, bit rate directly influences the energy consumption. We try to minimize the battery usage due to transmission, while constraining the noise reduction performance. This results in an efficient rate allocation strategy, which depends on the underlying signal statistics, as well as the distance from sensors to a fusion center (FC). Under the utilization of a linearly constrained minimum variance (LCMV) beamformer, the problem is derived as a semi-definite program. Furthermore, we show that rate allocation is more general than sensor selection, and sensor selection can be seen as a special case of the presented rate-allocation solution, e.g., the best microphone subset can be determined by thresholding the rates. Finally, numerical simulations for the application of estimating several target sources in a WASN demonstrate that the proposed method outperforms the microphone subset selection based approaches in the sense of energy usage, and we find that the sensors close to the FC and close to point sources are allocated with higher rates.

@en

Knowledge graphs (KGs) have proven to be effective to improve recommendation. Existing methods mainly rely on hand-engineered features from KGs (e.g., meta paths), which requires domain knowledge. This paper presents RKGE, a KG embedding approach that automatically learns semantic representations of both entities and paths between entities for characterizing user preferences towards items. Specifically, RKGE employs a novel recurrent network architecture that contains a batch of recurrent networks to model the semantics of paths linking a same entity pair, which are seamlessly fused into recommendation. It further employs a pooling operator to discriminate the saliency of different paths in characterizing user preferences towards items. Extensive validation on real-world datasets shows the superiority of RKGE against state-of-the-art methods. Furthermore, we show that RKGE provides meaningful explanations for recommendation results.

@en

In large-scale wireless acoustic sensor networks (WASNs), many of the sensors will only have a marginal contribution to a certain estimation task. Involving all sensors increases the energy budget unnecessarily and decreases the lifetime of the WASN. Using microphone subset selection, also termed as sensor selection, the most informative sensors can be chosen from a set of candidate sensors to achieve a prescribed inference performance. In this paper, we consider microphone subset selection for minimum variance distortionless response (MVDR) beamformer based noise reduction. The best subset of sensors is determined by minimizing the transmission cost while constraining the output noise power (or signal-to-noise ratio). Assuming the statistical information on correlation matrices of the sensor measurements is available, the sensor selection problem for this model-driven scheme is first solved by utilizing convex optimization techniques. In addition, to avoid estimating the statistics related to all the candidate sensors beforehand, we also propose a data-driven approach to select the best subset using a greedy strategy. The performance of the greedy algorithm converges to that of the model-driven method, while it displays advantages in dynamic scenarios as well as on computational complexity. Compared to a sparse MVDR or radius-based beamformer, experiments show that the proposed methods can guarantee the desired performance with significantly less transmission costs.

@en

In this paper, we propose a rate-distributed linearly constrained minimum variance (LCMV) beamformer for joint noise reduction and spatial cue preservation for assistive hearing in wireless acoustic sensor networks (WASNs). The WASN can consist of wireless communicating hearing AIDS, extended with additional wireless microphones. Due to the fact that each sensor node has a limited power budget, it is essential to consider the energy usage when designing algorithms for such WASNs. As the energy usage in terms of data transmission is directly affected by the communication rate, the proposed method optimally distributes the bit rate for each microphone node. The rate distribution is obtained by minimizing the total transmission costs under constraints on the noise reduction performance and spatial cue preservation of interfering sources. In contrast to sensor selection, i.e., binary decisions on the usefulness of a node, rate distribution allows for soft decisions, and, will lead to more degrees of freedom for joint noise reduction and spatial cue preservation than sensor selection. Numerical results show that given a certain noise reduction requirement, the proposed method displays improved energy efficiency and can preserve the spatial cues of more interferers compared to sensor selection approaches.

@en

The "International Workshop on Recommender Systems for Citizens" (CitRec) is focused on a novel type of recommender systems both in terms of ownership and purpose: recommender systems run by citizens and serving society as a whole.

@en

Binaural sound source localization is an important technique for speech enhancement, video conferencing, and human-robot interaction, etc. However, in realistic scenarios, the reverberation and environmental noise would degrade the precision of sound direction estimation. Therefore, reliable sound localization is essential to practical applications. To deal with these disturbances, this paper presents a novel binaural sound source localization approach based on reverberation weighting and generalized parametric mapping. First, the reverberation weighting as a preprocessing stage, is used to separately suppress the early and late reverberation, while preserving interaural cues. Then, two binaural cues, i.e., interaural time and intensity differences, are extracted from the frequency-domain representations of dereverberated binaural signals for the online localization. Their corresponding templates are established using the training data. Furthermore, the generalized parametric mapping is proposed to build a generalized parametric model for describing relationships between azimuth and binaural cues analytically. Finally, a two-step sound localization process is introduced to refine azimuth estimation based on the generalized parametric model and template matching. Experiments in both simulated and real scenarios validate that the proposed method can achieve better localization performance compared to state-of-the-art methods.

@en

Auto-localization in wireless acoustic sensor networks (WASNs) can be achieved by time-of-arrival (TOA) measurements between sensors and sources. Most existing approaches are centralized, and they require a fusion center to communicate with other nodes. In practice, WASN topologies are time-varying with nodes joining or leaving the network, which poses scalability issues for such algorithms. In particular, for an increasing number of nodes, the total transmission power required to reach the fusion center increases. Therefore, in order to facilitate scalability, we present a structured total least squares (STLS) based internal delay estimation for distributed microphone localization where the internal delay refers to the time taken for a source signal reaching a sensor to that it is registered as received by the capture device. Each node only needs to communicate with its neighbors instead of with a remote host, and they run an STLS algorithm locally to estimate local internal delays and positions (i.e., its own and those of its neighbors), such that the original centralized computation is divided into many subproblems. Experiments demonstrate that the decentralized internal delay estimation converges to the centralized results with increasing signal-to-noise ratio (SNR). More importantly, less computational complexity and transmission power are required to obtain comparable localization accuracy.@en

Randomized gossip (RG) based distributed averaging has been popular for wireless sensor networks (WSNs) in multiple areas. With RG, randomly two adjacent nodes are selected to communicate and exchange information iteratively until consensus is reached. One way to improve the convergence speed of RG is to use greedy gossip with eavesdropping (GGE). Instead of randomly selecting two nodes, GGE selects the two nodes based on the maximum difference between nodes in each iteration. To further increase the convergence speed in terms of transmissions, we present in this paper a synchronous version of the GGE algorithm, called greedy gossip with synchronous communication (GGwSC). The presented algorithm allows multiple node pairs to exchange their values synchronously. Because of the selection criterion of the maximum difference between the values at the nodes, there is at least one node pair with different information, such that the relative error must be reduced after each iteration. The convergence rate in terms of the number of transmissions is demonstrated to be improved compared to GGE. Experimental results validate that the proposed GGwSC is quite e↵ective for the random geometric graph (RGG) as well as for several other special network topologies.@en