Pub Date : 2021-08-23DOI: 10.1109/SAS51076.2021.9530149
F. Zonzini, Matteo Zauli, Mauro Mangia, N. Testoni, L. De Marchi
Nowadays, there is an increasing demand for resilient and long-term monitoring solutions, capable to enhance the safety of aging structures against man-made and built-in hazards. Nonetheless, the widespread deployment of full-scale and dense sensor networks might be incompatible with the available energy budget. Besides, the massive amount of data which is acquired might cause network congestion. To address these issues, the Compressed Sensing (CS) technique represents a solution that is cost-effective and specifically suited for the vibration diagnostics field. This work investigates the feasibility of a model-based CS technique, exploiting the so-called rakeness (Rak-CS) approach, which is robust against noise uncertainty in the context of pure ambient vibrations. Experimental results proved that the accuracy of the reconstructed structural parameters is up to 95 % (i.e. modal shape correlation equal to 0.95) with a compression ratio equal to 10.
{"title":"HW-Oriented Compressed Sensing for Operational Modal Analysis: The Impact of Noise in MEMS Accelerometer Networks","authors":"F. Zonzini, Matteo Zauli, Mauro Mangia, N. Testoni, L. De Marchi","doi":"10.1109/SAS51076.2021.9530149","DOIUrl":"https://doi.org/10.1109/SAS51076.2021.9530149","url":null,"abstract":"Nowadays, there is an increasing demand for resilient and long-term monitoring solutions, capable to enhance the safety of aging structures against man-made and built-in hazards. Nonetheless, the widespread deployment of full-scale and dense sensor networks might be incompatible with the available energy budget. Besides, the massive amount of data which is acquired might cause network congestion. To address these issues, the Compressed Sensing (CS) technique represents a solution that is cost-effective and specifically suited for the vibration diagnostics field. This work investigates the feasibility of a model-based CS technique, exploiting the so-called rakeness (Rak-CS) approach, which is robust against noise uncertainty in the context of pure ambient vibrations. Experimental results proved that the accuracy of the reconstructed structural parameters is up to 95 % (i.e. modal shape correlation equal to 0.95) with a compression ratio equal to 10.","PeriodicalId":224327,"journal":{"name":"2021 IEEE Sensors Applications Symposium (SAS)","volume":"66 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129636912","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 : 2021-08-23DOI: 10.1109/SAS51076.2021.9530059
B. Andò, S. Baglio, S. Castorina, S. Graziani, M. Messina, S. Petralia, S. V. Gupta Tondepu
In this article, the information reported is of a genetical metabolic disorder named Phenylketonuria PKU and the necessity for the development of a biosensor with a complete Point-of-Care system. The choice of the biosensor is a low-cost rapid prototyping sensor fabricated by Inkjet Printed (IJP) Interdigitated Capacitive Transducer (IDT) on a PET substrate. The work curtailed a detailed explanation on the chosen methodology indirect phenylalanine quantification by making use the dielectric permittivity variation caused by the chemical reaction between the functional layer with the residual Ammonia (NH3) generated by an enzymatic reaction involving phenylalanine. This work includes a clear elaboration of the sensor design flow and the modelling of the capacitive sensor with multiple layers on top of the electrodes (Functional layer and Material Under Test). Simulation results are shown which allow for the choice of optimal design parameters for the Capacitive sensor.
{"title":"A Capacitive Readout Strategy for Ammonia Detection: Design Flow, Modeling and Simulation","authors":"B. Andò, S. Baglio, S. Castorina, S. Graziani, M. Messina, S. Petralia, S. V. Gupta Tondepu","doi":"10.1109/SAS51076.2021.9530059","DOIUrl":"https://doi.org/10.1109/SAS51076.2021.9530059","url":null,"abstract":"In this article, the information reported is of a genetical metabolic disorder named Phenylketonuria PKU and the necessity for the development of a biosensor with a complete Point-of-Care system. The choice of the biosensor is a low-cost rapid prototyping sensor fabricated by Inkjet Printed (IJP) Interdigitated Capacitive Transducer (IDT) on a PET substrate. The work curtailed a detailed explanation on the chosen methodology indirect phenylalanine quantification by making use the dielectric permittivity variation caused by the chemical reaction between the functional layer with the residual Ammonia (NH3) generated by an enzymatic reaction involving phenylalanine. This work includes a clear elaboration of the sensor design flow and the modelling of the capacitive sensor with multiple layers on top of the electrodes (Functional layer and Material Under Test). Simulation results are shown which allow for the choice of optimal design parameters for the Capacitive sensor.","PeriodicalId":224327,"journal":{"name":"2021 IEEE Sensors Applications Symposium (SAS)","volume":"58 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122038866","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 : 2021-08-23DOI: 10.1109/SAS51076.2021.9530034
Burcu Arman Kuzubasoglu, S. Bahadir
In this study, the behavior of the sensor printed on the textile surface with carbon nanotube (CNT)-based ink formulated for wearable sensor applications against temperature and humidity was modeled using artificial neural networks. While humidity and temperature are defined as network input variables, the linear electrical resistance value is defined as network output variable. In the study, 167 experimental results were entered as data set, 70% of them were used for ANN training, 15% for validation of the proposed model, and 15% for testing. Levenberg Marquardt (LM) and Bayesian Regularization (BR) were used as the learning algorithm. The logarithmic sigmoid has been used in hidden layers and fitnet in output neurons have been used as an activation function. It has been observed that the developed artificial neural network model exhibits a significant performance in estimating the electrical resistance value against temperature for textile-based sensors developed in different humidity conditions from 50 % relative humidity to 80 % relative humidity and a good agreement with experimental data.
{"title":"Modeling of Wearable Sensor in Various Temperature and Humidity Conditions by Artificial Neural Networks","authors":"Burcu Arman Kuzubasoglu, S. Bahadir","doi":"10.1109/SAS51076.2021.9530034","DOIUrl":"https://doi.org/10.1109/SAS51076.2021.9530034","url":null,"abstract":"In this study, the behavior of the sensor printed on the textile surface with carbon nanotube (CNT)-based ink formulated for wearable sensor applications against temperature and humidity was modeled using artificial neural networks. While humidity and temperature are defined as network input variables, the linear electrical resistance value is defined as network output variable. In the study, 167 experimental results were entered as data set, 70% of them were used for ANN training, 15% for validation of the proposed model, and 15% for testing. Levenberg Marquardt (LM) and Bayesian Regularization (BR) were used as the learning algorithm. The logarithmic sigmoid has been used in hidden layers and fitnet in output neurons have been used as an activation function. It has been observed that the developed artificial neural network model exhibits a significant performance in estimating the electrical resistance value against temperature for textile-based sensors developed in different humidity conditions from 50 % relative humidity to 80 % relative humidity and a good agreement with experimental data.","PeriodicalId":224327,"journal":{"name":"2021 IEEE Sensors Applications Symposium (SAS)","volume":"107 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115951264","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 : 2021-08-23DOI: 10.1109/SAS51076.2021.9530177
Jobish John, Vinay S. Palaparthy, Apoorv Dethe, M. Baghini
Dielectric based capacitive soil moisture sensors are widely used because of their affordability and ease of use. We propose a simple temperature compensated soil specific in-field calibration method for frequency-domain soil moisture sensors and is implemented using the in-house developed soil moisture sensors. The proposed approach produces two different look-up table based calibration models, one corresponding to 22° C and another corresponding to 32° C. The sensor output frequency is mapped to the soil moisture with the help of linear interpolation using both the models whenever the soil temperature is in the range of 22° C - 32° C. If the soil temperature is outside this range, the calibration model closer to the temperature is used for soil moisture measurements. With the proposed calibration approach, the maximum difference between the gravimetric soil moisture and the measured values is observed as 3 % in comparison with the conventional oven-dry laboratory calibration approach, a labour-intensive method. Field experiments were carried out for five consecutive days using a wireless sensor network consisting of 3 sensor nodes where each node reported its sensor data every 3 hours. The field measurements with the proposed calibration approach showed a maximum deviation of 3.17% in comparison with gravimetric measurements.
{"title":"A temperature compensated soil specific calibration approach for frequency domain soil moisture sensors for in-situ agricultural applications","authors":"Jobish John, Vinay S. Palaparthy, Apoorv Dethe, M. Baghini","doi":"10.1109/SAS51076.2021.9530177","DOIUrl":"https://doi.org/10.1109/SAS51076.2021.9530177","url":null,"abstract":"Dielectric based capacitive soil moisture sensors are widely used because of their affordability and ease of use. We propose a simple temperature compensated soil specific in-field calibration method for frequency-domain soil moisture sensors and is implemented using the in-house developed soil moisture sensors. The proposed approach produces two different look-up table based calibration models, one corresponding to 22° C and another corresponding to 32° C. The sensor output frequency is mapped to the soil moisture with the help of linear interpolation using both the models whenever the soil temperature is in the range of 22° C - 32° C. If the soil temperature is outside this range, the calibration model closer to the temperature is used for soil moisture measurements. With the proposed calibration approach, the maximum difference between the gravimetric soil moisture and the measured values is observed as 3 % in comparison with the conventional oven-dry laboratory calibration approach, a labour-intensive method. Field experiments were carried out for five consecutive days using a wireless sensor network consisting of 3 sensor nodes where each node reported its sensor data every 3 hours. The field measurements with the proposed calibration approach showed a maximum deviation of 3.17% in comparison with gravimetric measurements.","PeriodicalId":224327,"journal":{"name":"2021 IEEE Sensors Applications Symposium (SAS)","volume":"73 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130564837","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 : 2021-08-23DOI: 10.1109/SAS51076.2021.9530111
H. Noshahri, Y. Wijnant, Catalin Cernat, E. Dertien, L. O. Scholtenhuis
Detecting voids in pipe surroundings is essential to structural condition assessment of concrete sewer pipelines. Impact-echo is a non-destructive testing method that can be used for this purpose. This method works based on exciting the surface of concrete and using a contact-based sensor to monitor the propagation of the resulting stress waves. However, the presence of deposits and humidity inside the sewer pipe makes establishing a contact between the sensor and the pipe wall very difficult. Therefore, the goal of this study is to compare the performance of contactless sensors for this application. Specifically, we assess how microphones, laser vibrometers, and particle velocity meters support void detection. To this end, we first investigate the requirements for excitation of stress waves in the concrete in terms of impact duration and energy. Next, we suggest a data analysis method for void detection based on the difference in the acoustic impedances of concrete, sand, and air. Both numerical modeling and experimental results show the supremacy of microphones in detecting voids behind concrete. We suggest that future studies conduct in-situ experiments to explore how pipe wall reflections and noise influence the performance of a microphone in detecting voids surrounding the concrete sewer pipes.
{"title":"Comparison of sensors for contactless detection of void behind concrete using stress waves","authors":"H. Noshahri, Y. Wijnant, Catalin Cernat, E. Dertien, L. O. Scholtenhuis","doi":"10.1109/SAS51076.2021.9530111","DOIUrl":"https://doi.org/10.1109/SAS51076.2021.9530111","url":null,"abstract":"Detecting voids in pipe surroundings is essential to structural condition assessment of concrete sewer pipelines. Impact-echo is a non-destructive testing method that can be used for this purpose. This method works based on exciting the surface of concrete and using a contact-based sensor to monitor the propagation of the resulting stress waves. However, the presence of deposits and humidity inside the sewer pipe makes establishing a contact between the sensor and the pipe wall very difficult. Therefore, the goal of this study is to compare the performance of contactless sensors for this application. Specifically, we assess how microphones, laser vibrometers, and particle velocity meters support void detection. To this end, we first investigate the requirements for excitation of stress waves in the concrete in terms of impact duration and energy. Next, we suggest a data analysis method for void detection based on the difference in the acoustic impedances of concrete, sand, and air. Both numerical modeling and experimental results show the supremacy of microphones in detecting voids behind concrete. We suggest that future studies conduct in-situ experiments to explore how pipe wall reflections and noise influence the performance of a microphone in detecting voids surrounding the concrete sewer pipes.","PeriodicalId":224327,"journal":{"name":"2021 IEEE Sensors Applications Symposium (SAS)","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133618560","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 : 2021-08-23DOI: 10.1109/SAS51076.2021.9530114
P.M. Janssou, J. Schmalzel, N. Graneau, P. Kaladius, L. Baramidze, I.R. Maduka, J. Medina, E. Jansson, W. McGrath
This is the sixth peer-reviewed publication to report data from our novel, spherical Mach Effect electromagnetic sensor apparatus. The 2020–2021 current results were collected during the recent conjunction of Jupiter and Saturn on December 25th 2020 and again during an experimental run on February 23rd 2021. The observations of these three (3) additional high-sigma $(geq 4sigma)$ anomalies suggest that the Mach Effect is detectable and electromagnetic in its nature. These recent electromagnetic interactions represent observations 9, 10 and 11 completed by a third set of researchers. The novel sensor/detector creates reproducible results and, in such a manner, as to directionally “point” to significant “local” matter implicated as potential sources of inertia in a physical system. The probability that these results could be achieved randomly with zero false positives to date is less than 3.2 E-6. Applying Z-statistic probabilities to only the February outlier reduces this probability further to 8.4 E-9.
{"title":"Novel Mach Effect Sensor's ‘Improbable’ Observations (2016–2021)","authors":"P.M. Janssou, J. Schmalzel, N. Graneau, P. Kaladius, L. Baramidze, I.R. Maduka, J. Medina, E. Jansson, W. McGrath","doi":"10.1109/SAS51076.2021.9530114","DOIUrl":"https://doi.org/10.1109/SAS51076.2021.9530114","url":null,"abstract":"This is the sixth peer-reviewed publication to report data from our novel, spherical Mach Effect electromagnetic sensor apparatus. The 2020–2021 current results were collected during the recent conjunction of Jupiter and Saturn on December 25th 2020 and again during an experimental run on February 23rd 2021. The observations of these three (3) additional high-sigma $(geq 4sigma)$ anomalies suggest that the Mach Effect is detectable and electromagnetic in its nature. These recent electromagnetic interactions represent observations 9, 10 and 11 completed by a third set of researchers. The novel sensor/detector creates reproducible results and, in such a manner, as to directionally “point” to significant “local” matter implicated as potential sources of inertia in a physical system. The probability that these results could be achieved randomly with zero false positives to date is less than 3.2 E-6. Applying Z-statistic probabilities to only the February outlier reduces this probability further to 8.4 E-9.","PeriodicalId":224327,"journal":{"name":"2021 IEEE Sensors Applications Symposium (SAS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131055347","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 : 2021-08-23DOI: 10.1109/SAS51076.2021.9530061
J. Schmalzel, Russell Trafford
Efforts to provide coherent support for interoperability for devices using new technologies remains an ongoing process. Those involved in standards development know this all too well. The challenge remains: How best to keep up with rapidly evolving paradigms such as the IoT, IIoT, Industry 4.0, and similar major shifts? Recent work as part of the IEEE P1451.x standards development looked backwards to identify standards that are mature and extensible, offering a rich functional suite adaptable to new devices and architectures. Standard Commands for Programmable Instruments (SCPI-1999) is one such industry standard, which long ago solved the problem of communicating with devices from many manufacturers. SCPI was an elegant solution for converting unreadable vendor-specific commands to a universal set adaptable by all. Developing interoperability between IoT devices could benefit from a similar SCPI-like approach. To investigate this potential, a typical IoT scenario was developed. Existing nonreadable commands were converted into a SCPI-like command structure. The results of this effort were compelling and suggest that such an approach could achieve interoperability among a wide variety of IoT devices and vendors.
{"title":"SCPI: IoT and the Déjà Vu of Instrument Control","authors":"J. Schmalzel, Russell Trafford","doi":"10.1109/SAS51076.2021.9530061","DOIUrl":"https://doi.org/10.1109/SAS51076.2021.9530061","url":null,"abstract":"Efforts to provide coherent support for interoperability for devices using new technologies remains an ongoing process. Those involved in standards development know this all too well. The challenge remains: How best to keep up with rapidly evolving paradigms such as the IoT, IIoT, Industry 4.0, and similar major shifts? Recent work as part of the IEEE P1451.x standards development looked backwards to identify standards that are mature and extensible, offering a rich functional suite adaptable to new devices and architectures. Standard Commands for Programmable Instruments (SCPI-1999) is one such industry standard, which long ago solved the problem of communicating with devices from many manufacturers. SCPI was an elegant solution for converting unreadable vendor-specific commands to a universal set adaptable by all. Developing interoperability between IoT devices could benefit from a similar SCPI-like approach. To investigate this potential, a typical IoT scenario was developed. Existing nonreadable commands were converted into a SCPI-like command structure. The results of this effort were compelling and suggest that such an approach could achieve interoperability among a wide variety of IoT devices and vendors.","PeriodicalId":224327,"journal":{"name":"2021 IEEE Sensors Applications Symposium (SAS)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115187096","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 : 2021-08-23DOI: 10.1109/SAS51076.2021.9530029
Marc Schroth, Timuçin Etkin, Wilhelm Stork
Recognising human activity can be advantageous in a number of different scenarios including elder care, healthcare or for training purposes. It can be of direct use to support humans in doing different activities, but is still a challenge for systems to correctly classify the activity in a way that is valuable for the user, as they often times lack the robustness or simplicity for day-to-day use. In this paper an approach for human activity recognition based on object interactions is presented. The proposed system consists of a wireless sensor network, with each sensor node measuring the received signal strength indication (RSSI) to its neighbouring nodes. The accumulated RSSI data is then analyzed by a machine learning algorithm which tries to infer one of several cooked dishes from that data. Experimental studies demonstrate promising results and therefore potential for this technology for recognising human activity in the form of cooking, but its generalised approach makes it suitable for other environments, too.
{"title":"A novel approach for human activity recognition using object interactions and machine learning","authors":"Marc Schroth, Timuçin Etkin, Wilhelm Stork","doi":"10.1109/SAS51076.2021.9530029","DOIUrl":"https://doi.org/10.1109/SAS51076.2021.9530029","url":null,"abstract":"Recognising human activity can be advantageous in a number of different scenarios including elder care, healthcare or for training purposes. It can be of direct use to support humans in doing different activities, but is still a challenge for systems to correctly classify the activity in a way that is valuable for the user, as they often times lack the robustness or simplicity for day-to-day use. In this paper an approach for human activity recognition based on object interactions is presented. The proposed system consists of a wireless sensor network, with each sensor node measuring the received signal strength indication (RSSI) to its neighbouring nodes. The accumulated RSSI data is then analyzed by a machine learning algorithm which tries to infer one of several cooked dishes from that data. Experimental studies demonstrate promising results and therefore potential for this technology for recognising human activity in the form of cooking, but its generalised approach makes it suitable for other environments, too.","PeriodicalId":224327,"journal":{"name":"2021 IEEE Sensors Applications Symposium (SAS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115439562","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 : 2021-08-23DOI: 10.1109/SAS51076.2021.9530181
Marcello Zanghieri, A. Burrello, S. Benatti, Kaspar Anton Schindler, L. Benini
Epilepsy is a severe neurological disorder that affects about 1 % of the world population, and one-third of cases are drug-resistant. Apart from surgery, drug-resistant patients can benefit from closed-loop brain stimulation, eliminating or mitigating the epileptic symptoms. For the closed-loop to be accurate and safe, it is paramount to couple stimulation with a detection system able to recognize seizure onset with high sensitivity and specificity and short latency, while meeting the strict computation and energy constraints of always-on realtime monitoring platforms. We propose a novel setup for iEEG-based epilepsy detection, exploiting a Temporal Convolutional Network (TCN) optimized for deployability on low-power edge devices for real-time monitoring. We test our approach on the Short- Term SWEC-ETHZ iEEG Database, containing a total of 100 epileptic seizures from 16 patients (from 2 to 14 per patient) comparing it with the state-of-the-art (SoA) approach, represented by Hyperdimensional Computing (HD). Our TCN attains a detection delay which is 10s better than SoA, without performance drop in sensitivity and specificity. Contrary to previous literature, we also enforce a time-consistent setup, where training seizures always precede testing seizures chronologically. When deployed on a commercial low-power parallel microcontroller unit (MCU), each inference with our model has a latency of only 5.68 ms and an energy cost of only 124.5 μJ if executed on 1 core, and latency 1.46 ms and an energy cost 51.2 μJ if parallelized on 8 cores. These latency and energy consumption, lower than the current SoA, demonstrates the suitability of our solution for real-time long-term embedded epilepsy monitoring.
{"title":"Low-Latency Detection of Epileptic Seizures from iEEG with Temporal Convolutional Networks on a Low-Power Parallel MCU","authors":"Marcello Zanghieri, A. Burrello, S. Benatti, Kaspar Anton Schindler, L. Benini","doi":"10.1109/SAS51076.2021.9530181","DOIUrl":"https://doi.org/10.1109/SAS51076.2021.9530181","url":null,"abstract":"Epilepsy is a severe neurological disorder that affects about 1 % of the world population, and one-third of cases are drug-resistant. Apart from surgery, drug-resistant patients can benefit from closed-loop brain stimulation, eliminating or mitigating the epileptic symptoms. For the closed-loop to be accurate and safe, it is paramount to couple stimulation with a detection system able to recognize seizure onset with high sensitivity and specificity and short latency, while meeting the strict computation and energy constraints of always-on realtime monitoring platforms. We propose a novel setup for iEEG-based epilepsy detection, exploiting a Temporal Convolutional Network (TCN) optimized for deployability on low-power edge devices for real-time monitoring. We test our approach on the Short- Term SWEC-ETHZ iEEG Database, containing a total of 100 epileptic seizures from 16 patients (from 2 to 14 per patient) comparing it with the state-of-the-art (SoA) approach, represented by Hyperdimensional Computing (HD). Our TCN attains a detection delay which is 10s better than SoA, without performance drop in sensitivity and specificity. Contrary to previous literature, we also enforce a time-consistent setup, where training seizures always precede testing seizures chronologically. When deployed on a commercial low-power parallel microcontroller unit (MCU), each inference with our model has a latency of only 5.68 ms and an energy cost of only 124.5 μJ if executed on 1 core, and latency 1.46 ms and an energy cost 51.2 μJ if parallelized on 8 cores. These latency and energy consumption, lower than the current SoA, demonstrates the suitability of our solution for real-time long-term embedded epilepsy monitoring.","PeriodicalId":224327,"journal":{"name":"2021 IEEE Sensors Applications Symposium (SAS)","volume":"220 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123974027","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 : 2021-08-23DOI: 10.1109/SAS51076.2021.9530094
A. Depari, P. Bellagente, P. Ferrari, A. Flammini, M. Pasetti, S. Rinaldi, E. Sisinni
The monitoring of pollutants in industrial plants is a major concern, in order to satisfy the requirements dictated by the related norms. Recently, the problem of odor monitoring gained importance since, despite the generally low dangerous nature of the emission, people usually correlate bad smell to unhealthy air condition. In this paper, we focus on the wastewater treatment application scenario and propose a versatile air pollution control solution. In particular, a distributed eN ose, based on low-cost, but highly configurable sensing probes, is suggested. The Internet of Things (IoT) paradigm has been followed, thanks to the adoption of a communication infrastructure based on the widely adopted LoRaWAN technology. Such an approach, on one hand, would allow to easily cope with the administration and local population; on the other hand, it paves the way to use analytics to predict emission events in advance. The need for managing multiple transducers per each probe has been solved using an innovative acquisition strategy, exploiting both the volt-amperometric and resistance-to-time (integral) approaches, implemented by low-cost microcontroller and ancillary electronics. A proof-of-concept prototype has been realized and preliminary experiments have been carried out. Experimental results have demonstrated the solidity of the proposed approach, with relative error below 1 % and relative standard deviation below 0.5% over the whole considered resistive range of more than two decades.
{"title":"Versatile and low-cost sensor interface for IoT-ready odor monitoring in wastewater management","authors":"A. Depari, P. Bellagente, P. Ferrari, A. Flammini, M. Pasetti, S. Rinaldi, E. Sisinni","doi":"10.1109/SAS51076.2021.9530094","DOIUrl":"https://doi.org/10.1109/SAS51076.2021.9530094","url":null,"abstract":"The monitoring of pollutants in industrial plants is a major concern, in order to satisfy the requirements dictated by the related norms. Recently, the problem of odor monitoring gained importance since, despite the generally low dangerous nature of the emission, people usually correlate bad smell to unhealthy air condition. In this paper, we focus on the wastewater treatment application scenario and propose a versatile air pollution control solution. In particular, a distributed eN ose, based on low-cost, but highly configurable sensing probes, is suggested. The Internet of Things (IoT) paradigm has been followed, thanks to the adoption of a communication infrastructure based on the widely adopted LoRaWAN technology. Such an approach, on one hand, would allow to easily cope with the administration and local population; on the other hand, it paves the way to use analytics to predict emission events in advance. The need for managing multiple transducers per each probe has been solved using an innovative acquisition strategy, exploiting both the volt-amperometric and resistance-to-time (integral) approaches, implemented by low-cost microcontroller and ancillary electronics. A proof-of-concept prototype has been realized and preliminary experiments have been carried out. Experimental results have demonstrated the solidity of the proposed approach, with relative error below 1 % and relative standard deviation below 0.5% over the whole considered resistive range of more than two decades.","PeriodicalId":224327,"journal":{"name":"2021 IEEE Sensors Applications Symposium (SAS)","volume":"45 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122547678","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：