Wan Du, Zikun Xing, Mo Li, Bingsheng He, L. Chua, Haiyan Miao
We collaborate with environmental scientists to study the hydrodynamics and water quality in an urban district, where the surface wind distribution is an essential input but undergoes high spatial and temporal variations due to the complex urban landform created by surrounding buildings. In this work, we study an optimal sensor placement scheme to measure the wind distribution over a large urban reservoir with a limited number of wind sensors. Unlike existing sensor placement solutions that assume Gaussian process of target phenomena, this study measures the wind which inherently exhibits strong non-Gaussian yearly distribution. By leveraging the local monsoon characteristics of wind, we segment a year into different monsoon seasons which follow a unique distribution respectively. We also use computational fluid dynamics to learn the spatial correlation of wind in the presence of surrounding buildings. The output of sensor placement is a set of the most informative locations to deploy the wind sensors, based on the readings of which we can accurately predict the wind over the entire reservoir surface in real time. 10 wind sensors are finally deployed around or on the water surface of an urban reservoir. The in-field measurement results of more than 3 months suggest that the proposed sensor placement and spatial prediction approach provides accurate wind measurement which outperforms the state-of-the-art Gaussian model based or interpolation based approaches.
{"title":"Optimal sensor placement and measurement of wind for water quality studies in urban reservoirs","authors":"Wan Du, Zikun Xing, Mo Li, Bingsheng He, L. Chua, Haiyan Miao","doi":"10.1145/2700265","DOIUrl":"https://doi.org/10.1145/2700265","url":null,"abstract":"We collaborate with environmental scientists to study the hydrodynamics and water quality in an urban district, where the surface wind distribution is an essential input but undergoes high spatial and temporal variations due to the complex urban landform created by surrounding buildings. In this work, we study an optimal sensor placement scheme to measure the wind distribution over a large urban reservoir with a limited number of wind sensors. Unlike existing sensor placement solutions that assume Gaussian process of target phenomena, this study measures the wind which inherently exhibits strong non-Gaussian yearly distribution. By leveraging the local monsoon characteristics of wind, we segment a year into different monsoon seasons which follow a unique distribution respectively. We also use computational fluid dynamics to learn the spatial correlation of wind in the presence of surrounding buildings. The output of sensor placement is a set of the most informative locations to deploy the wind sensors, based on the readings of which we can accurately predict the wind over the entire reservoir surface in real time. 10 wind sensors are finally deployed around or on the water surface of an urban reservoir. The in-field measurement results of more than 3 months suggest that the proposed sensor placement and spatial prediction approach provides accurate wind measurement which outperforms the state-of-the-art Gaussian model based or interpolation based approaches.","PeriodicalId":297218,"journal":{"name":"IPSN-14 Proceedings of the 13th International Symposium on Information Processing in Sensor Networks","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-02-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127754354","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 : 2014-04-15DOI: 10.1109/IPSN.2014.6846793
Tong Kun Lai, An-Ping Wang, Chun-Min Chang, Hua-Min Tseng, Kailing Huang, Jo-Ping Li, Wen-Chan Shih, P. Chou
This demo presents the world's smallest wireless motion-sensing platform based on Bluetooth 4.0 Low Energy (BLE) Technology. It is merely 8×8 mm2 in area but is complete with a user-programmable microcontroller and integrated RF, a digital triaxial accelerometer with programmable threshold detection, and a real-time clock and calendar chip, and a magnetic sensor/switch. This system has been used in a number of applications, including a proximity tag, a pedometer, an air mouse with gesture recognition, and a BLE-to-IR remote controller.
{"title":"Demonstration abstract: An 8×8 mm2 Bluetooth Low Energy wireless motion-sensing platform","authors":"Tong Kun Lai, An-Ping Wang, Chun-Min Chang, Hua-Min Tseng, Kailing Huang, Jo-Ping Li, Wen-Chan Shih, P. Chou","doi":"10.1109/IPSN.2014.6846793","DOIUrl":"https://doi.org/10.1109/IPSN.2014.6846793","url":null,"abstract":"This demo presents the world's smallest wireless motion-sensing platform based on Bluetooth 4.0 Low Energy (BLE) Technology. It is merely 8×8 mm2 in area but is complete with a user-programmable microcontroller and integrated RF, a digital triaxial accelerometer with programmable threshold detection, and a real-time clock and calendar chip, and a magnetic sensor/switch. This system has been used in a number of applications, including a proximity tag, a pedometer, an air mouse with gesture recognition, and a BLE-to-IR remote controller.","PeriodicalId":297218,"journal":{"name":"IPSN-14 Proceedings of the 13th International Symposium on Information Processing in Sensor Networks","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-04-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128798887","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 : 2014-04-15DOI: 10.1109/IPSN.2014.6846751
M. Cattani, Marco Zúñiga, Andreas Loukas, K. Langendoen
We address the problem of estimating the neighborhood cardinality of nodes in dynamic wireless networks. Different from previous studies, we consider networks with high densities (a hundred neighbors per node) and where all nodes estimate cardinality concurrently. Performing concurrent estimations on dense mobile networks is hard; we need estimators that are not only accurate, but also fast, asynchronous (due to mobility) and lightweight (due to concurrency and high density). To cope with these requirements, we propose Estreme, a neighborhood cardinality estimator with extremely low overhead that leverages the rendezvous time of low-power medium access control (MAC) protocols. We implemented Estreme on the Contiki OS and show a significant improvement over the state-of-the-art. With Estreme, 100 nodes can concurrently estimate their neighborhood cardinality with an error of ≈10%. State-of-the-art solutions provide a similar accuracy, but on networks consisting of a few tens of nodes and where only a fraction of nodes estimate the cardinality concurrently.
{"title":"Lightweight neighborhood cardinality estimation in dynamic wireless networks","authors":"M. Cattani, Marco Zúñiga, Andreas Loukas, K. Langendoen","doi":"10.1109/IPSN.2014.6846751","DOIUrl":"https://doi.org/10.1109/IPSN.2014.6846751","url":null,"abstract":"We address the problem of estimating the neighborhood cardinality of nodes in dynamic wireless networks. Different from previous studies, we consider networks with high densities (a hundred neighbors per node) and where all nodes estimate cardinality concurrently. Performing concurrent estimations on dense mobile networks is hard; we need estimators that are not only accurate, but also fast, asynchronous (due to mobility) and lightweight (due to concurrency and high density). To cope with these requirements, we propose Estreme, a neighborhood cardinality estimator with extremely low overhead that leverages the rendezvous time of low-power medium access control (MAC) protocols. We implemented Estreme on the Contiki OS and show a significant improvement over the state-of-the-art. With Estreme, 100 nodes can concurrently estimate their neighborhood cardinality with an error of ≈10%. State-of-the-art solutions provide a similar accuracy, but on networks consisting of a few tens of nodes and where only a fraction of nodes estimate the cardinality concurrently.","PeriodicalId":297218,"journal":{"name":"IPSN-14 Proceedings of the 13th International Symposium on Information Processing in Sensor Networks","volume":"58 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-04-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124501456","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 : 2014-04-15DOI: 10.1109/IPSN.2014.6846742
Syed Monowar Hossain, A. Ali, Md. Mahbubur Rahman, Emre Ertin, D. Epstein, Ashley Kennedy, K. Preston, A. Umbricht, Yixin Chen, Santosh Kumar
A variety of health and behavioral states can potentially be inferred from physiological measurements that can now be collected in the natural free-living environment. The major challenge, however, is to develop computational models for automated detection of health events that can work reliably in the natural field environment. In this paper, we develop a physiologically-informed model to automatically detect drug (cocaine) use events in the free-living environment of participants from their electrocardiogram (ECG) measurements. The key to reliably detecting drug use events in the field is to incorporate the knowledge of autonomic nervous system (ANS) behavior in the model development so as to decompose the activation effect of cocaine from the natural recovery behavior of the parasympathetic nervous system (after an episode of physical activity). We collect 89 days of data from 9 active drug users in two residential lab environments and 922 days of data from 42 active drug users in the field environment, for a total of 11,283 hours. We develop a model that tracks the natural recovery by the parasympathetic nervous system and then estimates the dampening caused to the recovery by the activation of the sympathetic nervous system due to cocaine. We develop efficient methods to screen and clean the ECG time series data and extract candidate windows to assess for potential drug use. We then apply our model on the recovery segments from these windows. Our model achieves 100% true positive rate while keeping the false positive rate to 0.87/day over (9+ hours/day of) lab data and to 1.13/day over (11+ hours/day of) field data.
{"title":"Identifying drug (cocaine) intake events from acute physiological response in the presence of free-living physical activity","authors":"Syed Monowar Hossain, A. Ali, Md. Mahbubur Rahman, Emre Ertin, D. Epstein, Ashley Kennedy, K. Preston, A. Umbricht, Yixin Chen, Santosh Kumar","doi":"10.1109/IPSN.2014.6846742","DOIUrl":"https://doi.org/10.1109/IPSN.2014.6846742","url":null,"abstract":"A variety of health and behavioral states can potentially be inferred from physiological measurements that can now be collected in the natural free-living environment. The major challenge, however, is to develop computational models for automated detection of health events that can work reliably in the natural field environment. In this paper, we develop a physiologically-informed model to automatically detect drug (cocaine) use events in the free-living environment of participants from their electrocardiogram (ECG) measurements. The key to reliably detecting drug use events in the field is to incorporate the knowledge of autonomic nervous system (ANS) behavior in the model development so as to decompose the activation effect of cocaine from the natural recovery behavior of the parasympathetic nervous system (after an episode of physical activity). We collect 89 days of data from 9 active drug users in two residential lab environments and 922 days of data from 42 active drug users in the field environment, for a total of 11,283 hours. We develop a model that tracks the natural recovery by the parasympathetic nervous system and then estimates the dampening caused to the recovery by the activation of the sympathetic nervous system due to cocaine. We develop efficient methods to screen and clean the ECG time series data and extract candidate windows to assess for potential drug use. We then apply our model on the recovery segments from these windows. Our model achieves 100% true positive rate while keeping the false positive rate to 0.87/day over (9+ hours/day of) lab data and to 1.13/day over (11+ hours/day of) field data.","PeriodicalId":297218,"journal":{"name":"IPSN-14 Proceedings of the 13th International Symposium on Information Processing in Sensor Networks","volume":"65 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-04-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127617706","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 : 2014-04-15DOI: 10.1109/IPSN.2014.6846797
S. Thomas, Viswam Nathan, Chengzhi Zong, A. L. P. Aroul, Lijoy Philipose, K. Soundarapandian, Xiangrong Shi, R. Jafari
A wrist watch based system, which can measure electrocardiogram (ECG) and photoplethysmogram (PPG), is presented in this work. By using both ECG and PPG we measure pulse transit time (PTT), which is known to correlate well with the blood pressure (BP) [1]. This system, called BioWatch, can potentially facilitate continuous and ubiquitous monitoring of ECG, PPG and BP.
{"title":"Demonstration abstract: BioWatch — A wrist watch based physiological signal acquisition system","authors":"S. Thomas, Viswam Nathan, Chengzhi Zong, A. L. P. Aroul, Lijoy Philipose, K. Soundarapandian, Xiangrong Shi, R. Jafari","doi":"10.1109/IPSN.2014.6846797","DOIUrl":"https://doi.org/10.1109/IPSN.2014.6846797","url":null,"abstract":"A wrist watch based system, which can measure electrocardiogram (ECG) and photoplethysmogram (PPG), is presented in this work. By using both ECG and PPG we measure pulse transit time (PTT), which is known to correlate well with the blood pressure (BP) [1]. This system, called BioWatch, can potentially facilitate continuous and ubiquitous monitoring of ECG, PPG and BP.","PeriodicalId":297218,"journal":{"name":"IPSN-14 Proceedings of the 13th International Symposium on Information Processing in Sensor Networks","volume":"270 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-04-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133842432","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 : 2014-04-15DOI: 10.1109/IPSN.2014.6846771
Abdelraouf Ouadjaout, Noureddine Lasla, Miloud Bagaa, N. Badache
In this paper, we present SADA, a static analysis tool to verify device drivers for TinyOS applications. Its broad goal is to certify that the execution paths of the application complies with a given hardware specification. SADA can handle a broad spectrum of hardware specifications, ranging from simple assertions about the values of configuration registers, to complex behaviors of possibly several connected hardware components. The hardware specification is expressed in BIP, a language for describing easily complex interacting discrete components. The analysis of the joint behavior of the application and the hardware specification is then performed using the theory of Abstract Interpretation. We have done a set of experiments on some TinyOS applications. Encouraging results are obtained that confirm the effectiveness of our approach.
{"title":"Poster abstract: Static analysis of device drivers in TinyOS","authors":"Abdelraouf Ouadjaout, Noureddine Lasla, Miloud Bagaa, N. Badache","doi":"10.1109/IPSN.2014.6846771","DOIUrl":"https://doi.org/10.1109/IPSN.2014.6846771","url":null,"abstract":"In this paper, we present SADA, a static analysis tool to verify device drivers for TinyOS applications. Its broad goal is to certify that the execution paths of the application complies with a given hardware specification. SADA can handle a broad spectrum of hardware specifications, ranging from simple assertions about the values of configuration registers, to complex behaviors of possibly several connected hardware components. The hardware specification is expressed in BIP, a language for describing easily complex interacting discrete components. The analysis of the joint behavior of the application and the hardware specification is then performed using the theory of Abstract Interpretation. We have done a set of experiments on some TinyOS applications. Encouraging results are obtained that confirm the effectiveness of our approach.","PeriodicalId":297218,"journal":{"name":"IPSN-14 Proceedings of the 13th International Symposium on Information Processing in Sensor Networks","volume":"74 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-04-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124888975","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 : 2014-04-15DOI: 10.1109/IPSN.2014.6846777
K. Subbu, Neethu Thomas
Smart control systems infer user locations through wireless sensor networks, active and passive RFIDs, NFCs etc., to name a few. This involves considerable amount of installation and sensor data managing. We present a personalized smart control system that learns user location and controls appliances present at the user's location. The only sensor we use is the smartphone itself through its embedded magnetic field and light sensors for classifying user locations and detecting light intensity in rooms respectively.
{"title":"Poster abstract: A location aware personalized smart control system","authors":"K. Subbu, Neethu Thomas","doi":"10.1109/IPSN.2014.6846777","DOIUrl":"https://doi.org/10.1109/IPSN.2014.6846777","url":null,"abstract":"Smart control systems infer user locations through wireless sensor networks, active and passive RFIDs, NFCs etc., to name a few. This involves considerable amount of installation and sensor data managing. We present a personalized smart control system that learns user location and controls appliances present at the user's location. The only sensor we use is the smartphone itself through its embedded magnetic field and light sensors for classifying user locations and detecting light intensity in rooms respectively.","PeriodicalId":297218,"journal":{"name":"IPSN-14 Proceedings of the 13th International Symposium on Information Processing in Sensor Networks","volume":"41 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-04-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128519769","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 : 2014-04-15DOI: 10.1109/IPSN.2014.6846778
Manoj R. Rege, V. Handziski, A. Wolisz
Predicting performance of crowd-sensing applications at large scale, in the pre-deployment phase, represents a significant challenge for developers. We demonstrate a solution to this problem in the form of a cloud-based emulation platform called CrowdMeter. Our platform emulates mobile devices and access network links, models human factors in crowd-sensing, and leverages virtualization through cloud infrastructure-as-service resources to model large scale crowd-sensing. In this demo we exhibit the capabilities of CrowdMeter by deploying VideoQuest, a simple crowd-sensing application, on hundreds of emulated mobile devices, and by measuring its performance.
{"title":"Demonstration abstract: CrowdMeter — Predicting performance of crowd-sensing applications using emulations","authors":"Manoj R. Rege, V. Handziski, A. Wolisz","doi":"10.1109/IPSN.2014.6846778","DOIUrl":"https://doi.org/10.1109/IPSN.2014.6846778","url":null,"abstract":"Predicting performance of crowd-sensing applications at large scale, in the pre-deployment phase, represents a significant challenge for developers. We demonstrate a solution to this problem in the form of a cloud-based emulation platform called CrowdMeter. Our platform emulates mobile devices and access network links, models human factors in crowd-sensing, and leverages virtualization through cloud infrastructure-as-service resources to model large scale crowd-sensing. In this demo we exhibit the capabilities of CrowdMeter by deploying VideoQuest, a simple crowd-sensing application, on hundreds of emulated mobile devices, and by measuring its performance.","PeriodicalId":297218,"journal":{"name":"IPSN-14 Proceedings of the 13th International Symposium on Information Processing in Sensor Networks","volume":"165 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-04-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127390892","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 : 2014-04-15DOI: 10.1109/IPSN.2014.6846775
Liqiong Chang, Ju Wang, Dingyi Fang, Xiaojiang Chen, Tianzhang Xing, Weike Nie
While device-free localization (DFL), i.e., target without carrying any devices, is attractive, the need for a significant degree of pre-deployment effort is a key challenge. In this paper, we proposed the NDP method, which do not assume knowledge of the placement layout, including the parameters of the transceivers. The key intuition is that all of the distorted wireless signals caused by the target, even the many from unknown transceivers, are constrained the presence of the target(s). Despite the absence of any explicit pre-deployment calibration, NDP yields a 80th percentile error of 1.8m which is much butter than the 6.2m, 5.8m and 5m yielded by the three state-of-the-art algorithms from prior work.
{"title":"Poster abstract: NDP — A novel device-free localization method with little efforts","authors":"Liqiong Chang, Ju Wang, Dingyi Fang, Xiaojiang Chen, Tianzhang Xing, Weike Nie","doi":"10.1109/IPSN.2014.6846775","DOIUrl":"https://doi.org/10.1109/IPSN.2014.6846775","url":null,"abstract":"While device-free localization (DFL), i.e., target without carrying any devices, is attractive, the need for a significant degree of pre-deployment effort is a key challenge. In this paper, we proposed the NDP method, which do not assume knowledge of the placement layout, including the parameters of the transceivers. The key intuition is that all of the distorted wireless signals caused by the target, even the many from unknown transceivers, are constrained the presence of the target(s). Despite the absence of any explicit pre-deployment calibration, NDP yields a 80th percentile error of 1.8m which is much butter than the 6.2m, 5.8m and 5m yielded by the three state-of-the-art algorithms from prior work.","PeriodicalId":297218,"journal":{"name":"IPSN-14 Proceedings of the 13th International Symposium on Information Processing in Sensor Networks","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-04-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130112141","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 : 2014-04-15DOI: 10.1109/IPSN.2014.6846756
Yiran Shen, W. Hu, Mingrui Yang, Bo Wei, S. Lucey, C. Chou
Face recognition is an element of many smartphone apps, e.g. face unlocking, people tagging and games. Sparse Representation Classification (SRC) is a state-of-the-art face recognition algorithm, which has been shown to outperform many classical face recognition algorithms in OpenCV. The success of SRC is due to its use of ℓ1 optimisation, which makes SRC robust to noise and occlusions. Since ℓ1 optimisation is computationally intensive, SRC uses random projection matrices to reduce the dimension of the ℓ1 problem. However, random projection matrices do not give consistent classification accuracy. In this paper, we propose a method to optimise the projection matrix for ℓ1-based classification1. Our evaluations, based on publicly available databases and real experiment, show that face recognition based on the optimised projection matrix can be 5-17% more accurate than its random counterpart and OpenCV algorithms. Furthermore, the optimised projection matrix does not have to be re-calculated even if new faces are added to the training set. We implement the SRC with optimised projection matrix on Android smartphones and find that the computation of residuals in SRC is a severe bottleneck, taking up 85-90% of the computation time. To address this problem, we propose a method to compute the residuals approximately, which is 50 times faster but without sacrificing recognition accuracy. Lastly, we demonstrate the feasibility of our new algorithm by the implementation and evaluation of a new face unlocking app and show its robustness to variation to poses, facial expressions, lighting changes and occlusions.
{"title":"Face recognition on smartphones via optimised Sparse Representation Classification","authors":"Yiran Shen, W. Hu, Mingrui Yang, Bo Wei, S. Lucey, C. Chou","doi":"10.1109/IPSN.2014.6846756","DOIUrl":"https://doi.org/10.1109/IPSN.2014.6846756","url":null,"abstract":"Face recognition is an element of many smartphone apps, e.g. face unlocking, people tagging and games. Sparse Representation Classification (SRC) is a state-of-the-art face recognition algorithm, which has been shown to outperform many classical face recognition algorithms in OpenCV. The success of SRC is due to its use of ℓ1 optimisation, which makes SRC robust to noise and occlusions. Since ℓ1 optimisation is computationally intensive, SRC uses random projection matrices to reduce the dimension of the ℓ1 problem. However, random projection matrices do not give consistent classification accuracy. In this paper, we propose a method to optimise the projection matrix for ℓ1-based classification1. Our evaluations, based on publicly available databases and real experiment, show that face recognition based on the optimised projection matrix can be 5-17% more accurate than its random counterpart and OpenCV algorithms. Furthermore, the optimised projection matrix does not have to be re-calculated even if new faces are added to the training set. We implement the SRC with optimised projection matrix on Android smartphones and find that the computation of residuals in SRC is a severe bottleneck, taking up 85-90% of the computation time. To address this problem, we propose a method to compute the residuals approximately, which is 50 times faster but without sacrificing recognition accuracy. Lastly, we demonstrate the feasibility of our new algorithm by the implementation and evaluation of a new face unlocking app and show its robustness to variation to poses, facial expressions, lighting changes and occlusions.","PeriodicalId":297218,"journal":{"name":"IPSN-14 Proceedings of the 13th International Symposium on Information Processing in Sensor Networks","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-04-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122221396","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：