Y-Net: A Dual Path Model for High Accuracy Blind Source Separation

Wu, Huanzhuo; He, Jingwen; Tömösközi, Máté; Fitzek, Frank H. P.

doi:10.1109/gcwkshps50303.2020.9367428

Cited by 7 publications

(3 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As illustrated in Fig. 3, it follows three strategies: (i) depth-wise residual block [13] to reduce the computation; (ii) a large downsampling rate of four to squeeze the input data; (iii) a dual path [14] structure to obtain enriched semantic and temporal domain information and to obtain a balance between accuracy and size.…”

Section: A a Lightweight Deep Neural Network Model For Networkmentioning

confidence: 99%

In-Network Processing for Low-Latency Industrial Anomaly Detection in Softwarized Networks

Wu¹,

He²,

Tömösközi³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Modern manufacturers are currently undertaking the integration of novel digital technologies -such as 5Gbased wireless networks, the Internet of Things (IoT), and cloud computing -to elevate their production process to a brand new level, the level of smart factories. In the setting of a modern smart factory, time-critical applications are increasingly important to facilitate efficient and safe production. However, these applications suffer from delays in data transmission and processing due to the high density of wireless sensors and the large volumes of data that they generate. As the advent of nextgeneration networks has made network nodes intelligent and capable of handling multiple network functions, the increased computational power of the nodes makes it possible to offload some of the computational overhead. In this paper, we show for the first time our IA-Net-Lite industrial anomaly detection system with the novel capability of in-network data processing. IA-Net-Lite utilizes intelligent network devices to combine data transmission and processing, as well as to progressively filter redundant data in order to optimize service latency. By testing in a practical network emulator, we showed that the proposed approach can reduce the service latency by up to 40%. Moreover, the benefits of our approach could potentially be exploited in other large-volume and artificial intelligence applications.

show abstract

Section: A a Lightweight Deep Neural Network Model For Networkmentioning

confidence: 99%

In-Network Processing for Low-Latency Industrial Anomaly Detection in Softwarized Networks

Wu¹,

He²,

Tömösközi³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…When it comes to a BSS problem, many candidate options are available. One school is machine learning (ML) based on neural networks (NNs), such as Y-Net [9], Conv-TasNet [10].…”

Section: Related Workmentioning

confidence: 99%

“…A natural idea is to first transfer all data to a centralized node; when all data are received, a sort of Blind Source Separation (BSS) [5] algorithm is applied to separate mixed data. BSS candidates include Independent Component Analysis (ICA)-based methods [6]- [8] or neural network-based methods [9], [10]. However, forwarding and then analyzing could delay critical decision-making actions due to i) possibly long waiting time of transferring the data, and ii) possibly long execution time of running the algorithm on a single node.…”

Section: Introductionmentioning

confidence: 99%

In-Network Processing Acoustic Data for Anomaly Detection in Smart Factory

Wu¹,

Shen²,

Xiao³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Modern manufacturing is now deeply integrating new technologies such as 5G, Internet-of-things (IoT), and cloud/edge computing to shape manufacturing to a new level -Smart Factory. Autonomic anomaly detection (e.g., malfunctioning machines and hazard situations) in a factory hall is on the list and expects to be realized with massive IoT sensor deployments. In this paper, we consider acoustic data-based anomaly detection, which is widely used in factories because sound information reflects richer internal states while videos cannot; besides, the capital investment of an audio system is more economically friendly. However, a unique challenge of using audio data is that sounds are mixed when collecting thus source data separation is inevitable. A traditional way transfers audio data all to a centralized point for separation. Nevertheless, such a centralized manner (i.e., data transferring and then analyzing) may delay prompt reactions to critical anomalies. We demonstrate that this job can be transformed into an in-network processing scheme and thus further accelerated. Specifically, we propose a progressive processing scheme where data separation jobs are distributed as microservices on intermediate nodes in parallel with data forwarding. Therefore, collected audio data can be separated 43.75% faster with even less total computing resources. This solution is comprehensively evaluated with numerical simulations, compared with benchmark solutions, and results justify its advantages.

show abstract

In-Network Processing for Low-Latency Industrial Anomaly Detection in Softwarized Networks

Tömösközi

et al. 2021

2021 IEEE Global Communications Conference (GLOBECOM)

View full text Add to dashboard Cite

Y-Net: A Dual Path Model for High Accuracy Blind Source Separation

Cited by 7 publications

References 9 publications

In-Network Processing for Low-Latency Industrial Anomaly Detection in Softwarized Networks

In-Network Processing for Low-Latency Industrial Anomaly Detection in Softwarized Networks

In-Network Processing Acoustic Data for Anomaly Detection in Smart Factory

In-Network Processing for Low-Latency Industrial Anomaly Detection in Softwarized Networks

Contact Info

Product

Resources

About