Polychronis Koutsakis

Street furniture such as bins, seats and bus shelters can become “smart” with the inclusion of wireless sensor nodes, which consist of environmental sensors, wireless modules, processors and microcontrollers. One of the most crucial challenges for smart street furniture is how to manage power consumption efficiently without affecting data freshness. In this work, we propose a novel Wireless Sensor Network (WSN)architecture for smart street furniture. Unlike existing WSNs which are based on a one-way communication model between wireless sensor nodes and the server, the proposed architecture employs a two-way communication model and a dynamic adaptation of the time interval of measurements to balance between power consumption and data updates. Our approach also provides a real-time low-power design for wireless sensor nodes which efficiently communicate the updated data instead of sending the same data on a regular basis. To the best of our knowledge, this is the first work in the relevant literature which extends the functionality of the wireless module in wireless sensor nodes to act not only as a station sending environmental data but also as soft Access Point (AP), sensing MAC addresses and WiFi signal strengths from surrounding WiFi-enabled devices. We have conducted experiments on the Murdoch University campus and our results show that our proposal improves lifetime of wireless sensor nodes up to 293% compared to static architectures similar to the ones that have been proposed in the literature. Moreover, network bandwidth is improved up to 38% without affecting data freshness. Finally, storage space for the database at the server is reduced up to 99%.

The strain imposed by the bandwidth demands of multimedia applications on wired and wireless networks calls for efficient novel solutions to the problem of network resource allocation, to avoid significant packet losses. In this letter, we focus on a large variety of MPEG-4, H.264 and H.265-encoded video traces. We use the metaheuristic technique of Simulated Annealing to predict the size of B-frames, and compare it against an existing approach from the literature. B-frame size prediction can be used in order to reduce bandwidth requirements and smoothen the encoded video stream, by selective B-frame dropping. We show that Simulated Annealing can significantly improve the prediction accuracy.

Studies on the composition and nature of Internet protocols are crucial for continued research and innovation. This study used three different methods to investigate the presence and level of support for various Internet protocols. Internet traffic entering and exiting a university network was passively captured, anonymised and analysed to test protocol usage. Active tests probed the Internet's most popular websites and experiments on the default behaviour of popular client, server and mobile operating systems were performed to reconcile the findings of the passive data collection. These results are valuable to research areas, such as those using emulations and simulations, where realism is dependent on the accuracy of the underlying assumptions about Internet traffic. Prior work is leveraged to explore changes and protocol adoption trends. This study shows that the majority of Internet traffic is now encrypted. There has also been an increase in large UDP frames, which we attribute to the Google QUIC protocol. Support for TCP options such as Selective Acknowledgements (SACK) and Maximum Segment Size (MSS) can now be assumed. Explicit Congestion Notification (ECN) usage is still marginal, yet active measurement shows that many servers will support the protocol if requested. Recent IETF standards such as Multipath TCP and TCP Fast Open have small but measurable levels of adoption.

Motivated by recent advances in the area of Compo-sitional Distributional Semantic Models (CDSMs), we propose a compositional approach for estimating continuous affective ratings for adjective-noun (AN) and noun-noun (NN) pairs. The ratings are computed for the three basic dimensions of continuous affective spaces, namely, valence, arousal and dominance. We propose that similarly to the semantic modification that underlies CDSMs, affective modification may occur within the framework of affec-tive spaces, especially when the constituent words of the linguistic structures under investigation form modifier-head pairs (e.g., AN and NN). The affective content of the entire structure is determined from the interaction between the respective constituents, i.e., the affect conveyed by the head is altered by the modifier. In addition, we investigate the fusion of the proposed model with the semantic-affective model proposed in (Malandrakis et al., 2013) applied both at word-and phrase-level. The automatically computed affective ratings were evaluated against human ratings in terms of correlation. The most accurate estimates are achieved via fusion and absolute performance improvement up to 5% and 4% is reported for NN and AN, respectively.

Starting from the English affective lexicon ANEW (Bradley and Lang, 1999a) we have created the first Greek affective lexicon. It contains human ratings for the three continuous affective dimensions of valence, arousal and dominance for 1034 words. The Greek affective lexicon is compared with affective lexica in English, Spanish and Portuguese. The lexicon is automatically expanded by selecting a small number of manually annotated words to bootstrap the process of estimating affective ratings of unknown words. We experimented with the parameters of the semantic-affective model in order to investigate their impact to its performance, which reaches 85% binary classification accuracy (positive vs. negative ratings). We share the Greek affective lexicon that consists of 1034 words and the automatically expanded Greek affective lexicon that contains 407K words.

We investigate an affective saliency approach for speech emotion recognition of spoken dialogue utterances that estimates the amount of emotional information over time. The proposed saliency approach uses a regression model that combines features extracted from the acoustic signal and the posteriors of a segment-level classifier to obtain frame or segment-level ratings. The affective saliency model is trained using a minimum classification error (MCE) criterion that learns the weights by optimizing an objective loss function related to the classification error rate of the emotion recognition system. Affective saliency scores are then used to weight the contribution of frame-level posteriors and/or features to the speech emotion classification decision. The algorithm is evaluated for the task of anger detection on four call-center datasets for two languages, Greek and English, with good results.

E-mail has become a de-facto means of communication. Mail servers try to manage the explosive growth of e-mail usage and offer users good quality of service, while spam e-mails are expected to account for 90% of the e-mail traffic. The exceedingly heavy workload can lead to the replacement of existing e-mail servers due to their inability to cope with performance standards and storing capacity. In this study, we focus on modeling the workload of the email servers of a medium-sized Greek university, for all types of traffic (user and system e-mails, as well as spam). We collected a vast amount of e-mail logs with high variations in terms of size and volume over time. We tested some of the most popular distributions for workload characterization and used powerful statistical tests to evaluate our findings. Interestingly we come to different conclusions in comparison with previous works in the field. Our work indicates that, with the exception of some outliers, campus email traffic can be modeled and predicted quite accurately.

We propose and evaluate the use of an affective-semantic model to expand the affective lexica of German, Greek, English, Spanish and Portuguese. Motivated by the assumption that semantic similarity implies affective similarity, we use word level semantic similarity scores as semantic features to estimate their corresponding affective scores. Various context-based semantic similarity metrics are investigated using contextual features that include both words and character n-grams. The model produces continuous affective ratings in three dimensions (valence, arousal and dominance) for all five languages, achieving consistent performance. We achieve classification accuracy (valence polarity task) between 85% and 91% for all five languages. For morphologically rich languages the proposed use of character n-grams is shown to improve performance.