Pub Date : 2022-08-29DOI: 10.23919/eusipco55093.2022.9909836
Maria-Paula Comsa, R. Phlypo, P. Grangeat
This paper presents a dynamic transport model of a gaseous compound such as carbon dioxide based on the diffusion-convection through a three layered media composed of: a liquid medium (blood), a membrane (skin), a gaseous medium (air). The objective is to estimate the signal defined by the time variations in the concentration of the gaseous compound dissolved in the liquid medium based solely on the measurement signal defined by the time variations of the concentration of the gaseous compound in the gaseous medium. This dynamic model makes it possible to formulate the direct transport model in the form of a Markovian model with hidden states in order to generate synthetic data. We propose to implement a Kalman filter to calculate from the noisy observed variables, the hidden variables of the model, and in particular the concentration of the gaseous compound in the liquid medium. The challenge is to model the temporal evolution of a concentration profile as a function of time and depth taking into account the heterogeneity of the diffusion coefficients and the partition coefficients associated with the three media considered. The objective of this time recursive processing is to design an algorithm, which can be carried out on an embedded processor, taking into account the constraints of limited computing capacity. The application we are dealing with concerns the transcutaneous measurement of blood carbon dioxide in the forearm using an autonomous wristband-type worn device, in particular for monitoring respiratory diseases at home[1], [2].
{"title":"Inverting the diffusion-convection equation for gas desorption through an homogeneous membrane by Kalman filtering","authors":"Maria-Paula Comsa, R. Phlypo, P. Grangeat","doi":"10.23919/eusipco55093.2022.9909836","DOIUrl":"https://doi.org/10.23919/eusipco55093.2022.9909836","url":null,"abstract":"This paper presents a dynamic transport model of a gaseous compound such as carbon dioxide based on the diffusion-convection through a three layered media composed of: a liquid medium (blood), a membrane (skin), a gaseous medium (air). The objective is to estimate the signal defined by the time variations in the concentration of the gaseous compound dissolved in the liquid medium based solely on the measurement signal defined by the time variations of the concentration of the gaseous compound in the gaseous medium. This dynamic model makes it possible to formulate the direct transport model in the form of a Markovian model with hidden states in order to generate synthetic data. We propose to implement a Kalman filter to calculate from the noisy observed variables, the hidden variables of the model, and in particular the concentration of the gaseous compound in the liquid medium. The challenge is to model the temporal evolution of a concentration profile as a function of time and depth taking into account the heterogeneity of the diffusion coefficients and the partition coefficients associated with the three media considered. The objective of this time recursive processing is to design an algorithm, which can be carried out on an embedded processor, taking into account the constraints of limited computing capacity. The application we are dealing with concerns the transcutaneous measurement of blood carbon dioxide in the forearm using an autonomous wristband-type worn device, in particular for monitoring respiratory diseases at home[1], [2].","PeriodicalId":231263,"journal":{"name":"2022 30th European Signal Processing Conference (EUSIPCO)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122018129","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 : 2022-08-29DOI: 10.23919/eusipco55093.2022.9909587
François Gaits, Nicolas Mellado, Adrian Basarab
Ultrasound image simulation is a well-explored field with the main objective of generating realistic synthetic images, further used as ground truth (e.g. for training databases in machine learning), or for radiologists' training. Several ultrasound simulators are already available, most of them consisting in similar steps: (i) generate a collection of tissue mimicking individual scatterers with random spatial positions and random amplitudes, (ii) model the ultrasound probe and the emission and reception schemes, (iii) generate the RF signals resulting from the interaction between the scatterers and the propagating ultrasound waves. To ensure fully developed speckle, a few tens of scatterers by resolution cell are needed, demanding to handle high amounts of data (especially in 3D) and resulting into important computational time. The objective of this work is to explore new scatterer spatial distributions, with application to 2D slice simulation from 3D volumes. More precisely, lazy evaluation of pseudo-random schemes proves them to be highly computationally efficient compared to uniform random distribution commonly used. A statistical analysis confirms the visual impression of the results.
{"title":"Efficient 2D ultrasound simulation based on dart-throwing 3D scatterer sampling","authors":"François Gaits, Nicolas Mellado, Adrian Basarab","doi":"10.23919/eusipco55093.2022.9909587","DOIUrl":"https://doi.org/10.23919/eusipco55093.2022.9909587","url":null,"abstract":"Ultrasound image simulation is a well-explored field with the main objective of generating realistic synthetic images, further used as ground truth (e.g. for training databases in machine learning), or for radiologists' training. Several ultrasound simulators are already available, most of them consisting in similar steps: (i) generate a collection of tissue mimicking individual scatterers with random spatial positions and random amplitudes, (ii) model the ultrasound probe and the emission and reception schemes, (iii) generate the RF signals resulting from the interaction between the scatterers and the propagating ultrasound waves. To ensure fully developed speckle, a few tens of scatterers by resolution cell are needed, demanding to handle high amounts of data (especially in 3D) and resulting into important computational time. The objective of this work is to explore new scatterer spatial distributions, with application to 2D slice simulation from 3D volumes. More precisely, lazy evaluation of pseudo-random schemes proves them to be highly computationally efficient compared to uniform random distribution commonly used. A statistical analysis confirms the visual impression of the results.","PeriodicalId":231263,"journal":{"name":"2022 30th European Signal Processing Conference (EUSIPCO)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117234317","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 : 2022-08-29DOI: 10.23919/eusipco55093.2022.9909731
Q. Nguyen, M. Unoki
Bone-conducted (BC) speech potentially avoids the undesired effects on recorded speech due to background noise or reverberation; however, BC speech has lower quality and intelligibility than air-conducted (AC) speech. Since a large-scale BC speech database is hard to obtain (low-resource), current BC speech enhancement methods hardly improve the speech of speakers outside the training dataset. We proposed a method for enhancing BC speech from speakers outside of the training dataset in such a low-resource scenario. The proposed method contained a feature conversion model based on a vector-quantized variational autoencoder incorporating the gammachirp filterbank cepstral coefficients. The proposed method exploited the large-scale clean AC speech database to improve the quality of the BC speech. We conducted three evaluations to determine the effectiveness of the proposed method: perceptual evaluation of speech quality, short-time objective intelligibility, and the syllable error rate of the automatic speech recognition system. The results indicated that the proposed method could improve the sound quality and intelligibility of the BC speech from speakers outside of the training dataset.
{"title":"Bone-conducted Speech Enhancement Using Vector-quantized Variational Autoencoder and Gammachirp Filterbank Cepstral Coefficients","authors":"Q. Nguyen, M. Unoki","doi":"10.23919/eusipco55093.2022.9909731","DOIUrl":"https://doi.org/10.23919/eusipco55093.2022.9909731","url":null,"abstract":"Bone-conducted (BC) speech potentially avoids the undesired effects on recorded speech due to background noise or reverberation; however, BC speech has lower quality and intelligibility than air-conducted (AC) speech. Since a large-scale BC speech database is hard to obtain (low-resource), current BC speech enhancement methods hardly improve the speech of speakers outside the training dataset. We proposed a method for enhancing BC speech from speakers outside of the training dataset in such a low-resource scenario. The proposed method contained a feature conversion model based on a vector-quantized variational autoencoder incorporating the gammachirp filterbank cepstral coefficients. The proposed method exploited the large-scale clean AC speech database to improve the quality of the BC speech. We conducted three evaluations to determine the effectiveness of the proposed method: perceptual evaluation of speech quality, short-time objective intelligibility, and the syllable error rate of the automatic speech recognition system. The results indicated that the proposed method could improve the sound quality and intelligibility of the BC speech from speakers outside of the training dataset.","PeriodicalId":231263,"journal":{"name":"2022 30th European Signal Processing Conference (EUSIPCO)","volume":"65 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124708186","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 : 2022-08-29DOI: 10.23919/eusipco55093.2022.9909951
Othmane-Latif Ouabi, Jiawei Yi, Neil Zeghidour, N. Declercq, M. Geist, C. Pradalier
Mapping a structure, such as a metal plate in a ship hull, using acoustic echoes typically requires making assumptions on its shape (e.g. rectangular). This work introduces a more general approach based on beamforming to recover the geometry of arbitrary polygonal-shaped bounded areas from acoustic reflections. Our method only requires a single omni-directional emitter-receiver acoustic device mounted on a mobile platform. We apply beamforming to the acoustic measurements in the geometrical boundary space. We subsequently retrieve the edges from the beamforming results via the minimization of a regularized cost criterion, using a simulated annealing optimizer. We also design a boundary rejection criterion so that their exact number can be recovered based only on a specified upper bound. We assess our method on different geometries in a simulation environment and a real-world setting using ultrasonic guided waves measurements. The results demonstrate that it is efficient for achieving the targeted objectives.
{"title":"Polygonal Shapes Reconstruction from Acoustic Echoes Using a Mobile Sensor and Beamforming","authors":"Othmane-Latif Ouabi, Jiawei Yi, Neil Zeghidour, N. Declercq, M. Geist, C. Pradalier","doi":"10.23919/eusipco55093.2022.9909951","DOIUrl":"https://doi.org/10.23919/eusipco55093.2022.9909951","url":null,"abstract":"Mapping a structure, such as a metal plate in a ship hull, using acoustic echoes typically requires making assumptions on its shape (e.g. rectangular). This work introduces a more general approach based on beamforming to recover the geometry of arbitrary polygonal-shaped bounded areas from acoustic reflections. Our method only requires a single omni-directional emitter-receiver acoustic device mounted on a mobile platform. We apply beamforming to the acoustic measurements in the geometrical boundary space. We subsequently retrieve the edges from the beamforming results via the minimization of a regularized cost criterion, using a simulated annealing optimizer. We also design a boundary rejection criterion so that their exact number can be recovered based only on a specified upper bound. We assess our method on different geometries in a simulation environment and a real-world setting using ultrasonic guided waves measurements. The results demonstrate that it is efficient for achieving the targeted objectives.","PeriodicalId":231263,"journal":{"name":"2022 30th European Signal Processing Conference (EUSIPCO)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124745599","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 : 2022-08-29DOI: 10.23919/eusipco55093.2022.9909779
Abdulrahman Takiddin, Muhammad Ismail, E. Serpedin
Malicious customers hack into their smart meters to reduce their electricity bills using various cyberattack types. Such actions lead to financial losses and stability issues in the power grid. Existing research on machine learning-based detection offers promising detection performance. However, such detectors have been tested on a single type of cyberattacks and report performance accordingly, which is not a realistic setup since malicious customers may inject different types of cyberattacks. In this work, we examine the robustness of state-of-the-art machine learning-based electricity theft detectors against a combination of false data injection attacks (FDIAs). Specifically, we inject traditional, evasion, and data poisoning attacks with low, medium, and high injection levels then report the detection performance. Our results show that sequential ensemble learning-based detection offers the most stable detection performance that degrades only by 5.3% when subject to high injection levels of FDIAs compared to 15.7–18.5% degradation rates for the stand-alone detectors.
{"title":"Detection of Electricity Theft False Data Injection Attacks in Smart Grids","authors":"Abdulrahman Takiddin, Muhammad Ismail, E. Serpedin","doi":"10.23919/eusipco55093.2022.9909779","DOIUrl":"https://doi.org/10.23919/eusipco55093.2022.9909779","url":null,"abstract":"Malicious customers hack into their smart meters to reduce their electricity bills using various cyberattack types. Such actions lead to financial losses and stability issues in the power grid. Existing research on machine learning-based detection offers promising detection performance. However, such detectors have been tested on a single type of cyberattacks and report performance accordingly, which is not a realistic setup since malicious customers may inject different types of cyberattacks. In this work, we examine the robustness of state-of-the-art machine learning-based electricity theft detectors against a combination of false data injection attacks (FDIAs). Specifically, we inject traditional, evasion, and data poisoning attacks with low, medium, and high injection levels then report the detection performance. Our results show that sequential ensemble learning-based detection offers the most stable detection performance that degrades only by 5.3% when subject to high injection levels of FDIAs compared to 15.7–18.5% degradation rates for the stand-alone detectors.","PeriodicalId":231263,"journal":{"name":"2022 30th European Signal Processing Conference (EUSIPCO)","volume":"63 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125062768","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 : 2022-08-29DOI: 10.23919/eusipco55093.2022.9909603
Beatriz Sanabria-Barradas, S. Sanei, D. Granados-Ramos
Diagnosis of autism spectrum disorder (ASD) in children is often achieved by estimating the amplitudes and latencies of visual event-related potentials (ERPs). This requires accurate detection of desired ERPs, in our case P1 and N170, which are sensitive to visual stimuli. We aim to develop a hybrid of tensor factorization (TF) and singular spectrum analysis (SSA) to detect these components from electroencephalograms (EEGs) and restore the inherent noise and artifacts. The application of single-channel SSA to the detected sources by TF results in the removal of brain beta activity considerably enhancing the accuracy. The ERP parameters (amplitudes and latencies) are automatically estimated and applied to a decision-tree classifier leading to 100% accuracy.
{"title":"A Hybrid Tensor Factorization - Singular Spectrum Analysis Approach for ERP-based Assessment of Autism in Children","authors":"Beatriz Sanabria-Barradas, S. Sanei, D. Granados-Ramos","doi":"10.23919/eusipco55093.2022.9909603","DOIUrl":"https://doi.org/10.23919/eusipco55093.2022.9909603","url":null,"abstract":"Diagnosis of autism spectrum disorder (ASD) in children is often achieved by estimating the amplitudes and latencies of visual event-related potentials (ERPs). This requires accurate detection of desired ERPs, in our case P1 and N170, which are sensitive to visual stimuli. We aim to develop a hybrid of tensor factorization (TF) and singular spectrum analysis (SSA) to detect these components from electroencephalograms (EEGs) and restore the inherent noise and artifacts. The application of single-channel SSA to the detected sources by TF results in the removal of brain beta activity considerably enhancing the accuracy. The ERP parameters (amplitudes and latencies) are automatically estimated and applied to a decision-tree classifier leading to 100% accuracy.","PeriodicalId":231263,"journal":{"name":"2022 30th European Signal Processing Conference (EUSIPCO)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129452613","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 : 2022-08-29DOI: 10.23919/eusipco55093.2022.9909839
Yanbin He, M. Coutiño, E. Isufi, G. Leus
In this work, we focus on partitioning dynamic graphs with two types of nodes (bi-colored), though not necessarily bipartite graphs. They commonly appear in communication network applications, e.g., one color being base stations, the other users, and the dynamic process being the varying connection status between base stations and moving users. We introduce a partition cost function that incorporates the coloring of the graph and propose solutions based on the generalized eigenvalue problem (GEVP) for the static two-way partition problem. The static multi-way partition problem is then handled by a heuristic based on the two-way partition problem. Regarding the adaptive partition, an eigenvector update-based method is proposed. Numerical experiments demonstrate the performance of the devised approaches.
{"title":"Dynamic Bi-Colored Graph Partitioning","authors":"Yanbin He, M. Coutiño, E. Isufi, G. Leus","doi":"10.23919/eusipco55093.2022.9909839","DOIUrl":"https://doi.org/10.23919/eusipco55093.2022.9909839","url":null,"abstract":"In this work, we focus on partitioning dynamic graphs with two types of nodes (bi-colored), though not necessarily bipartite graphs. They commonly appear in communication network applications, e.g., one color being base stations, the other users, and the dynamic process being the varying connection status between base stations and moving users. We introduce a partition cost function that incorporates the coloring of the graph and propose solutions based on the generalized eigenvalue problem (GEVP) for the static two-way partition problem. The static multi-way partition problem is then handled by a heuristic based on the two-way partition problem. Regarding the adaptive partition, an eigenvector update-based method is proposed. Numerical experiments demonstrate the performance of the devised approaches.","PeriodicalId":231263,"journal":{"name":"2022 30th European Signal Processing Conference (EUSIPCO)","volume":"34 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129695845","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 : 2022-08-29DOI: 10.23919/eusipco55093.2022.9909642
Jyoti Maggu, Shalini Sharma, A. Majumdar
This work addresses the problem of solving a linear inverse problem. Conventional inversion techniques are model based (transductive). The advent of deep learning led the way for data-driven (inductive) inversion techniques. The main issue with inductive inversion is that unless the unseen signal (to be inverted) is similar to the training data, the learnt model fails to generalize rendering poor inversion results. A recent study on deep dictionary learning has shown how it can combine the best of both worlds – deep learning with transductive inversion. In this work, we show how the analysis counterpart of dictionary learning, called transform learning, can be extended deeper for transductive inversion. Results on dynamic MRI reconstruction, show that the proposed technique improves over the state-of-the-art.
{"title":"Transductive Inversion via Deep Transform Learning","authors":"Jyoti Maggu, Shalini Sharma, A. Majumdar","doi":"10.23919/eusipco55093.2022.9909642","DOIUrl":"https://doi.org/10.23919/eusipco55093.2022.9909642","url":null,"abstract":"This work addresses the problem of solving a linear inverse problem. Conventional inversion techniques are model based (transductive). The advent of deep learning led the way for data-driven (inductive) inversion techniques. The main issue with inductive inversion is that unless the unseen signal (to be inverted) is similar to the training data, the learnt model fails to generalize rendering poor inversion results. A recent study on deep dictionary learning has shown how it can combine the best of both worlds – deep learning with transductive inversion. In this work, we show how the analysis counterpart of dictionary learning, called transform learning, can be extended deeper for transductive inversion. Results on dynamic MRI reconstruction, show that the proposed technique improves over the state-of-the-art.","PeriodicalId":231263,"journal":{"name":"2022 30th European Signal Processing Conference (EUSIPCO)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128329815","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 : 2022-08-29DOI: 10.23919/eusipco55093.2022.9909697
M. Panagiotou, Athanasia Zlatintsi, P. Filntisis, A. J. Roumeliotis, Niki Efthymiou, P. Maragos
In this study, the application of deep learning models for the detection of relapses in patients with psychotic disorders (i.e., bipolar disorder and schizophrenia) is examined, using physiological signals, collected by smartwatches. In order to tackle the problem of relapse detection, which in our case is handled as an anomaly detection task, four different autoencoder architectures, based on Transformers, Fully connected Neural Networks (FNN), Convolution Neural Networks (CNN) and Gated Recurrent Unit (GRU), are implemented as personalized and global models. In this work, time-scaled data of total duration of 1569 days, segmented into five minutes intervals, from ten patients suffering from psychotic disorders have been examined yielding encouraging results. Furthermore, since the patients' relapses were appropriately annotated by clinicians as low, moderate or severe, we conducted a post hoc analysis using the models that performed best, to examine the importance of the severity level among three participants who relapsed multiple times with different severity level, providing important evidence.
{"title":"A comparative study of autoencoder architectures for mental health analysis using wearable sensors data","authors":"M. Panagiotou, Athanasia Zlatintsi, P. Filntisis, A. J. Roumeliotis, Niki Efthymiou, P. Maragos","doi":"10.23919/eusipco55093.2022.9909697","DOIUrl":"https://doi.org/10.23919/eusipco55093.2022.9909697","url":null,"abstract":"In this study, the application of deep learning models for the detection of relapses in patients with psychotic disorders (i.e., bipolar disorder and schizophrenia) is examined, using physiological signals, collected by smartwatches. In order to tackle the problem of relapse detection, which in our case is handled as an anomaly detection task, four different autoencoder architectures, based on Transformers, Fully connected Neural Networks (FNN), Convolution Neural Networks (CNN) and Gated Recurrent Unit (GRU), are implemented as personalized and global models. In this work, time-scaled data of total duration of 1569 days, segmented into five minutes intervals, from ten patients suffering from psychotic disorders have been examined yielding encouraging results. Furthermore, since the patients' relapses were appropriately annotated by clinicians as low, moderate or severe, we conducted a post hoc analysis using the models that performed best, to examine the importance of the severity level among three participants who relapsed multiple times with different severity level, providing important evidence.","PeriodicalId":231263,"journal":{"name":"2022 30th European Signal Processing Conference (EUSIPCO)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128599108","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 : 2022-08-29DOI: 10.23919/eusipco55093.2022.9909768
S. Sykiotis, Maria Kaselimi, A. Doulamis, N. Doulamis
In this study, we present TransformNILM, a novel Transformer based model for Non-Intrusive Load Monitoring (NILM). To infer the consumption signal of household appliances, TransformNILM employs Transformer layers, which utilize attention mechanisms to successfully draw global dependencies between input and output sequences. Trans-formNILM does not require data balancing and operates with minimal dataset pre-processing. Compared to other Transformer-based architectures, TransformNILM instigates an efficient training scheme, where model training consists of unsupervised pre-training and supervised model fine-tuning, thus leading to decreased training time and improved predictive performance. Experimental results validate Trans-formNILM's superiority compared to several state of the art methods.
{"title":"An Efficient Deep Bidirectional Transformer Model for Energy Disaggregation","authors":"S. Sykiotis, Maria Kaselimi, A. Doulamis, N. Doulamis","doi":"10.23919/eusipco55093.2022.9909768","DOIUrl":"https://doi.org/10.23919/eusipco55093.2022.9909768","url":null,"abstract":"In this study, we present TransformNILM, a novel Transformer based model for Non-Intrusive Load Monitoring (NILM). To infer the consumption signal of household appliances, TransformNILM employs Transformer layers, which utilize attention mechanisms to successfully draw global dependencies between input and output sequences. Trans-formNILM does not require data balancing and operates with minimal dataset pre-processing. Compared to other Transformer-based architectures, TransformNILM instigates an efficient training scheme, where model training consists of unsupervised pre-training and supervised model fine-tuning, thus leading to decreased training time and improved predictive performance. Experimental results validate Trans-formNILM's superiority compared to several state of the art methods.","PeriodicalId":231263,"journal":{"name":"2022 30th European Signal Processing Conference (EUSIPCO)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129352011","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：