Sachi Desai - Academia.edu (original) (raw)
Papers by Sachi Desai
Feature extraction methods based on the discrete wavelet transform (DWT) and multiresolution anal... more Feature extraction methods based on the discrete wavelet transform (DWT) and multiresolution analysis are used to develop a robust classification algorithm that reliably discriminates between launch and impact artillery and/or mortar events via acoustic signals produced during detonation. Distinct characteristics are found within the acoustic signatures since impact events emphasize concussive and shrapnel effects, while launch events are similar to explosions, designed to expel and propel artillery round from a gun. The ensuing blast waves are readily characterized by variations in the corresponding peak pressure and rise time of the waveform, differences in the ratio of positive pressure amplitude to the negative amplitude, variations in the prominent frequencies associated with the varying blast events and variations in the overall duration of the resulting waveform. Unique attributes can also be identified that depend upon the properties of the gun tube, projectile speed at the ...
The major problem of online learning or incremental learning is that, target function is frequent... more The major problem of online learning or incremental learning is that, target function is frequently changing over time. This problem is commonly known as concept drift. Concept drift can be is further complicated if the dataset is classimbalanced. There are different learning methods presented so far to handle concept drift like rule-based systems, decision trees, Naive Bayes, support vector machines, instance based learning, ensemble of classifiers, etc. This learning method requires further to combined with methods of drift detection in order to constantly monitor the performance of concept drift, however online changes detection was failed. In literature there are many methods presented for learning from data streams and drift detection, but most of methods failed to achieve speed and accuracy due to data inconsistency. In this project our goal is to present efficient method for online and non-parametric drift detection. This proposed method is based on recently presented Hoeffdings Bounds and HDDM. It handles concept drift regardless of the learning model to monitor the performance metrics measured during the learning process, to trigger drift signals when a significant variation has been detected. The existing system however as Naive Bayes classifier are having limitations, there is no scope to improve accuracy of HDDM. The Propose system will be efficiently provide drift detection method for data stream mining to improve accuracy.
Unattended Ground, Sea, and Air Sensor Technologies and Applications X, 2008
In this paper we discuss a random walk model to characterize the pulse discharge battery process.... more In this paper we discuss a random walk model to characterize the pulse discharge battery process. Several theoretical results are derived including the mean and variance of an unattended battery-driven sensor lifetime. Some numerical results are presented.
spie.org
On-Site Proceedings Abstract Due Date: 3 October 2005 Manuscript Due Date: 23 January 2006 Cooper... more On-Site Proceedings Abstract Due Date: 3 October 2005 Manuscript Due Date: 23 January 2006 Cooperating Organizations [logo] International Neural Network Society (INNS) [logo] IEEE Computational.
: Agenda: Welcome Remarks, Joint Science and Technology, Doing Business with USSOCOM, Reducing th... more : Agenda: Welcome Remarks, Joint Science and Technology, Doing Business with USSOCOM, Reducing the Threat of Nuclear and Radiological Terrorism, Avon M53 Protective Mask for USSOCOM, Chemical Homeland Security Suite (C-HoSS), Radiological Emergency Response, Use of Recombinant Butyrylcholinesterase in Responding to Chemical Weapon Attack, Reliable Discrimination of High Explosive/Chem/Bio/Artillery Using Acoustic IGS, Real-Time First Bite Detection, Polymer Technology for the Lock-Down/Removal of Radiological Contamination, Modeling Tool for Prediction and Mitigation of CBRNE Events, Terrorist Motivations to Employ CBRN Weapons, USAF Counter-Biological Warfare Effort, Keynote Presentation, Question and Answer Session, Responding to Multiple Ebola Attacks: The Need for Coordinated Preparedness, Capture, Contain, Treat and Dispose of Decontamination Runoff on Site, National Guard CBRN Response: Achieving Unity of Effort Between Local/State/Federal, CBRN Detectors for Early Warning of ...
SPIE Proceedings, 2007
Feature extraction methods based on the discrete wavelet transform and multiresolution analysis f... more Feature extraction methods based on the discrete wavelet transform and multiresolution analysis facilitate the development of a robust classification algorithm that reliably discriminates mortar and artillery variants via acoustic signals produced during the launch/impact events. Utilizing acoustic sensors to exploit the sound waveform generated from the blast for the identification of mortar and artillery variants. Distinct characteristics arise within the
Unattended Ground, Sea, and Air Sensor Technologies and Applications XI, 2009
Herein is described the U.S. Army RDECOM-ARDEC's purpose and series of activities conducted a... more Herein is described the U.S. Army RDECOM-ARDEC's purpose and series of activities conducted at the 2008 NATO SET-093 TG-53 experiment and field test. The overall purpose of the field test as stated by SET-093 panel was to provide a baseline test capable of providing relevant scenarios and data regarding a variety of impulsive generated acoustic events. As organized, the field
Chemical, Biological, Radiological, Nuclear, and Explosives (CBRNE) Sensing IX, 2008
Integrating a sensor suite with ability to discriminate potential Chemical/Biological (CB) events... more Integrating a sensor suite with ability to discriminate potential Chemical/Biological (CB) events from high-explosive (HE) events employing an acoustic sensor array with a Time Difference of Arrival (TDOA) algorithm. Developing a cueing mechanism for more power intensive and range limited sensing CB techniques. Enabling the event detection algorithm to locate to a blast event using TDOA further information is provided
2012 IEEE International Symposium on Precision Clock Synchronization for Measurement, Control and Communication Proceedings, 2012
ABSTRACT Detection of events using a network of simple field sensors has gained interest due to i... more ABSTRACT Detection of events using a network of simple field sensors has gained interest due to its low cost and robustness. Sensor networks have been extensively analyzed recently in terms of stability, robustness, and efficiency. Time synchronization has proven to be critical in sensor fusion applications where time of arrival is a decision property, and thus an accurate common time reference is required. In this work, we analyze the dependence on time synchronization of an acoustic event detection system, and we present a bio-inspired synchronization algorithm for wireless sensor networks capable of providing the system with a common time reference to enable accurate detection.
SPIE Proceedings, 2008
ABSTRACT Explosion detection and recognition is a critical capability to provide situational awar... more ABSTRACT Explosion detection and recognition is a critical capability to provide situational awareness to the war-fighters in battlefield. Acoustic sensors are frequently deployed to detect such events and to trigger more expensive sensing/sensor modalities (i.e. radar, laser spectroscope, IR etc.). Acoustic analysis of explosions has been intensively studied to reliably discriminate mortars, artillery, round variations, and type of blast (i.e. chemical/biological or high-explosive). One of the major challenges is high level of noise, which may include non-coherent noise generated from the environmental background and coherent noise induced by possible mobile acoustic sensor platform. In this work, we introduce a new acoustic scene analysis method to effectively enhance explosion classification reliability and reduce the false alarm rate at low SNR and with high coherent noise. The proposed method is based on acoustic signature modeling using Hidden Markov Models (HMMs). Special frequency domain acoustic features characterizing explosions as well as coherent noise are extracted from each signal segment, which forms an observation vector for HMM training and test. Classification is based on a unique model similarity measure between the HMM estimated from the test observations and the trained HMMs. Experimental tests are based on the acoustic explosion dataset from US ARMY ARDEC, and experimental results have demonstrated the effectiveness of the proposed method.
Author (s): Daniel Lee; Mark McClelland; Joseph Schneider; Tsung-Lin Yang; Dan Gallagher; John Wa... more Author (s): Daniel Lee; Mark McClelland; Joseph Schneider; Tsung-Lin Yang; Dan Gallagher; John Wang; Danelle Shah; Nisar Ahmed; Pete Moran; Brandon Jones; Tung-Sing Leung; Aaron Nathan; Hadas Kress-Gazit; Mark Campbell
SPIE Proceedings, 2008
The coordination of a multi-robot system searching for multi targets is challenging under dynamic... more The coordination of a multi-robot system searching for multi targets is challenging under dynamic environment since the multi-robot system demands group coherence (agents need to have the incentive to work together faithfully) and group competence (agents need to know how to work together well). In our previous proposed bio-inspired coordination method, Local Interaction through Virtual Stigmergy (LIVS), one problem is the considerable randomness of the robot movement during coordination, which may lead to more power consumption and longer searching time. To address these issues, an adaptive LIVS (ALIVS) method is proposed in this paper, which not only considers the travel cost and target weight, but also predicting the target/robot ratio and potential robot redundancy with respect to the detected targets. Furthermore, a dynamic weight adjustment is also applied to improve the searching performance. This new method a truly distributed method where each robot makes its own decision based on its local sensing information and the information from its neighbors. Basically, each robot only communicates with its neighbors through a virtual stigmergy mechanism and makes its local movement decision based on a Particle Swarm Optimization (PSO) algorithm. The proposed ALIVS algorithm has been implemented on the embodied robot simulator, Player/Stage, in a searching target. The simulation results demonstrate the efficiency and robustness in a power-efficient manner with the real-world constraints.
Control Engineering Practice, 2017
Time synchronization has proven to be critical in sensor fusion applications where the time of ar... more Time synchronization has proven to be critical in sensor fusion applications where the time of arrival is utilized as a decision variable. Herein, the application of pulse-coupled synchronization to an acoustic event detection system based on a wireless sensor network is presented. The aim of the system is to locate the source of acoustic events utilizing time of arrival measurements for different formations of the sensor network. A distributed localization algorithm is introduced that solves the problem locally using only a subset of the time of arrival measurements and then fuses the local guesses using averaging consensus techniques. It is shown that the pulse-coupled strategy provides the system with the proper level of synchronization needed to enable accurate localization, even when there exists drift between the internal clocks and the formation is not perfectly maintained. Moreover, the distributed nature of pulse-coupled synchronization allows coordinated synchronization and distributed localization over an infrastructure-free ad-hoc network.
2012 15th International Conference on Information Fusion, 2012
In this paper several methods and models for improving small arms localization are investigated. ... more In this paper several methods and models for improving small arms localization are investigated. Each acoustic sensor is placed at a disparate location and it is assumed that each system may or may not return an estimated range and/or azimuth shooter. Various simple geometric based data fusion methods are proposed and their performance evaluated. Models of localization errors are also proposed and these models are used herein to develop a maximum likelihood approach to data fusion. The parameters of these statistical distributions are estimated from real world data. Comparing / contrasting the results of both methods side by side, it can be shown that while the maximum likelihood based approach performs the best, decent results can be achieved with the simpler geometric based approach.
Independent Component Analyses, Wavelets, Unsupervised Smart Sensors, and Neural Networks IV, 2006
Feature extraction methods based on the discrete wavelet transform (DWT) and multiresolution anal... more Feature extraction methods based on the discrete wavelet transform (DWT) and multiresolution analysis are used to develop a robust classification algorithm that reliably discriminates between launch and impact artillery and/or mortar events via acoustic signals produced during detonation. Distinct characteristics are found within the acoustic signatures since impact events emphasize concussive and shrapnel effects, while launch events are similar to explosions, designed to expel and propel artillery round from a gun. The ensuing blast waves are readily characterized by variations in the corresponding peak pressure and rise time of the waveform, differences in the ratio of positive pressure amplitude to the negative amplitude, variations in the prominent frequencies associated with the varying blast events and variations in the overall duration of the resulting waveform. Unique attributes can also be identified that depend upon the properties of the gun tube, projectile speed at the muzzle, and the explosive/concussive properties associated with the events. In this work, the discrete wavelet transform is used to extract the predominant components and distinct characteristics from the aforementioned acoustic signatures at ranges exceeding 1km. The resulting time-frequency decomposition of the acoustic transient signals is used to produce a separable feature space representation. Highly reliable classification is achieved with a feedforward neural network classifier trained on a sample space derived from the distribution of wavelet coefficients and higher frequency details found within different levels of the multiresolution decomposition. The neural network developed herein provides a capability to classify events (as either launch (LA) or impact (IM)) with a high level of reliability.
Sensors, and Command, Control, Communications, and Intelligence (C3I) Technologies for Homeland Security and Homeland Defense V, 2006
Feature extraction methods based on the discrete wavelet transform and multiresolution analysis f... more Feature extraction methods based on the discrete wavelet transform and multiresolution analysis facilitate the development of a robust classification algorithm that reliably discriminates between launch and impact mortar events via acoustic signals produced during these events. Distinct characteristics arise within the different explosive events because impact events emphasize concussive and shrapnel effects, while launch events result from blasts that expel
Unmanned/Unattended Sensors and Sensor Networks II, 2005
Feature extraction methods based on the discrete wavelet transform (DWT) and multiresolution anal... more Feature extraction methods based on the discrete wavelet transform (DWT) and multiresolution analysis are used to develop a robust classification algorithm that reliably discriminates between launch and impact artillery and/or mortar events via acoustic signals produced during detonation. Distinct characteristics are found within the acoustic signatures since impact events emphasize concussive and shrapnel effects, while launch events are similar to explosions, designed to expel and propel artillery round from a gun. The ensuing blast waves are readily characterized by variations in the corresponding peak pressure and rise time of the waveform, differences in the ratio of positive pressure amplitude to the negative amplitude, variations in the prominent frequencies associated with the varying blast events and variations in the overall duration of the resulting waveform. Unique attributes can also be identified that depend upon the properties of the gun tube, projectile speed at the ...
The major problem of online learning or incremental learning is that, target function is frequent... more The major problem of online learning or incremental learning is that, target function is frequently changing over time. This problem is commonly known as concept drift. Concept drift can be is further complicated if the dataset is classimbalanced. There are different learning methods presented so far to handle concept drift like rule-based systems, decision trees, Naive Bayes, support vector machines, instance based learning, ensemble of classifiers, etc. This learning method requires further to combined with methods of drift detection in order to constantly monitor the performance of concept drift, however online changes detection was failed. In literature there are many methods presented for learning from data streams and drift detection, but most of methods failed to achieve speed and accuracy due to data inconsistency. In this project our goal is to present efficient method for online and non-parametric drift detection. This proposed method is based on recently presented Hoeffdings Bounds and HDDM. It handles concept drift regardless of the learning model to monitor the performance metrics measured during the learning process, to trigger drift signals when a significant variation has been detected. The existing system however as Naive Bayes classifier are having limitations, there is no scope to improve accuracy of HDDM. The Propose system will be efficiently provide drift detection method for data stream mining to improve accuracy.
Unattended Ground, Sea, and Air Sensor Technologies and Applications X, 2008
In this paper we discuss a random walk model to characterize the pulse discharge battery process.... more In this paper we discuss a random walk model to characterize the pulse discharge battery process. Several theoretical results are derived including the mean and variance of an unattended battery-driven sensor lifetime. Some numerical results are presented.
spie.org
On-Site Proceedings Abstract Due Date: 3 October 2005 Manuscript Due Date: 23 January 2006 Cooper... more On-Site Proceedings Abstract Due Date: 3 October 2005 Manuscript Due Date: 23 January 2006 Cooperating Organizations [logo] International Neural Network Society (INNS) [logo] IEEE Computational.
: Agenda: Welcome Remarks, Joint Science and Technology, Doing Business with USSOCOM, Reducing th... more : Agenda: Welcome Remarks, Joint Science and Technology, Doing Business with USSOCOM, Reducing the Threat of Nuclear and Radiological Terrorism, Avon M53 Protective Mask for USSOCOM, Chemical Homeland Security Suite (C-HoSS), Radiological Emergency Response, Use of Recombinant Butyrylcholinesterase in Responding to Chemical Weapon Attack, Reliable Discrimination of High Explosive/Chem/Bio/Artillery Using Acoustic IGS, Real-Time First Bite Detection, Polymer Technology for the Lock-Down/Removal of Radiological Contamination, Modeling Tool for Prediction and Mitigation of CBRNE Events, Terrorist Motivations to Employ CBRN Weapons, USAF Counter-Biological Warfare Effort, Keynote Presentation, Question and Answer Session, Responding to Multiple Ebola Attacks: The Need for Coordinated Preparedness, Capture, Contain, Treat and Dispose of Decontamination Runoff on Site, National Guard CBRN Response: Achieving Unity of Effort Between Local/State/Federal, CBRN Detectors for Early Warning of ...
SPIE Proceedings, 2007
Feature extraction methods based on the discrete wavelet transform and multiresolution analysis f... more Feature extraction methods based on the discrete wavelet transform and multiresolution analysis facilitate the development of a robust classification algorithm that reliably discriminates mortar and artillery variants via acoustic signals produced during the launch/impact events. Utilizing acoustic sensors to exploit the sound waveform generated from the blast for the identification of mortar and artillery variants. Distinct characteristics arise within the
Unattended Ground, Sea, and Air Sensor Technologies and Applications XI, 2009
Herein is described the U.S. Army RDECOM-ARDEC's purpose and series of activities conducted a... more Herein is described the U.S. Army RDECOM-ARDEC's purpose and series of activities conducted at the 2008 NATO SET-093 TG-53 experiment and field test. The overall purpose of the field test as stated by SET-093 panel was to provide a baseline test capable of providing relevant scenarios and data regarding a variety of impulsive generated acoustic events. As organized, the field
Chemical, Biological, Radiological, Nuclear, and Explosives (CBRNE) Sensing IX, 2008
Integrating a sensor suite with ability to discriminate potential Chemical/Biological (CB) events... more Integrating a sensor suite with ability to discriminate potential Chemical/Biological (CB) events from high-explosive (HE) events employing an acoustic sensor array with a Time Difference of Arrival (TDOA) algorithm. Developing a cueing mechanism for more power intensive and range limited sensing CB techniques. Enabling the event detection algorithm to locate to a blast event using TDOA further information is provided
2012 IEEE International Symposium on Precision Clock Synchronization for Measurement, Control and Communication Proceedings, 2012
ABSTRACT Detection of events using a network of simple field sensors has gained interest due to i... more ABSTRACT Detection of events using a network of simple field sensors has gained interest due to its low cost and robustness. Sensor networks have been extensively analyzed recently in terms of stability, robustness, and efficiency. Time synchronization has proven to be critical in sensor fusion applications where time of arrival is a decision property, and thus an accurate common time reference is required. In this work, we analyze the dependence on time synchronization of an acoustic event detection system, and we present a bio-inspired synchronization algorithm for wireless sensor networks capable of providing the system with a common time reference to enable accurate detection.
SPIE Proceedings, 2008
ABSTRACT Explosion detection and recognition is a critical capability to provide situational awar... more ABSTRACT Explosion detection and recognition is a critical capability to provide situational awareness to the war-fighters in battlefield. Acoustic sensors are frequently deployed to detect such events and to trigger more expensive sensing/sensor modalities (i.e. radar, laser spectroscope, IR etc.). Acoustic analysis of explosions has been intensively studied to reliably discriminate mortars, artillery, round variations, and type of blast (i.e. chemical/biological or high-explosive). One of the major challenges is high level of noise, which may include non-coherent noise generated from the environmental background and coherent noise induced by possible mobile acoustic sensor platform. In this work, we introduce a new acoustic scene analysis method to effectively enhance explosion classification reliability and reduce the false alarm rate at low SNR and with high coherent noise. The proposed method is based on acoustic signature modeling using Hidden Markov Models (HMMs). Special frequency domain acoustic features characterizing explosions as well as coherent noise are extracted from each signal segment, which forms an observation vector for HMM training and test. Classification is based on a unique model similarity measure between the HMM estimated from the test observations and the trained HMMs. Experimental tests are based on the acoustic explosion dataset from US ARMY ARDEC, and experimental results have demonstrated the effectiveness of the proposed method.
Author (s): Daniel Lee; Mark McClelland; Joseph Schneider; Tsung-Lin Yang; Dan Gallagher; John Wa... more Author (s): Daniel Lee; Mark McClelland; Joseph Schneider; Tsung-Lin Yang; Dan Gallagher; John Wang; Danelle Shah; Nisar Ahmed; Pete Moran; Brandon Jones; Tung-Sing Leung; Aaron Nathan; Hadas Kress-Gazit; Mark Campbell
SPIE Proceedings, 2008
The coordination of a multi-robot system searching for multi targets is challenging under dynamic... more The coordination of a multi-robot system searching for multi targets is challenging under dynamic environment since the multi-robot system demands group coherence (agents need to have the incentive to work together faithfully) and group competence (agents need to know how to work together well). In our previous proposed bio-inspired coordination method, Local Interaction through Virtual Stigmergy (LIVS), one problem is the considerable randomness of the robot movement during coordination, which may lead to more power consumption and longer searching time. To address these issues, an adaptive LIVS (ALIVS) method is proposed in this paper, which not only considers the travel cost and target weight, but also predicting the target/robot ratio and potential robot redundancy with respect to the detected targets. Furthermore, a dynamic weight adjustment is also applied to improve the searching performance. This new method a truly distributed method where each robot makes its own decision based on its local sensing information and the information from its neighbors. Basically, each robot only communicates with its neighbors through a virtual stigmergy mechanism and makes its local movement decision based on a Particle Swarm Optimization (PSO) algorithm. The proposed ALIVS algorithm has been implemented on the embodied robot simulator, Player/Stage, in a searching target. The simulation results demonstrate the efficiency and robustness in a power-efficient manner with the real-world constraints.
Control Engineering Practice, 2017
Time synchronization has proven to be critical in sensor fusion applications where the time of ar... more Time synchronization has proven to be critical in sensor fusion applications where the time of arrival is utilized as a decision variable. Herein, the application of pulse-coupled synchronization to an acoustic event detection system based on a wireless sensor network is presented. The aim of the system is to locate the source of acoustic events utilizing time of arrival measurements for different formations of the sensor network. A distributed localization algorithm is introduced that solves the problem locally using only a subset of the time of arrival measurements and then fuses the local guesses using averaging consensus techniques. It is shown that the pulse-coupled strategy provides the system with the proper level of synchronization needed to enable accurate localization, even when there exists drift between the internal clocks and the formation is not perfectly maintained. Moreover, the distributed nature of pulse-coupled synchronization allows coordinated synchronization and distributed localization over an infrastructure-free ad-hoc network.
2012 15th International Conference on Information Fusion, 2012
In this paper several methods and models for improving small arms localization are investigated. ... more In this paper several methods and models for improving small arms localization are investigated. Each acoustic sensor is placed at a disparate location and it is assumed that each system may or may not return an estimated range and/or azimuth shooter. Various simple geometric based data fusion methods are proposed and their performance evaluated. Models of localization errors are also proposed and these models are used herein to develop a maximum likelihood approach to data fusion. The parameters of these statistical distributions are estimated from real world data. Comparing / contrasting the results of both methods side by side, it can be shown that while the maximum likelihood based approach performs the best, decent results can be achieved with the simpler geometric based approach.
Independent Component Analyses, Wavelets, Unsupervised Smart Sensors, and Neural Networks IV, 2006
Feature extraction methods based on the discrete wavelet transform (DWT) and multiresolution anal... more Feature extraction methods based on the discrete wavelet transform (DWT) and multiresolution analysis are used to develop a robust classification algorithm that reliably discriminates between launch and impact artillery and/or mortar events via acoustic signals produced during detonation. Distinct characteristics are found within the acoustic signatures since impact events emphasize concussive and shrapnel effects, while launch events are similar to explosions, designed to expel and propel artillery round from a gun. The ensuing blast waves are readily characterized by variations in the corresponding peak pressure and rise time of the waveform, differences in the ratio of positive pressure amplitude to the negative amplitude, variations in the prominent frequencies associated with the varying blast events and variations in the overall duration of the resulting waveform. Unique attributes can also be identified that depend upon the properties of the gun tube, projectile speed at the muzzle, and the explosive/concussive properties associated with the events. In this work, the discrete wavelet transform is used to extract the predominant components and distinct characteristics from the aforementioned acoustic signatures at ranges exceeding 1km. The resulting time-frequency decomposition of the acoustic transient signals is used to produce a separable feature space representation. Highly reliable classification is achieved with a feedforward neural network classifier trained on a sample space derived from the distribution of wavelet coefficients and higher frequency details found within different levels of the multiresolution decomposition. The neural network developed herein provides a capability to classify events (as either launch (LA) or impact (IM)) with a high level of reliability.
Sensors, and Command, Control, Communications, and Intelligence (C3I) Technologies for Homeland Security and Homeland Defense V, 2006
Feature extraction methods based on the discrete wavelet transform and multiresolution analysis f... more Feature extraction methods based on the discrete wavelet transform and multiresolution analysis facilitate the development of a robust classification algorithm that reliably discriminates between launch and impact mortar events via acoustic signals produced during these events. Distinct characteristics arise within the different explosive events because impact events emphasize concussive and shrapnel effects, while launch events result from blasts that expel
Unmanned/Unattended Sensors and Sensor Networks II, 2005