Pub Date : 2021-06-14DOI: 10.1109/WF-IoT51360.2021.9595077
M. Bundas, Chasity Nadeau, T. Nguyen, Jeannine Shantz, M. Balduccini, Tran Cao Son
While Artificial Intelligence (AI) and Machine Learning provide a pathway of new and exciting possibilities for AI-Enabled Cyber-Physical and Internet of Things systems, these technology solutions are not without challenges that may hinder adoption. We do not always understand why AI components behave in the way they do, nor can we always predict what they will do under new circumstances. In this paper, we discuss possible approaches for extending the NIST CPS Framework in a way that provides designers, operators and other stakeholders with a shared vocabulary and a collaborative framework allowing them to discuss, identify, express, and verify requirements on the behavior of AI-enabled Cyber-Physical and Internet of Things Systems.
{"title":"Towards a Framework for Characterizing the Behavior of AI-Enabled Cyber-Physical and IoT Systems","authors":"M. Bundas, Chasity Nadeau, T. Nguyen, Jeannine Shantz, M. Balduccini, Tran Cao Son","doi":"10.1109/WF-IoT51360.2021.9595077","DOIUrl":"https://doi.org/10.1109/WF-IoT51360.2021.9595077","url":null,"abstract":"While Artificial Intelligence (AI) and Machine Learning provide a pathway of new and exciting possibilities for AI-Enabled Cyber-Physical and Internet of Things systems, these technology solutions are not without challenges that may hinder adoption. We do not always understand why AI components behave in the way they do, nor can we always predict what they will do under new circumstances. In this paper, we discuss possible approaches for extending the NIST CPS Framework in a way that provides designers, operators and other stakeholders with a shared vocabulary and a collaborative framework allowing them to discuss, identify, express, and verify requirements on the behavior of AI-enabled Cyber-Physical and Internet of Things Systems.","PeriodicalId":184138,"journal":{"name":"2021 IEEE 7th World Forum on Internet of Things (WF-IoT)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128050600","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-06-14DOI: 10.1109/WF-IoT51360.2021.9595769
Meenakshi Sethunath, Yang Peng
Serverless computing functions typically execute in the cloud. However, the high latency of accessing the cloud may require running them on edge servers, which have limited computing power and memory availability. This paper proposes a joint resource allocation and request dispatch scheme to execute serverless computing functions over edge and cloud collaboratively. This new scheme explicitly considers how to allocate server memory and operation budget for concurrent serverless computing requests considering the cold-start latency in design. The proposed scheme has been evaluated through extensive simulations. Its effectiveness has been proved by comparison with the upper-bound results.
{"title":"A Joint Resource Allocation and Request Dispatch Scheme for Performing Serverless Computing over Edge and Cloud","authors":"Meenakshi Sethunath, Yang Peng","doi":"10.1109/WF-IoT51360.2021.9595769","DOIUrl":"https://doi.org/10.1109/WF-IoT51360.2021.9595769","url":null,"abstract":"Serverless computing functions typically execute in the cloud. However, the high latency of accessing the cloud may require running them on edge servers, which have limited computing power and memory availability. This paper proposes a joint resource allocation and request dispatch scheme to execute serverless computing functions over edge and cloud collaboratively. This new scheme explicitly considers how to allocate server memory and operation budget for concurrent serverless computing requests considering the cold-start latency in design. The proposed scheme has been evaluated through extensive simulations. Its effectiveness has been proved by comparison with the upper-bound results.","PeriodicalId":184138,"journal":{"name":"2021 IEEE 7th World Forum on Internet of Things (WF-IoT)","volume":"40 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128293838","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-06-14DOI: 10.1109/WF-IoT51360.2021.9595496
Ali Shoker, Peter Moertl, Ramiro Robles
Truck platooning is a form of convoy cooperative driving of connected trucks assisted by a lead truck. The aim is to reduce the fuel and driving costs, improve road safety, and reduce CO2 emission. Being semi-autonomous, platoons must be trustworthy in many perspectives. This paper presents a high-level trustworthy requirements analysis on three key perspectives: driver, communication, and security. In addition, we observed that any trustworthy requirement analysis is incomplete if perspectives are addressed independently. Therefore, we propose a simple holistic methodology that addresses the different perspectives as well as their dependencies, and we exemplify the use of the methodology with two use cases presented in the paper. However, we draw attention to the importance of more research to drive a more exhaustive and validated methodology1.
{"title":"A First Step Towards Holistic Trustworthy Platoons","authors":"Ali Shoker, Peter Moertl, Ramiro Robles","doi":"10.1109/WF-IoT51360.2021.9595496","DOIUrl":"https://doi.org/10.1109/WF-IoT51360.2021.9595496","url":null,"abstract":"Truck platooning is a form of convoy cooperative driving of connected trucks assisted by a lead truck. The aim is to reduce the fuel and driving costs, improve road safety, and reduce CO2 emission. Being semi-autonomous, platoons must be trustworthy in many perspectives. This paper presents a high-level trustworthy requirements analysis on three key perspectives: driver, communication, and security. In addition, we observed that any trustworthy requirement analysis is incomplete if perspectives are addressed independently. Therefore, we propose a simple holistic methodology that addresses the different perspectives as well as their dependencies, and we exemplify the use of the methodology with two use cases presented in the paper. However, we draw attention to the importance of more research to drive a more exhaustive and validated methodology1.","PeriodicalId":184138,"journal":{"name":"2021 IEEE 7th World Forum on Internet of Things (WF-IoT)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124626280","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-06-14DOI: 10.1109/WF-IoT51360.2021.9594925
Jacob Hunte, H. Lutfiyya, A. Haque
This paper looks at providing an efficient method of synchronizing devices deployed in an electrical grid. The proposed method focuses on device synchronization specifically for localizing faults on distribution networks. It analyses the travelling waves that are present on the electrical grid at and around the time of the fault. It is a synchronization method which uses external signals to synchronize the fault events detected by the devices without reliance on accuracy of clocks used in each device. Initial experimental results shows that this is a promising approach.
{"title":"Device Synchronization for Fault Localization in Electrical Distribution Grids","authors":"Jacob Hunte, H. Lutfiyya, A. Haque","doi":"10.1109/WF-IoT51360.2021.9594925","DOIUrl":"https://doi.org/10.1109/WF-IoT51360.2021.9594925","url":null,"abstract":"This paper looks at providing an efficient method of synchronizing devices deployed in an electrical grid. The proposed method focuses on device synchronization specifically for localizing faults on distribution networks. It analyses the travelling waves that are present on the electrical grid at and around the time of the fault. It is a synchronization method which uses external signals to synchronize the fault events detected by the devices without reliance on accuracy of clocks used in each device. Initial experimental results shows that this is a promising approach.","PeriodicalId":184138,"journal":{"name":"2021 IEEE 7th World Forum on Internet of Things (WF-IoT)","volume":"190 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121842264","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-06-14DOI: 10.1109/WF-IoT51360.2021.9596006
R. Kureshi, D. Thakker, B. Mishra, Baseer Ahmad
On average, we spend around 90% of the time in indoor environments. Indoor Air Quality (IAQ) has been receiving increased attention from the environmental bodies, local authorities and citizens as it is becoming clearer that poor IAQ has public health implications. Therefore, monitoring of indoor environment and involving citizens becomes crucial to enhance IAQ and managing their indoor environments by raising awareness – a goal of many Citizen Science (CS) projects. In this work, we present a use case of IAQ monitoring in a European project with a focus on Smart Cities with citizen engagement and involvement. It is well known that the cost of Air Quality (AQ) monitoring stations, which are often stationary, and generally produce reliable, and high-quality data is a non-starter for CS projects as cost prohibits the scaling of deployment and citizen involvement. On the other hand, it is widely assumed that low-cost devices for AQ, although available in abundance, often produce low-quality data, putting the credibility of basing any analysis on low-cost sensors. There is an increasing number of research efforts that look at how to ascertain the data quality of such sensors so that they could still be used reliably, often to provide indicative readings, and for analytics. In this work, we present data science-based techniques that have been utilised for selecting low-cost sensors based on their data quality indicators, and an integrated visualisation system that utilises structure data for IAQ to support multi-city trials in a CS project. The sensors are selected after analysing their consistency over a period by applying different approaches such as statistical analysis and graphical plots.
{"title":"Use Case of Building an Indoor Air Quality Monitoring System","authors":"R. Kureshi, D. Thakker, B. Mishra, Baseer Ahmad","doi":"10.1109/WF-IoT51360.2021.9596006","DOIUrl":"https://doi.org/10.1109/WF-IoT51360.2021.9596006","url":null,"abstract":"On average, we spend around 90% of the time in indoor environments. Indoor Air Quality (IAQ) has been receiving increased attention from the environmental bodies, local authorities and citizens as it is becoming clearer that poor IAQ has public health implications. Therefore, monitoring of indoor environment and involving citizens becomes crucial to enhance IAQ and managing their indoor environments by raising awareness – a goal of many Citizen Science (CS) projects. In this work, we present a use case of IAQ monitoring in a European project with a focus on Smart Cities with citizen engagement and involvement. It is well known that the cost of Air Quality (AQ) monitoring stations, which are often stationary, and generally produce reliable, and high-quality data is a non-starter for CS projects as cost prohibits the scaling of deployment and citizen involvement. On the other hand, it is widely assumed that low-cost devices for AQ, although available in abundance, often produce low-quality data, putting the credibility of basing any analysis on low-cost sensors. There is an increasing number of research efforts that look at how to ascertain the data quality of such sensors so that they could still be used reliably, often to provide indicative readings, and for analytics. In this work, we present data science-based techniques that have been utilised for selecting low-cost sensors based on their data quality indicators, and an integrated visualisation system that utilises structure data for IAQ to support multi-city trials in a CS project. The sensors are selected after analysing their consistency over a period by applying different approaches such as statistical analysis and graphical plots.","PeriodicalId":184138,"journal":{"name":"2021 IEEE 7th World Forum on Internet of Things (WF-IoT)","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124424517","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-06-14DOI: 10.1109/WF-IoT51360.2021.9595143
Youssef Osman, Reed Dennis, Khalid Elgazzar
We perform fruit counting on video footage by following a two-stage pipeline that consists of detecting the fruits, then tracking them frame-by-frame. Detection is done through the use of You Only Look Once model (YOLO). Bounding boxes are extracted from detection and Non Max Suppression (NMS) is performed to get final detections. The boxes are then input into the tracking pipeline. For tracking, we apply a custom-developed DeepSORT algorithm to work with fruits. Using the box coordinates, every detected object is cropped out of the original image, and a separate feature extraction using a convolutional neural network (CNN) called ResNet is performed on that image crop to get the feature map. New detections are associated with old detections by comparing their features as a distance metric, where two objects with minimal distance are associated together. Input objects with no association are treated as new objects to be tracked. By keeping track of the fruits throughout the video frames, we ensure that we’re counting them appropriately when they are first detected. We demonstrate the approach on videos from an apple orchard to test the performance of the proposed pipeline in natural light. Experimental results show high accuracy of fruit counting on real-time video feeds. The new approach can be efficiently applied on any type of fruit and vegetables with no changes in the algorithms.
{"title":"Yield Estimation using Deep Learning for Precision Agriculture","authors":"Youssef Osman, Reed Dennis, Khalid Elgazzar","doi":"10.1109/WF-IoT51360.2021.9595143","DOIUrl":"https://doi.org/10.1109/WF-IoT51360.2021.9595143","url":null,"abstract":"We perform fruit counting on video footage by following a two-stage pipeline that consists of detecting the fruits, then tracking them frame-by-frame. Detection is done through the use of You Only Look Once model (YOLO). Bounding boxes are extracted from detection and Non Max Suppression (NMS) is performed to get final detections. The boxes are then input into the tracking pipeline. For tracking, we apply a custom-developed DeepSORT algorithm to work with fruits. Using the box coordinates, every detected object is cropped out of the original image, and a separate feature extraction using a convolutional neural network (CNN) called ResNet is performed on that image crop to get the feature map. New detections are associated with old detections by comparing their features as a distance metric, where two objects with minimal distance are associated together. Input objects with no association are treated as new objects to be tracked. By keeping track of the fruits throughout the video frames, we ensure that we’re counting them appropriately when they are first detected. We demonstrate the approach on videos from an apple orchard to test the performance of the proposed pipeline in natural light. Experimental results show high accuracy of fruit counting on real-time video feeds. The new approach can be efficiently applied on any type of fruit and vegetables with no changes in the algorithms.","PeriodicalId":184138,"journal":{"name":"2021 IEEE 7th World Forum on Internet of Things (WF-IoT)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132363413","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Our research work describes how deep learning techniques can be applied with success to create a non-invasive method to control the dynamic density of dough rising during the fermentation process. This paper explains in detail the steps performed to train and apply on the field a Convolutional Neural Network (CNN) to monitor the leavening of a traditional Christmas Italian type of sweet bread made in Milan, called Panettone, that usually needs an accurate and supervised leavening process of around three days. One of our main goals was to prove how these CNNs and their learned inner representations could easily become the foundation for developing a remote-supervision framework capable to monitor the leavening process. Since the duration of this crucial phase is not exactly predictable, as it depends on many external factors, and usually takes place during the night, it makes sense to adopt a not supervised approach that is able to autonomously detect and notify to the bakery personnel (sending sms, email, whatsapp, etc.) when the density of dough is considered optimal, appropriate and ready to start the baking phases.Results demonstrated that a CNN based paradigm is more effective and more accurate than the current used empirical methods. The model converged and the average loss value was near to zero, even if the set of images and examples adopted to train and test the classifier was limited. Though applied in the bakery products context, the designed approach can be easily adapted to other monitoring tasks or industry domains, and its independence from User expert knowledge and specific artisanal skills can be considered one of the major advantages.
{"title":"Leavening control system based on machine learning techniques","authors":"Desilda Toska, Alfredo Pulla, Stefano Robustelli, Gianmarco Fiamma","doi":"10.1109/WF-IoT51360.2021.9595709","DOIUrl":"https://doi.org/10.1109/WF-IoT51360.2021.9595709","url":null,"abstract":"Our research work describes how deep learning techniques can be applied with success to create a non-invasive method to control the dynamic density of dough rising during the fermentation process. This paper explains in detail the steps performed to train and apply on the field a Convolutional Neural Network (CNN) to monitor the leavening of a traditional Christmas Italian type of sweet bread made in Milan, called Panettone, that usually needs an accurate and supervised leavening process of around three days. One of our main goals was to prove how these CNNs and their learned inner representations could easily become the foundation for developing a remote-supervision framework capable to monitor the leavening process. Since the duration of this crucial phase is not exactly predictable, as it depends on many external factors, and usually takes place during the night, it makes sense to adopt a not supervised approach that is able to autonomously detect and notify to the bakery personnel (sending sms, email, whatsapp, etc.) when the density of dough is considered optimal, appropriate and ready to start the baking phases.Results demonstrated that a CNN based paradigm is more effective and more accurate than the current used empirical methods. The model converged and the average loss value was near to zero, even if the set of images and examples adopted to train and test the classifier was limited. Though applied in the bakery products context, the designed approach can be easily adapted to other monitoring tasks or industry domains, and its independence from User expert knowledge and specific artisanal skills can be considered one of the major advantages.","PeriodicalId":184138,"journal":{"name":"2021 IEEE 7th World Forum on Internet of Things (WF-IoT)","volume":"8 4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132220210","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-06-14DOI: 10.1109/WF-IoT51360.2021.9595503
Y. Ohba, Jing Yi Koh, N. Ng, S. Keoh
This paper studies the performance of a proximity content distribution scheme over IEEE 802.11s mesh networks with reasonable user density among combinations of three network configurations and two transport mechanisms. For content access control, Hyperledger Sawtooth Blockchain with PoET (Proof of Elapsed Time) consensus algorithm is used as a decentralised storage of non-repudiated and rapid transactions for granting content access and distributing the content decryption key. An extensive performance evaluation of the content distribution and content access control protocols using ns-3 simulator was conducted. The results show that the integration of Blockchain and UDP multicast content distribution in a hybrid mesh network topology is highly feasible.
{"title":"Performance Evaluation of a Blockchain-based Content Distribution over Wireless Mesh Networks","authors":"Y. Ohba, Jing Yi Koh, N. Ng, S. Keoh","doi":"10.1109/WF-IoT51360.2021.9595503","DOIUrl":"https://doi.org/10.1109/WF-IoT51360.2021.9595503","url":null,"abstract":"This paper studies the performance of a proximity content distribution scheme over IEEE 802.11s mesh networks with reasonable user density among combinations of three network configurations and two transport mechanisms. For content access control, Hyperledger Sawtooth Blockchain with PoET (Proof of Elapsed Time) consensus algorithm is used as a decentralised storage of non-repudiated and rapid transactions for granting content access and distributing the content decryption key. An extensive performance evaluation of the content distribution and content access control protocols using ns-3 simulator was conducted. The results show that the integration of Blockchain and UDP multicast content distribution in a hybrid mesh network topology is highly feasible.","PeriodicalId":184138,"journal":{"name":"2021 IEEE 7th World Forum on Internet of Things (WF-IoT)","volume":"84 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134222868","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-06-14DOI: 10.1109/WF-IoT51360.2021.9595827
H. Herath, N. M. P. M. Nishshanka, P. V. N. U. Madhumali, Subhodha Gunawardena
Motor dysfunction is a common outcome of strokes, spinal cord injuries, head injuries and multiple sclerosis. Their occupational therapies bring a lot of difficulties as they are labor-intensive, time-consuming and expensive. Robots play a major role in rehabilitation by replacing traditional therapies and offer ideal customized therapies. Further, wearable robots such as exoskeletons make the rehabilitation process simpler. Most of the existing rehabilitation robots use joysticks as their control method, which requires hand movement form the patient or a helper. However, introducing voice control mechanisms to these rehabilitation robots would raise the independence of individuals in robot controlling. This paper introduces a model to control robotic devices using voice commands which is based on Recurrent Neural Networks (RNN). Here, Long Short-Term Memory (LSTM) machine learning technique is implemented on “RehaBot” exoskeleton robot which is used for upper-limb rehabilitation with two-Degree of Freedom (2-DOF). Ten different voice commands are used to design the voice control system contemplating the movements of the upper limb. As the voice commands could be affected by the background noise, gender and data input source (microphone), their effects on voice commands are analyzed and discussed here.
{"title":"Voice Control System for Upper Limb Rehabilitation Robots using Machine Learning","authors":"H. Herath, N. M. P. M. Nishshanka, P. V. N. U. Madhumali, Subhodha Gunawardena","doi":"10.1109/WF-IoT51360.2021.9595827","DOIUrl":"https://doi.org/10.1109/WF-IoT51360.2021.9595827","url":null,"abstract":"Motor dysfunction is a common outcome of strokes, spinal cord injuries, head injuries and multiple sclerosis. Their occupational therapies bring a lot of difficulties as they are labor-intensive, time-consuming and expensive. Robots play a major role in rehabilitation by replacing traditional therapies and offer ideal customized therapies. Further, wearable robots such as exoskeletons make the rehabilitation process simpler. Most of the existing rehabilitation robots use joysticks as their control method, which requires hand movement form the patient or a helper. However, introducing voice control mechanisms to these rehabilitation robots would raise the independence of individuals in robot controlling. This paper introduces a model to control robotic devices using voice commands which is based on Recurrent Neural Networks (RNN). Here, Long Short-Term Memory (LSTM) machine learning technique is implemented on “RehaBot” exoskeleton robot which is used for upper-limb rehabilitation with two-Degree of Freedom (2-DOF). Ten different voice commands are used to design the voice control system contemplating the movements of the upper limb. As the voice commands could be affected by the background noise, gender and data input source (microphone), their effects on voice commands are analyzed and discussed here.","PeriodicalId":184138,"journal":{"name":"2021 IEEE 7th World Forum on Internet of Things (WF-IoT)","volume":"124 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134238558","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-06-14DOI: 10.1109/WF-IoT51360.2021.9596007
Steven Wyatt, David Elliott, A. Aravamudan, C. Otero, L. D. Otero, G. Anagnostopoulos, Anthony O. Smith, A. Peter, Wesley Jones, Steven Leung, Eric Lam
The unprecedented growth of edge sensor infrastructure is driving the demand function for in situ analytics, i.e. automated decision support at the point of data collection. In the present work, we detail our state-of-the-art Environmental Sound Classification (ESC) framework that is capable of near real-time acoustic categorization directly at the edge. Existing ESC algorithms primarily train and test on pristine datasets that fail in real-world deployments due their inability to handle real-world noisy environments. Methods to denoise the sounds are often computationally expensive for edge devices and do not guarantee performance improvements. To this end, we investigate a way to make existing ESC models robust and make them work in operational resource-constrained settings. Our framework employs a noisy classification model consisting of a tiny BERT-based Transformer (less than 20,000 parameters) and considers hardening of this model through the use of transmission channel noise augmentation. We detail real-world results through its deployment on a Raspberry Pi Zero and demonstrate its classification performance.
边缘传感器基础设施的空前增长推动了现场分析的需求功能，即在数据收集点的自动化决策支持。在目前的工作中，我们详细介绍了我们最先进的环境声音分类(ESC)框架，该框架能够直接在边缘进行近乎实时的声学分类。现有的ESC算法主要在原始数据集上进行训练和测试，这些数据集由于无法处理真实的噪声环境而在实际部署中失败。对于边缘设备来说，去噪声音的方法通常在计算上是昂贵的，并且不能保证性能的提高。为此，我们研究了一种方法，使现有的ESC模型具有鲁棒性，并使其在操作资源受限的环境下工作。我们的框架采用了一个由基于bert的小型变压器(小于20,000个参数)组成的噪声分类模型，并考虑通过使用传输通道噪声增强来强化该模型。我们通过在Raspberry Pi Zero上的部署详细介绍了实际结果，并演示了其分类性能。
{"title":"Environmental Sound Classification with Tiny Transformers in Noisy Edge Environments","authors":"Steven Wyatt, David Elliott, A. Aravamudan, C. Otero, L. D. Otero, G. Anagnostopoulos, Anthony O. Smith, A. Peter, Wesley Jones, Steven Leung, Eric Lam","doi":"10.1109/WF-IoT51360.2021.9596007","DOIUrl":"https://doi.org/10.1109/WF-IoT51360.2021.9596007","url":null,"abstract":"The unprecedented growth of edge sensor infrastructure is driving the demand function for in situ analytics, i.e. automated decision support at the point of data collection. In the present work, we detail our state-of-the-art Environmental Sound Classification (ESC) framework that is capable of near real-time acoustic categorization directly at the edge. Existing ESC algorithms primarily train and test on pristine datasets that fail in real-world deployments due their inability to handle real-world noisy environments. Methods to denoise the sounds are often computationally expensive for edge devices and do not guarantee performance improvements. To this end, we investigate a way to make existing ESC models robust and make them work in operational resource-constrained settings. Our framework employs a noisy classification model consisting of a tiny BERT-based Transformer (less than 20,000 parameters) and considers hardening of this model through the use of transmission channel noise augmentation. We detail real-world results through its deployment on a Raspberry Pi Zero and demonstrate its classification performance.","PeriodicalId":184138,"journal":{"name":"2021 IEEE 7th World Forum on Internet of Things (WF-IoT)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132963793","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：