While transformers demonstrate outstanding performance across various audio tasks, their application to neural vocoders remains challenging. Neural vocoders require the generation of long audio signals at the sample level, which demands high temporal resolution. This results in significant computational costs for attention map generation and limits their ability to efficiently process both global and local information. In addition, the sequential nature of sample generation in neural vocoders poses difficulties for real-time processing, making the direct adoption of transformers impractical. To address these challenges, we propose RingFormer, a neural vocoder that incorporates the ring attention mechanism into a lightweight transformer variant, the convolution-augmented transformer. Ring attention effectively captures local details while integrating global information, making it well suited for processing long sequences and enabling real-time audio generation. RingFormer is trained using adversarial training with two discriminators. It is integrated as a replacement for the decoder module in the variational inference with adversarial learning for end-to-end text-to-speech (VITS) text-to-speech framework, enabling fair comparisons with state-of-the-art vocoders such as HiFi-GAN, iSTFT-Net, and BigVGAN under identical conditions using various objective and subjective metrics. Experimental results show that RingFormer achieves comparable or superior performance to existing models, particularly excelling in real-time audio generation.

IEEE 802.15.4 low-rate WPAN defines GTS (Guaranteed Time Slot) allocation based on FCFS (First Come First Serve) basis. In case of an emergency traffic, however, it may be hard to receive time slots because of other traffic flows already in service. In this paper we propose an efficient way of allocating GTS to emergency data so that the emergency traffic can get time slots whenever they request. Traffic with a higher data rate is preferred. Extensive simulation results show that our proposed scheme gives a guaranteed GTS allocation to urgent nodes and also enhances overall throughput.

Abstract In this paper, we propose a practical bandwidth reprovisioning scheme in an integrated third generation (3G) andwireless local area network(WLAN) environment. We investigate how to increase available bandwidth for conversational class while keeping the packet loss probability of background class below a threshold. The proposed scheme is implemented as a simple network management protocol (SNMP) compliant management information base (MIB) and deployed atpacket data serving node(PDSN) equipment. The experiments conducted on1x evolution for data and voice(1xEV‐DV) testbed show that the proposed scheme increases the bandwidth of the conversational class and guarantee adequate service quality of the background class. Copyright © 2005 John Wiley & Sons, Ltd.

Current network management needs an end-to-end overview of various connections rather than the information that is purely local to the individual devices. The typical manager-centric polling approach, however, is not adequate to understand network-wide behavior of a large-scale broadband network. We propose a new management information base (MIB) approach, called operation, administration, and maintenance (OAM) MIB. The MIB provides a network manager with dynamic end-to-end management information by utilizing special standard ATM cells. The MIB makes end-to-end management feasible while it reduces management-related traffic and manager-to-manager interactions. In our model, a customer network management system accesses the MIB through M1/M2 reference points of the ATM Forum management architecture with simple network management protocol (SNMP).

A QoS restoration scheme for VP-based ATM networks is proposed and its related characteristics are evaluated. The proposed scheme is a state-independent restoration scheme using a disjoint path group. A VP assignment and capacity planning problem using a disjoint path group are formulated. When a failure occurs, failed working VP are switched to protection paths with the same QoS level using a simple and fast restoration algorithm. The simulated restoration characteristics of the proposed algorithm are presented. A scheme of excess capacity assignment to a protection path for dynamic traffic demands is also presented and analyzed.

Asynchronous transfer mode (ATM) network management is an urgent issue for the fast and successful proliferation of ATM networks. Especially, performance management of ATM virtual connections is urgently required because a different quality of service should be guaranteed on each connection. In most cases, the customer's virtual connections traverse multiple subnetworks, therefore, performance information related to specific user connection is distributed in the networks. This makes performance management of ATM network more difficult. In order to obtain this distributed performance information with current ATM-related management information bases (MIBs), interactions among managers and frequent polling of switches are inevitable. This frequent polling and manager to manager interactions may cause management traffic and data retrieval time to increase. In this article, we propose a performance management method using OAM MIB. The MIB is a set of managed objects which have the information provided by OAM (Operation Administration and Maintenance) cells. The basic concept and structure of the MIB are similar to the RMON MIB. By using the OAM MIB, management traffic and manager to manager interactions are much reduced. In our model, customer network management system (NMS) accesses the MIB via M1/M2 reference points of the ATM Forum management architecture with simple network management protocol (SNMP).

In this presentation, we introduce a fast restoration scheme against a link or node failure. Specifically, we apply this scheme to VP (Virtual Path) restoration in ATM (Asynchronous Transfer Mode) networks. Proposed restoration scheme operates on VP-based ATM network. This scheme is a distributed algorithm performed concurrently by network nodes neighboring failed node or link. When a node or link fails, restoration is performed according to the pre-assigned local, regional, and end-to-end restoration scheme. For this, we assigned a restoration level to each network nodes and VPs, respectively. Also, we devise and construct a VP loop, and fast restoration is performed along this loop. This loop is expanded when a failure occurs, and this expanded loop is used for the restoration of the failed VP connection. We define the required VP loop architecture, restoration messages, and restoration procedures. Proposed algorithm is simulated and compared with general flooding restoration scheme. Results show that the proposed loop-expanding restoration scheme can achieve fast restoration goal with a lower burden to network.