Digital Twin Networks (DTNs) are an emerging paradigm where DTs collaborate to share knowledge and deliver intelligent services. To meet the need for low-latency and localized processing, DTs can be deployed at the network edge. However, effective mechanisms for inter-twin communication and service provisioning remain largely unexplored. In this paper, we address these challenges by leveraging a name-based communication paradigm, namely Information Centric Networking (ICN), enhanced with semantic-awareness. To overcome the limitations of exact name matching, we integrate deep learning models into the ICN forwarding fabric to compute semantic similarity. This enables the discovery of relevant cached service results or routes to DT service providers, even under naming heterogeneity. A preliminary evaluation shows improved service provisioning success and reduced latency compared to standard ICN delivery.

Federated Learning (FL) is gaining momentum as a prominent solution to enable on-device training while preserving data privacy. In a typical FL architecture, the function of aggregating model updates from thousands of client devices, each locally training the shared model on its own data, is placed at a dedicated node within the edge domain. Such a placement decision remains static throughout the entire training process. However, the time-varying, heterogeneous, and limited availability of computing capabilities in edge environments, coupled with fluctuating populations of participating clients across training rounds, calls for dynamic per-round placement of the aggregation function. To further optimize the usage of heterogeneous resources while accounting for FL aggregation tasks, which may require different scales of computing capabilities, a fine-grained allocation of computing resources is needed across edge nodes. The objective of this work is to jointly optimize the FL aggregation function placement at each training round and the allocated computing resources in order to minimize the overall per-round training time while not exceeding the computing capabilities of each edge node. An optimization problem is formulated, and an efficient and effective heuristic algorithm is proposed, based on local search techniques. Experimental results demonstrate that the proposed solution significantly outperforms benchmark approaches, achieving a reduction in per-round training time and bandwidth consumption. Under the considered settings, the proposal yields up to a 27% gain in training time compared to methods that optimize aggregation placement only, without performing fine-grained resource allocation.

In-Network Computing (INC) is emerging as a key enabler of Sixth-Generation (6 G) systems, allowing programmable network nodes to provide not only connectivity, but also storage and processing across the cloud-to-edge continuum. Machine learning (ML) tasks, particularly Deep Neural Network (DNN) inference, stand to benefit significantly from this shift. Under the Split Inference (SI) paradigm, different layers of a DNN can be distributed across multiple in-network nodes that cooperate with the end-device requesting inference. In this work, we explore the potential of Named Data Networking (NDN) as an enabler for in-network SI. We demonstrate how NDN’s native features, such as in-network caching and routing-by name, can reduce inference delays and improve robustness under lossy edge connectivity, compared to traditional host-centric networking. Simulation results validate the effectiveness of NDN-based innetwork SI, highlighting its potential to enable resilient and efficient ML services in future 6 G environments.

Split inference (SI) has been devised as a valuable solution to enable the execution of computation-heavy deep neural network (DNN) inference models on resource-constrained edge devices. The different layers of a DNN model are distributed to one or several nodes (mainly edge/cloud servers) cooperating with the end-device requesting the inference. Boosted by the sixth generation (6 G) trends, programmable network nodes equipped with computing, caching and intelligence capabilities can be involved in such a cooperative task. In this work, we propose Named Data Networking (NDN) as a key enabler of in-network SI. Its connectionless communication model coupled with routing-by-name and native in-network caching capabilities can facilitate dynamic splitting operations on nodes throughout the cloud-to-things continuum. We show how NDN design principles and communication primitives can be leveraged to support in-network SI. Then, preliminary results are reported to showcase the benefits of the conceived proposal.

Federated Learning (FL) is gaining momentum as a promising solution to enable the efficient and privacy-preserving distributed training of Machine Learning (ML) models. Unlike centralized ML solutions, only the ML model and its updates are transferred between the clients and the aggregator server, eliminating the need to share large datasets. Notwithstanding, poor connectivity conditions experienced over the path that interconnects the FL clients and the aggregator server, either due to (wireless) channel losses or congestion, may deteriorate the training convergence. Several methods have been devised to reduce the training duration, primarily by minimizing data transfer through the design of ML algorithms at the application level. However, these solutions still exhibit unsettled issues, as they may only reduce the communication footprint but do not improve the communication process as a whole. Differently, in this work, our aim is to improve FL data exchange from a networking perspective by promoting Information Centric Networking (ICN) approaches rather than host-centric TCP/IP-based solutions. To this aim, we analyse the impact that host-centric transport protocols as well as ICN approaches have on the FL performance, in terms of duration of the model training and exchanged data (model and updates) load, under different channel loss settings. We show that ICN-based FL solutions significantly reduce the network data load and decrease the duration of the training round by up to an order of magnitude for high channel loss rates.

Federated Learning (FL) enables multiple devices (clients) training a shared machine learning (ML) model on local datasets and then sending the updated models to a central server, whose task is aggregating the locally-computed updates and sharing the learned global model again with the clients in an iterative process. The population of clients may change at each round, whereas the node executing the aggregation function is typically placed at an edge domain and remains static until the end of the overall FL training process. Indeed, the computing capabilities of the edge node hosting the aggregation function and the distance (latency) of such a node from the selected clients can highly affect the convergence rate of the FL training procedure. Moreover, the heterogeneous time-varying capabilities of edge nodes, coupled with the dynamic client population selected at each round, call for the optimal dynamic placement of the aggregation function across the available nodes in an edge domain. In this work, we formulate an optimization problem for the placement of the FL aggregation function, which aims to select at each round the edge node able to minimize the overall per-round training time, encompassing the aggregation time, the local training time at the clients and the time for exchanging the global model and the model updates. A time-efficient greedy heuristics is proposed, which is shown to well approximate the optimal solution and outperform the considered benchmark solutions.

Deep Recommender Models (DLRMs) inference is a fundamental AI workload accounting for more than 79% of the total AI workload in Meta's data centers. DLRMs' performance bottleneck is found in the embedding layers, which perform many random memory accesses to retrieve small embedding vectors from tables of various sizes. We propose the design of tailored data flows to speedup embedding look-ups. Namely, we propose four strategies to look up an embedding table effectively on one core, and a framework to automatically map the tables asymmetrically to the multiple cores of a SoC. We assess the effectiveness of our method using the Huawei Ascend AI accelerators, comparing it with the default Ascend compiler, and we perform high-level comparisons with Nvidia A100. Results show a speed-up varying from 1.5x up to 6.5x for real workload distributions, and more than 20x for extremely unbalanced distributions. Furthermore, the method proves to be much more independent of the query distribution than the baseline.

Digital Twin (DT) technology is expected to cover a crucial role in a variety of 6G application scenarios, including smart automotive, smart home and smart city. By leveraging advanced Artificial Intelligence (AI) modules alongside cutting-edge communication and networking architectures, DTs will be able to develop cognitive and social skills and build relationships with each other, thus facilitating the sharing of services and experience. However, in current implementations, DTs typically engage with their physical counterpart only, for predictive main-tenance and optimization, while protocols for inter-twin commu-nications and service discovery are still unexplored. In this paper, we focus on service discovery and provisioning in DT networks hosted at the network edge. In our design, the Social Internet of Things (SIoT) notion is applied to build social networks among DTs and, in parallel, name-based primitives, according to the Information Centric Networking (ICN) paradigm, are consid-ered to support inter-twin interactions. Two distributed name-based service discovery mechanisms are envisioned: a social-driven scheme, leveraging friendship and similarities among DTs' names, and a network-driven scheme, leveraging the ICN forwarding fabric only. A performance evaluation shows the benefits of the conceived solution in terms of reduced discovery latency compared to legacy centralized approaches.

Edge computing is aimed to support compute-intensive data-hungry interactive applications which can hardly run on resource-constrained consumer devices and may suffer from running in the cloud due to the long data transfer delay. The edge network nodes' heterogeneous and limited (compared to the cloud) capabilities make the computing task placement a challenge. In this paper, we propose a novelin-networktask placement strategy aimed at minimizing the edge network resources usage. The proposal specifically accounts fortime-limited reusablecomputing tasks, i.e., tasks whose output can be cached to serve requests from different consumers for a certain time. Caching such results, during their time validity, achieves the twofold benefit of reducing the service provisioning time and improving the edge resource utilization, by avoiding redundant computations and data exchange. The devised strategy is implemented as a network application of a Software-defined Networking Controller in charge of overseeing the edge domain. We formulate the optimal task placement through an integer linear programming problem, and we define an efficient heuristic algorithm that well approximates the solution achieved through a standard optimal solver. Achieved results show that the proposal successfully meets the targeted objectives in a wide variety of simulated scenarios, by outperforming benchmark solutions.

The increasing volume of Internet traffic is pushing the Internet Service Providers to deploy distributed caching services at the network edge, close to the end users, in order to speed up the data retrieval and reduce the bandwidth demands. In parallel, centralized paradigms like Software Defined Networking (SDN) are considered to optimize network management while supporting a variety of network applications like routing, load balancing and caching. In this paper, we extend the SDN control plane to support a novel content caching strategy. We consider a softwarised edge network domain where SDN nodes, augmented with storage capabilities, cache incoming data with the twofold target of limiting the retrieval delay and the inter-domain traffic. The caching decision is taken in a centralized mode by the SDN Controller, according to a newly defined content-driven closeness centrality metric, which identifies the importance of the SDN nodes as cachers based on their proximity to the majority of the clients requesting the most popular contents. Simulation results show the superiority of the solution in terms of higher cache hits and reduced latency, when compared against benchmark caching strategies.

Federated learning (FL) is gaining momentum as a prominent solution to perform training procedures without the need to move sensitive end-user data to a centralized third party server. In FL, models are locally trained at distributed end-devices, acting as clients, and only model updates are transferred from the clients to the aggregator, which is in charge of global model aggregation. Although FL can ensure better privacy preservation than centralized machine learning (ML), it exhibits still some concerns. First, clients need to be properly discovered and selected to ensure that highly accurate models are built. Second, huge models may still require to be exchanged from the aggregator to all the selected clients, incurring a not negligible network footprint. To tackle such issues, in this paper, we propose a framework built upon in-network caching, multicast and name based data delivery, natively provided by the Named Data Networking (NDN) paradigm, in order to support client discovery and aggregator-clients data exchange. Benefits of the proposal are showcased when compared to a conventional application-layer solution.

Technology is continually undergoing a constituent development caused by the appearance of billions new interconnected “things” and their entrenchment in our daily lives. One of the underlying versatile technologies, namely wearables, is able to capture rich contextual information produced by such devices and use it to deliver a legitimately personalized experience. The main aim of this paper is to shed light on the history of wearable devices and provide a state-of-the-art review on the wearable market. Moreover, the paper provides an extensive and diverse classification of wearables, based on various factors, a discussion on wireless communication technologies, architectures, data processing aspects, and market status, as well as a variety of other actual information on wearable technology. Finally, the survey highlights the critical challenges and existing/future solutions.

With more than 75 billions of objects connected by 2025, Internet of Things (IoT) is the catalyst for the digital revolution, contributing to the generation of big amounts of (transient) data, which calls into question the storage and processing performance of the conventional cloud. Moving storage resources at the edge can reduce the data retrieval latency and save core network resources, albeit the actual performance depends on the selected caching policy. Existing edge caching strategies mainly account for the content popularity as crucial decision metric and do not consider the transient feature of IoT data. In this article, we design a caching orchestration mechanism, deployed as a network application on top of a software-defined networking Controller in charge of the edge infrastructure, which accounts for the nodes’ storage capabilities, the network links’ available bandwidth, and the IoT data lifetime and popularity. The policy decideswhich IoT contentshave to be cached andin which nodeof a distributed edge deployment with limited storage resources, with the ultimate aim of minimizing the data retrieval latency. We formulate the optimal content placement through an Integer Linear Programming (ILP) problem and propose a heuristic algorithm to solve it. Results show that the proposal outperforms the considered benchmark solutions in terms of latency and cache hit probability, under all the considered simulation settings.

Software Defined Networking (SDN) and Named Data Networking (NDN) have been recently advocated as complementary paradigms to improve content distribution in the next-generation Internet. On the one hand, SDN offers a centralized control plane that can optimize routing decisions; on the other, the distinctive features at the NDN data plane, such as name-based delivery, in-network caching, and stateful forwarding, simplify data dissemination. In the integrated design, when a request cannot be handled locally at the NDN data plane in the Forwarding Information Base (FIB), the SDN Controller is contacted to inject the forwarding rule. Decisions such as which rules need to be stored in the node and for how long deeply affect the packet forwarding performance. This paper debates about the issues related to forwarding rules in the FIBs of SDN-controlled NDN nodes, by specifically accounting for their name-based nature, representing a key novelty compared to legacy SDN implementations. Quantitative results are reported to showcase the impact of crucial parameters, like the content popularity, the content requests rate, the table size, on the FIB performance in terms of valuable metrics (e.g., hit ratio, rejected requests, incurred signaling with the Controller).

The Multi-access Edge Computing (MEC) paradigm and the Cellular Vehicle-to-Everything (C-V2X) technology prove to be good candidates to address the vehicular applications' demands of high-performing connectivity and low-latency access to computing and storage resources. MEC provides cloud-like resources at the network edge, i.e., at the Multi-access Edge (ME) host. While vehicles move around, a service application instance running on a ME host may be triggered to move to another ME host to better support the application's demands. As a side effect, this migration may undermine service continuity. In this paper, we refer to the latest available ETSI MEC and 3GPP C-V2X specifications to investigate the issue of service migration between ME hosts in the context of vehicular communication. Early experimental results provide measures of the service migration latencies when ME applications run as Docker containers.

The latest advances in mobile devices and the widespread diffusion of networked objects are driving the evolution of traditional Mobile Cloud Computing (MCC) systems toward a new framework where storage, computing, sensing, and other device capabilities are offered as a service at the network edge. This visionary scenario, encompassing heterogeneous resources generated, shared, and consumed everywhere in the network, requires innovative architectural and protocol design. In this context, can the approaches recently formulated in the Future Internet research arena (e.g., middleware-based virtualization, Information Centric Networking, and Software-Defined Networking/Network Function Virtualization) support the evolution of mobile cloud systems? This paper provides an affirmative answer by proposing Future-MCC, a novel architecture that capitalizes on such promising approaches and re-thinks (when needed) their philosophy to better fit the evolution of MCC systems. The performance of Future-MCC has been investigated in a representative heterogeneous and dynamic Smart City scenario. Computer simulation results clearly demonstrate that the proposed solution ensures (i) a reduction of the bandwidth requirements spanning from 66 to 91 percent and (ii) an average energy saving equal to 99 percent with respect to a conventional cloud computing platform.

Pushed by the challenging demands of fifth generation (5G) use cases and the recent advancements in virtualization technologies, edge network devices are rapidly evolving from simple forwarders to softwarized infrastructures augmented with computing and storage capabilities. Such a trend blurs the distinction between IT and telco domains and paves the way for the integrated management of network and computing resources. In this paper, we focus on the interplay of the Software Defined Networking (SDN) and Named Data Networking (NDN) paradigms as key drivers for the orchestration of computing services in softwarized edge infrastructures. We propose a new framework that brings out the best of the SDN centralized intelligence to take “smart” decisions and inject rules for service allocation (e.g., retrieve input data, execute a function over them), and the best of the adaptive NDN forwarding plane and its native in-network caching to request and deliver services by name. Within the framework, we devise a service allocation strategy that aims at selecting as an executor, among the potential candidates, the one which is able to guarantee the shortest service provisioning delay for each service request, while accounting for the network topology, the links status, and the available computing resources in the edge nodes. Performance evaluations testify to the superiority of our proposal against benchmarking solutions from the literature under different operation settings.

Edge computing is a key paradigm to offload the core network and effectively process massive Internet of Things (IoT) raw data without sending them to the cloud. This paradigm normally relies on a set of purpose-built and pre-planned servers, which host storage and processing resources to provide IoT services close to the data sources, thus saving core network resources and offloading the remote cloud infrastructure. In this paper, we propose to turn the network edge into a dynamic, distributed computing environment that supports the provisioning of IoT services, by exploiting the recent evolution of named data networking (NDN), supporting both name-based data retrieval and computation. Specific name structure and novel NDN forwarding mechanisms are designed; a distributed strategy is also engineered to select the service executor among edge nodes, with the objectives to: 1) limit the raw IoT data traffic crossing the network and 2) allocate the service execution according to the nodes' available processing resources. Numerical analysis shows that the performance of the proposed framework approaches the one of the optimal solution of a formulated integer linear programming problem. System-level ndnSIM simulations confirm that the proposal also outperforms the considered state-of-the-art benchmark solutions in terms of service provisioning time.

Named Data Networking (NDN) has been recently extended to enable the discovery and provisioning by name of in-network cloud-like services such as processing and data storage. Such a feature is particularly helpful in distributed edge environments, where mobile end devices can be involved in service offering. In such a context, a challenging issue is to motivate a mobile device to behave as a provider and share its resources (e.g., CPU, memory) in order to assist other end devices (consumers), in wireless proximity, asking for a given service. In this paper, we propose a solution revolving around two concepts: enhanced NDN primitives and a social-driven stimulus for mobile devices to volunteer as service providers. A bio-inspired response function is used by a potential provider's device to rate both its available resources and the social ties with the current consumer's device, established according to the Social Internet of Things (SIoT) paradigm. An early evaluation showcases to which extent the conceived solution allows a consumer to find a nearby provider available to offer its services, under different social neighbourhood settings.

Large-scale smart environments (LSEs) are open and dynamic systems typically extending over a wide area and including a huge number of interacting devices with a heterogeneous nature. Thus, during their deployment scalability and interoperability are key requirements to be definitely taken into account. To these, discovery and reputation assessment of services and objects have to be added, given that new devices and functionalities continuously join LSEs. In spite of the increasing interest in this topic, effective approaches to develop LSEs are still missing. This paper proposes an agent-based approach that leverages edge computing and Social Internet of Things paradigms in order to address the above mentioned issues. The effectiveness of such an approach is assessed through a sample case study involving a commercial road environment.

Recently, Internet-of-Things (IoT) applications have been moving toward a network of intelligent objects with social capabilities, defined as the Social IoT (SIoT). Different types of relationships exist among things, thus forming social connections such as parental-object relationship (POR), co-work-object relationship (CWOR), co-location-object relationship, ownership-object relationship, and so on.

Modern vehicles are equipped with several interconnected sensors on board for monitoring and diagnosis purposes; their availability is a main driver for the development of novel applications in the smart vehicle domain. In this paper, we propose a Docker container-based platform as solution for implementing customized smart car applications. Through a proof-of-concept prototype-developed on a Raspberry Pi3 board-we show that a container-based virtualization approach is not only viable but also effective and flexible in the management of several parallel processes running on On Board Unit. More specifically, the platform can take priority-based decisions by handling multiple inputs, e.g., data from the CANbus based on the OBD II codes, video from the on-board webcam, and so on. Results are promising for the development of future in-vehicle virtualized platforms.

Leveraging the concept of controlled node mobility in this paper we develop an algorithm for tracking and controlling proximity malware propagation in Wireless Sensor Networks (WSN). Our proposal aims at: (i) notifying the nodes of malwarepropagation, (ii) leading a flying robot along a path in the WSNthat guarantees the minimum recovery time to (iii) heal the infected nodes. We formulate the targeted curing problem as a binary integer problem and determine the optimal solution by a central solver. We use the analytical result as a benchmark to evaluate the recovery time of the proposed solution. The achieved results show a satisfactory performance in terms of tracking the presence of an ongoing epidemic and healing the nodes.

Named Data Networking (NDN) is a new architecture for the Future Internet that supports efficient data delivery by directly using content names instead of IP addresses. This content-centric approach can be especially useful for IoT applications, where everyday objects embedding wirelessly connected constrained devices should be accessed, by both other machines and users. In such a context, we focus on the smart home domain and propose a framework called eDomus that (i) leverages functionalities by a popular social network, i.e., Facebook, to allow a user to remotely interact with the home network and (ii) is augmented with NDN concepts to properly monitor the domestic environment. In this paper, we present our framework and our preliminary prototype, by also providing some hints on future work.

Content-Centric Networking (CCN) provides a complete communication framework for data retrieval and dissemination. It leverages innovative naming, security schemes and novel routing strategies, augmented with caching at intermediate nodes. Content requests are forwarded towards the source(s) by direct use of content names (instead of IP addresses), matching the layout of the Wireless Sensor Network. Despite its potential CCN can not be directly applied to wireless environments, specifically resource-constrained sensor nodes. In this paper, a two-tier CCN architecture is proposed to manage the heterogeneity of involved devices (remote server, sink, sensor nodes). CCN is enhanced with some changes to the forwarding strategies to improve data collection.

In this paper, we address the problem of re-establishing the network connectivity in post-disaster scenarios, where the original wireless infrastructure has been partitioned into multiple network fragments (called islands), operating on different frequencies. To this purpose, we propose the utilization of swarms of dedicated repairing units, called Stem-Nodes (SNs). SNs are provided with Cognitive Radio (CR) and self-positioning capabilities, in order to offer maximum reconfigurability in terms of mobility and wireless technologies supported. Moreover, swarms of SNs can self-organize into STEM-Mesh structure, that works as a dynamic backbone to connect heterogeneous islands using different technologies (e.g. Wi-Fi, Wi-MAX, etc). In this paper, we present three contributions pertaining to STEM-Mesh: (i) we describe a distributed motion control scheme (based on virtual springs approach) that enables SNs to self-organize into dynamic STEM-Mesh structures, (ii) we introduce a discovery scheme, through which SNs can explore the scenario in both spatial and frequency domains, and possibly connect the islands to the STEM-Mesh backbone and (iii) we validate the correctness of the proposed scheme, by verifying the optimal placements of the SNs composing the STEM-Mesh on a simplified scenario (e.g. chain topology). Finally, we evaluate through Omnet++ simulations the ability of STEM-Mesh to maximally re-establish connectivity on partitioned network scenarios.

Content-centric is an emerging networking paradigm conceived for the future Internet. Data retrieval and distribution are based on content names instead of host addresses. In this paper we propose a content-centric architecture for energy-efficient multihop communications in a wireless mesh network and compare its performances against a legacy IP-based approach.

To provide QoS guarantees to a variety of traffic types is becoming a key feature for the success of wireless LANs. In this paper, we contribute to this issue by proposing a framework under which the nodes in an IEEE 802.11e ad hoc network may cooperate to provide the user with the desired QoS level

This paper contributes to the issue of QoS differentiation in wireless ad hoc networks. The focus is on the IEEE 802.11e MAC protocol that we propose to enhance by introducing a mechanism to "dynamically" assign priorities to traffic. Priorities are hop-by-hop assigned according to network resource availability and load in order to satisfy throughput requirements of each class of traffic

In wireless LANs, QoS provisioning and multimedia traffic support are critical elements for the successful deployment of such networks. In this paper, we propose a dynamic mechanism to adapt data rate and priority of multimedia wireless stations equipped with IEEE 802.11e network cards. We show how the introduced dynamicity is able to improve performance of the enhanced distributed channel access (EDCA) of IEEE 802.11e networks.

The recent introduction of new variable bit rate (VBR) speech coders has opened up new perspectives for the implementation of adaptive voice over IP (AVoIP) systems. The paper compares different VBR speech coding techniques in a scenario in which the rate of the single sources is dynamically adapted to the workload conditions. The coders compared are the AMR, the M/sup 3/R and the G.729. Using source and rate control mechanism models, performance was evaluated in terms of loss probability, offered throughput and mean CMOS (comparison mean opinion score) with varying numbers of sources and different background noise conditions. The use of header compression mechanisms was also evaluated.

In the near future, congestion control should be introduced for multimedia UDP-based traffic, in such a way that this traffic becomes "TCP-friendly". To this end, several TCP-friendly algorithms have been proposed in the literature. However, although these algorithms were introduced to support real-time applications on the Internet, the only target in optimizing them, until now, was to achieve fairness with TCP flows in the network. No attention has been paid to the quality of service (QoS) perceived by their users. The target of this paper is to analyze the problem of transmitting voice over IP when voice sources use two of the most promising TCP-friendly algorithms, RAP (rate adaptation protocol) and TFRC (TCP-friendly rate control). With this aim, a VoIP system architecture is introduced and the characteristics of each of its elements are discussed. To optimize the system, a voice multirate encoder, which is able to work over a TCP layer, is used, and a modification of both RAP and TFRC is proposed. Finally, in order to analyze the performance of the proposed system architecture and to compare the modified RAP and TFRC with the original algorithms, the sources have been modeled with an arrival process modulated by a Markov chain, and the model has been used to generate traffic in a simulation study performed with the ns-2 network simulator.

The paper proposes a performance evaluation and comparison of G.729, AMR, and fuzzy voice activity detection (FVAD) algorithms. The comparison was made using objective, psychoacoustic, and subjective parameters. A highly varied speech database was also set up to evaluate the extent to which VADs depend on language, the signal-to-noise ratio (SNR), or the power level.

The paper proposes a performance evaluation and comparison of recent ITU-T and ETSI voice activity detection algorithms. The comparison was made using both objective and psychoacoustic parameters, so as to have reliable judgements that were close to subjective ones. A highly varied speech database was also set up to evaluate the extent to which VAD depend on language, the signal to noise ratio, or the power level.

This paper presents a hybrid multi-mode/multi-rate, toll quality CS-ACELP coder developed for voice over IP applications. The coder uses coding modes compatible with the three 6.4, 8, and 11.8 kbit/s coding schemes standardised by ITU-T in G.729. In particular, the algorithm presents 4 coding categories, with an average bit rate ranging between about 3 and 8 kbit/s, that adapt the rate to changes in network conditions.